Beyond Effectiveness: A Multi-Criteria Framework for Comparing Practical Socio-Technical Interventions

# A Multi-Criteria Framework for Comparing Practical Socio-Technical Interventions
Source: [https://arxiv.org/html/2608.20649](https://arxiv.org/html/2608.20649)
## Beyond Effectiveness: A Multi\-Criteria Framework for Comparing Practical Socio\-Technical InterventionsDOI:[XXXXXXX\.XXXXXXX](https://doi.org/XXXXXXX.XXXXXXX)Conference:Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NYISBN:978\-1\-4503\-XXXX\-X/2018/06CCS:Social and professional topics Computing / technology policyCCS:Human\-centered computing Interaction design

Catherine King,Lynnette Hui Xian NgAffiliation:Carnegie Mellon University,Pittsburgh,USAemail:[lynnetteng@cmu\.edu](mailto:[email protected])andKathleen M\. CarleyAffiliation:Carnegie Mellon University,Pittsburgh,USAemail:[kathleen\.carley@cs\.cmu\.edu](mailto:[email protected])

2018; © , 2018; Received 5 June 2009

###### Abstract\.

Designers and policymakers in sociotechnical domains like content moderation, privacy interfaces, recommender systems and beyond, must choose among a growing menu of proposed interventions, but typically lack a principled basis for comparing them\. Prior work tends to evaluate interventions individually and mostly along the effectiveness criteria, while implementation constraints such as cost, effort and feasibility are often considered separately\. We present a multi\-criteria framework for evaluating sociotechnical interventions\. This framework is instantiated through the case of misinformation, a domain of intense focus for proposed countermeasures\. We surveyN=39N=39researchers on 40 operationalized interventions across five evaluative criteria: political feasibility, effectiveness, user acceptance, cost, and implementation effort\. We find that the interventions that experts judge to be the most effective are not always the most acceptable to the public or the most feasible to implement\. We also discuss how this tension has implications for the design of sociotechnical interventions beyond misinformation, and offer a decision framework for practitioners navigating the trade\-offs of sociotechnical interventions\.

###### Keywords:

socio\-technical interventions, effectiveness, misinformation, expert survey

## 1\.Introduction

Social media platforms and the emerging generative\-AI technology that shapes them are quickly transforming the online information environment, bringing with them various interventions used to address socio\-technical harms\([61](https://arxiv.org/html/2608.20649#bib.bib3);[98](https://arxiv.org/html/2608.20649#bib.bib26);[22](https://arxiv.org/html/2608.20649#bib.bib44)\)\. Yet these interventions, such as those for countering misinformation or detecting fraud, are often developed and treated individually rather than analyzed as a whole using a standardized set of criteria\([27](https://arxiv.org/html/2608.20649#bib.bib2);[41](https://arxiv.org/html/2608.20649#bib.bib13)\)\. Socio\-technical systems research has argued that such interventions cannot be meaningfully evaluated outside the social context in which they enter\([99](https://arxiv.org/html/2608.20649#bib.bib27)\), and value\-sensitive design calls for weighing a technology against multiple, sometimes competing values rather than a single metric\([33](https://arxiv.org/html/2608.20649#bib.bib28)\)\. Drawing on these two threads of research, we aim to establish a comprehensive framework for developing effective and practical countermeasures across the social media landscape\. Researchers, companies, and policymakers can use this framework to evaluate the socio\-technical interventions and consider their trade\-offs\.

We develop a multi\-criteria framework for socio\-technical intervention design through the case of misinformation\. Throughout, we use "intervention" to mean a deliberate countermeasure \(i\.e\., a policy, a platform design, an educational effort\) that is deployed to reduce a socio\-technical harm\. Misinformation is a domain where the evaluation gap is especially visible\. Countermeasures are proliferating faster than the field’s capacity to compare them\([106](https://arxiv.org/html/2608.20649#bib.bib30);[93](https://arxiv.org/html/2608.20649#bib.bib31)\), and platforms tend to implement interventions reactively and inconsistently\([35](https://arxiv.org/html/2608.20649#bib.bib29)\)\. Intervention design decisions typically require trading effectiveness against the practical and normative considerations that determine whether an intervention can or should be be deployed\([42](https://arxiv.org/html/2608.20649#bib.bib7);[95](https://arxiv.org/html/2608.20649#bib.bib9)\)\. However, to the best of our knowledge, no framework exists to support that reasoning\. To develop the framework, we integrate prior research with an expert elicitation survey ofN=39N=39researchers who evaluated 40 interventions across five criteria: political feasibility, effectiveness, user acceptance, cost, and effort level\. Our mainresearch questionis: How can we identify which interventions are both practical and effective, and under what circumstances? More specifically:

- •RQ1: What trade\-offs emerge when misinformation interventions are evaluated across political feasibility, effectiveness, acceptance, cost, and effort?
- •RQ2: Which intervention categories best balance user\-centered and implementation\-centered criteria?
- •RQ3: What design principles can guide practical misinformation intervention design?

The contributions of this work are threefold:

- •First, we introduce a taxonomy of forty operationalized interventions based off a harmonization of literature review\. These interventions are structured around eight categories, consisting of account moderation, content moderation, content distribution, generative AI\-specific measures, content labeling, user\-based measures, media literacy and educational efforts, and other external and institutional measures\.
- •Second, we provide a multi\-criteria expert evaluation of those interventions across five criteria: political feasibility, effectiveness, user acceptance, cost, and implementation effort\. This expert elicitation survey shows that interventions frequently vary in effectiveness, acceptance or political feasibility, and the administrative effort or cost involved\.
- •Third, we translate these judgments into a decision framework that maps practical operating constraints \(budget, political exposure, timeline, public trust\) onto a starting cluster of interventions, and the design investments needed to unlock the next cluster\.

Designers and policymakers often face similar tensions across many sociotechnical domains: how to moderate content\([50](https://arxiv.org/html/2608.20649#bib.bib14)\), how much friction to add to privacy choices\([2](https://arxiv.org/html/2608.20649#bib.bib32)\), or how to regulate dark patterns in interface design\([72](https://arxiv.org/html/2608.20649#bib.bib33)\)\. In each case, the question is not whether an intervention works, but which one to deploy given the contextual constraints\. Here, we study misinformation as a case study to build a tool for this selection problem\. This study provides insights that we discuss on how and which intervention types, scoped to the social media information environment, are most viable across different implementation constraints, and discusses how different stakeholders can use this framework\.

## 2\.Related Work

### 2\.1\.Misinformation interventions and their evaluation

Previous systematic review articles in the domain of misinformation interventions show that interventions are typically characterized by the part of the platform affected or the actor implementing them\.[17](https://arxiv.org/html/2608.20649#bib.bib1)groups interventions into four general categories \(informational, educational, sociopsychological, and institutional\) and evaluates them primarily on efficacy, while suggesting that other dimensions like feasibility, scalability, and durability should also be considered when selecting and designing interventions\.[27](https://arxiv.org/html/2608.20649#bib.bib2)find a mismatch between the countermeasures that are studied versus those that are implemented by platforms\. Specifically, fact\-checking interventions and interventions emphasizing source pre\-bunking or user inoculation are widely studied, while account\-level interventions such as removal and user notifications, and algorithmic interventions such as redirection and changes to monetization policy, remain understudied\. Finally[64](https://arxiv.org/html/2608.20649#bib.bib4)categorizes interventions into three general categories \(nudges, boosts and educational interventions, and refutation strategies\) and organizes implementation through the targeted outcome \(i\.e\. behaviors, competence, and beliefs\) and limitations\.

While these three key review articles anchor the space of misinformation interventions, they stop short of the comparative evaluation that design practice needs\. These review articles find that most prior research in this domain focuses narrowly on effectiveness or user acceptance of the interventions, but rarely both axes, and almost never alongside implementation constraints\. Our framework closes this gap by comparing multiple dimensions of interventions at once to facilitate identifying the appropriate interventions for the context\.

### 2\.2\.Intervention design features and trade\-offs

To design and evaluate interventions based on their features using multiple criteria, we first need to define what those features are\. A well\-defined intervention has a set of descriptive, objective characteristics, as well as potentially changing user perceptions of those characteristics and the intervention as a whole\. An intervention’s objective characteristics refer to the relatively static features that define how it is implemented and by whom\. These features include the implementer \(which organization\(s\) are implementing the intervention\), the targeted phase of the information pipeline \(such as the creation, spread, or belief phases\), the content being addressed, platform design changes, and the policy’s description \(how it is being implemented and communicated\)\([62](https://arxiv.org/html/2608.20649#bib.bib39)\)\. User perception refers to how individuals evaluate the different objective characteristics of interventions and includes factors such as perceived effectiveness, fairness, intrusiveness, transparency, and problem awareness\([42](https://arxiv.org/html/2608.20649#bib.bib7)\)\. Interventions that are viewed as more effective \(benefits outweigh costs\) and are more supported by users \(practical\) are more likely to be implemented by platforms\([40](https://arxiv.org/html/2608.20649#bib.bib111);[61](https://arxiv.org/html/2608.20649#bib.bib3);[68](https://arxiv.org/html/2608.20649#bib.bib112)\)\.

User perceptions are especially critical because they influence the overall level of policy support\. These perceptions affect whether people believe a problem even needs intervention, whether they think the intervention will be effective and worth doing, and whether they trust the proposed implementer to enforce the new policies fairly\. Public policy support often depends on problem awareness and the type of content addressed\([42](https://arxiv.org/html/2608.20649#bib.bib7)\)\. There is a higher level of support for anti\-smoking initiatives compared to alcohol consumption or diet, which is partially explained by the public’s awareness of the negative consequences associated with smoking\([28](https://arxiv.org/html/2608.20649#bib.bib10)\)\. Similarly, in a social media context, individuals have indicated they are more willing to engage in social corrections with close contacts or for seriously harmful content\([58](https://arxiv.org/html/2608.20649#bib.bib40);[102](https://arxiv.org/html/2608.20649#bib.bib83);[60](https://arxiv.org/html/2608.20649#bib.bib37)\)\. Additionally, some individuals may trust certain implementers more than others\. For example, partisanship is associated with lower levels of support for government regulation\([108](https://arxiv.org/html/2608.20649#bib.bib21)\)\.

Educational and messaging efforts could improve public perceptions and support for interventions\. A meta\-analysis of public policy interventions demonstrated that effectively communicating a potential policy’s effectiveness increased overall support by approximately 4%\([94](https://arxiv.org/html/2608.20649#bib.bib12)\)\. Interventions that are more transparently implemented also tend to gather more public support\([42](https://arxiv.org/html/2608.20649#bib.bib7);[50](https://arxiv.org/html/2608.20649#bib.bib14)\)\. Clearly marked and communicated interventions can shape a user’s perception of the effectiveness and intrusiveness of the intervention\([116](https://arxiv.org/html/2608.20649#bib.bib18)\)\. However, more opaque policies may in turn be less exploitable by malicious actors\([50](https://arxiv.org/html/2608.20649#bib.bib14)\), illustrating the trade off with transparency\.

### 2\.3\.Criteria selection for platform design

The field of Human\-Computer Interaction has a substantial body of work on how platform affordances and intervention design shape user information behavior\. Most of this work has independently arrived at the same trade\-off\-centered framing, where there are multiple criteria to balance simultaneously\. Content moderation work treats moderation not as a single optimization target but as a negotiation among competing values of efficiency, fairness, and user trust\([35](https://arxiv.org/html/2608.20649#bib.bib29)\)\. Ethnographic accounts of Reddit’s moderation tool describe this negotiation as part of the platform’s broader sociotechnical system, where automated and human judgment must be jointly accounted for\([112](https://arxiv.org/html/2608.20649#bib.bib35)\)\. Comparable trade\-offs structure privacy interface design, where the amount of friction a nudge introduces is weighed against its protective benefit rather than maximized outright\([2](https://arxiv.org/html/2608.20649#bib.bib32)\)\. Within misinformation specifically, recent reviews converge on effectiveness, feasibility, scalability, and durability as the relevant criteria\([17](https://arxiv.org/html/2608.20649#bib.bib1);[64](https://arxiv.org/html/2608.20649#bib.bib4)\)\. Feasibility refers to how easily an intervention can be implemented and whether it aligns with user expectations and the intervention’s intent\([41](https://arxiv.org/html/2608.20649#bib.bib13)\)\. Scalability and durability, both aspects of effectiveness, address how easily an intervention can be expanded to reach a larger audience and the longevity of its impact, respectively\([47](https://arxiv.org/html/2608.20649#bib.bib15);[70](https://arxiv.org/html/2608.20649#bib.bib16);[86](https://arxiv.org/html/2608.20649#bib.bib17)\)\.

These criteria focus on the technical components or user interactions with platforms, without considering external influences outside the platforms\. For that, we turn to public policy research\. In public policy, when comparing strategies for regulating socio\-technical platforms, research often evaluates proposals using measures like effectiveness, administrative difficulty or burden, cost, and political acceptability\([95](https://arxiv.org/html/2608.20649#bib.bib9);[65](https://arxiv.org/html/2608.20649#bib.bib8);[29](https://arxiv.org/html/2608.20649#bib.bib6)\)\. The public policy perspective on administrative burden relates to the concept of implementation feasibility\. Additionally, political acceptability is a similar concept to user acceptance of an intervention, which has been shown to influence support for that intervention, though it also incorporates external governmental constraints\([95](https://arxiv.org/html/2608.20649#bib.bib9);[61](https://arxiv.org/html/2608.20649#bib.bib3);[40](https://arxiv.org/html/2608.20649#bib.bib111)\)\.

## 3\.The PEACE Framework

In this work, we first define a categorization of intervention types and then develop a decision framework for evaluating intervention quality and balancing trade\-offs\. ThePEACEframework assesses each intervention’sPolitical feasibilityin the current environment, itsEffectivenessin achieving the intended outcome, the level of userAcceptanceit is likely to receive, itsCostor cheapness, and theEffortrequired for implementation\. These five evaluative criteria were drawn from literature, and insights from an expert elicitation survey was used to inform the design of a corresponding decision framework\.

### 3\.1\.Intervention Categorization

To categorize intervention types, we selected and examined all the studies cited by four recent, comprehensive review articles from different fields: the[27](https://arxiv.org/html/2608.20649#bib.bib2)article published in an interdisciplinary journal focused on social sciences, the[17](https://arxiv.org/html/2608.20649#bib.bib1)article highlighting interventions from both the Global North and Global South in a psychology journal, the[3](https://arxiv.org/html/2608.20649#bib.bib41)article published in the proceedings of a prominent HCI conference, and the[64](https://arxiv.org/html/2608.20649#bib.bib4)article in Nature Human Behavior\. We augmented the literature review by pulling additional, relevant articles on socio\-technical interventions identified through a Scopus search using the terms “misinformation” and “disinformation” combined with either “interventions” or “counter”\. We then categorized interventions on social media platforms by the stages of the information life cycle they target\([59](https://arxiv.org/html/2608.20649#bib.bib42)\)\. More details on our categorization and the specific literature referenced can be found in[Appendix A](https://arxiv.org/html/2608.20649#A1)\.

The information lifecycle on social media generally includes three phases: the creation of information\-disseminating accounts and the information itself, the dissemination of information through organic and viral amplification, and the engagement with the information by correcting, contesting, or contextualizing it\([61](https://arxiv.org/html/2608.20649#bib.bib3);[81](https://arxiv.org/html/2608.20649#bib.bib36);[110](https://arxiv.org/html/2608.20649#bib.bib104);[25](https://arxiv.org/html/2608.20649#bib.bib110)\)\. The life cycle lens provides a frame for thinking about where analysts could intervene\. Because the volume and speed of information spread are too fast for journalists and users to respond promptly to incorrect or contested claims, such as with fact\-checks, exploring ways to intervene earlier in the pipeline, like slowing the virality of misinformation, is an important consideration\([25](https://arxiv.org/html/2608.20649#bib.bib110)\)\. Table[1](https://arxiv.org/html/2608.20649#S3.T1)provides a high\-level overview of the intervention categories that will be discussed throughout the rest of the paper\([61](https://arxiv.org/html/2608.20649#bib.bib3)\)\. Refer to Appendix[B](https://arxiv.org/html/2608.20649#A2)for detailed information on the 40 specified intervention implementations\. The appendix outlines how governments, platforms, and institutions can implement these measures in practice\. The specific interventions were operationalized by reviewing the literature and platform policies to identify examples for each intervention type that have been implemented or proposed previously\([1](https://arxiv.org/html/2608.20649#bib.bib19);[95](https://arxiv.org/html/2608.20649#bib.bib9)\)\.

![The information lifecycle and corresponding intervention categories.](https://arxiv.org/html/2608.20649v1/figures/information_lifecycle.png)Figure 1\.The information lifecycle\([61](https://arxiv.org/html/2608.20649#bib.bib3)\)and corresponding intervention categories\.The information lifecycle and corresponding intervention categories\.The creation step is associated with account and content moderation\. The spread step is associated with content moderation, distribution, and generative AI\-specific interventions\. The belief step is associated with content labeling and user\-based measures\. Prevention can affect all the three steps and includes media literacy and education\. Institutional measures are external to the lifecycle\.Table 1\.Categories of Socio\-Technical Interventions, based on the information lifecycle\([61](https://arxiv.org/html/2608.20649#bib.bib3)\)The first six categories \(account moderation, content moderation, content distribution, generative AI\-specific, content labeling, user\-based measures\) are interventions that occur directly on the platforms\. These interventions can be implemented by social media companies, users on those platforms, or imposed on the platforms through government regulation\([95](https://arxiv.org/html/2608.20649#bib.bib9)\)\. These platform\-level interventions fall into two broad mechanisms: algorithmic interventions that govern what accounts and content can be created and distributed \(account moderation, content moderation, content distribution, generative AI\-specific\) and front\-end design choices that govern how content is displayed and the affordances in which it can be engaged with \(content labeling, user\-based measures\)\. More specifically, account moderation strategies target the information creation phase by controlling account and content creation permissions; content moderation, distribution, and generative AI\-assisted strategies target the information spread phase by reducing or removing algorithmic amplification and managing the information sharing process between users; and content labeling and user\-driven strategies occur after the information has been shared to verify or contextualize claims\.

The next two categories, media literacy and external institutional measures, are strategies that occur primarily beyond the platform, and can be implemented by platforms, governments or various civic organizations\. Media literacy and educational initiatives focus on the prevention phase in the information pipeline, teaching students how to identify and discern truth from misinformation and scams\. This includes learning how to recognize manipulation techniques, logical fallacies, and AI\-generated or edited content\. Structural strategies like regulation and data sharing may target any part of the pipeline, and are meant to support a healthier information environment\.

In this study, we specifically examine the case of misinformation spread through social media platforms\. The information lifecycle includes the creation of social media accounts and misinformation, the dissemination of misinformation, and the correction or prevention of false beliefs and actions\. The same lifecycle lens applies to adjacent harms – scams, fraud, coordinated inauthentic behavior, terrorist recruitment etc– that move through creation, spread, and belief on the same platforms\.

### 3\.2\.PEACE Criteria

For this study, five evaluative criteria were selected after reviewing previous comprehensive review articles and policy analyses\([17](https://arxiv.org/html/2608.20649#bib.bib1);[64](https://arxiv.org/html/2608.20649#bib.bib4);[95](https://arxiv.org/html/2608.20649#bib.bib9)\)\. These criteria are:political feasibility,effectiveness,acceptance,cost, andeffort level\. The five main criteria are defined as follows:

- •Political feasibility:The likelihood of implementing an intervention, incorporating the following factors as well as external, political factors, such as stakeholder perspectives, regulatory and legal constraints, and support from relevant officials or organizations\.
- •Effectiveness:The extent to which an intervention successfully targets the creation, spread, or belief phases in the information lifecycle under different circumstances, such as for some users, certain platforms, or particular types of content\.
- •Acceptance:The degree of user support for an intervention, which may be influenced by factors such as its perceived effectiveness, fairness, intrusiveness, transparency, and the level of trust in the implementer\.
- •Cost:The financial burden on the users, platforms, or government entities involved in implementing and maintaining the intervention\.
- •Effort:The level of administrative effort required to implement and maintain the intervention for the users, platforms, or government entities involved\.

The PEACE decision framework provides a balanced overview of each intervention’s overall viability\. The inherent characteristics of an intervention, along with how users perceive it, directly affect these five key evaluation criteria\. More specifically, user perceptions can impact the acceptance of the intervention, thereby affecting public support for the policy, which in turn increases policy compliance and effectiveness\([42](https://arxiv.org/html/2608.20649#bib.bib7)\)\. For example, political feasibility likely affected Meta’s US fact\-checking program in 2025, where despite its effectiveness and broad user acceptance\([91](https://arxiv.org/html/2608.20649#bib.bib22)\), external political pressures led to the program’s termination\([115](https://arxiv.org/html/2608.20649#bib.bib23);[53](https://arxiv.org/html/2608.20649#bib.bib20)\)\.

To provide a comprehensive set of recommendations across the intervention landscape that researchers, companies, and policymakers can use, we conducted an expert elicitation survey based on these five evaluative criteria\. Design and decision research have long relied on structured expert judgment when the outcome data for a given intervention is costly, slow or impossible to obtain at the scale of comparison against alternatives\([83](https://arxiv.org/html/2608.20649#bib.bib34)\)\. Within misinformation research specifically, survey\-based expert elicitation has previously been used to gauge which interventions experts believe should or would be effective, providing direct precedent for soliciting comparative judgments across our criteria, rather than waiting for outcome evidence for multiple interventions individually\([7](https://arxiv.org/html/2608.20649#bib.bib5);[17](https://arxiv.org/html/2608.20649#bib.bib1)\)\.

## 4\.Data and Methods

### 4\.1\.Survey Overview

We conducted an expert elicitation survey to obtain comparative judgments of across the five evaluative criteria of political feasibility, effectiveness, user acceptance, cost, and effort level\. The survey was promoted to participants and invitees of theSUMMIT 2025\[Full summit name, host location, and dates removed for peer review to preserve anonymity\.\]Our use of expert elicitation for survey judgment captures how experts anticipate the impact of interventions, including their effectiveness in combatting misinformation, how the public and users are likely to respond to the interventions, and their expert evaluation of each intervention’s feasibility, cost, and implementation effort\.

The survey data was collected between January 23 and March 24, 2025 from thirty\-nine researchers\. The experts provided their professional assessments of the effectiveness and acceptability of the general intervention categories \([Table 1](https://arxiv.org/html/2608.20649#S3.T1)\) using a 1\-7 Likert scale\. They also each evaluated 12 out of the 40 operationalized interventions \([Table 7](https://arxiv.org/html/2608.20649#A2.T7)\) across the five evaluative metrics using a 1\-5 Likert scale\. The specific interventions assigned to each participant were randomly selected\. Because of the survey’s length, participants were not assigned all 40 interventions\. Instead, 12 were assigned to each participant to ensure about 10 \(or more\) responses per intervention\. Informed consent was obtained from all participants\. See the supplemental file for a complete copy of the survey\.

In the survey, a "1" on the Likert scale indicated low political feasibility, effectiveness, acceptance, cost, and effort; and "5" indicated high\. During the analysis, cost and effort level scores were inverted \(i\.e\., a Likert score of "5" indicated low cost and effort\), ensuring all five metrics ran in the same direction, with higher scores reflecting better results\. The metrics were also renamed tocheapnessandeaseof implementation for clarity in the Results section\. This made it possible to analyze an overall average score for each intervention and conduct a cluster analysis\.

### 4\.2\.Demographics

The demographics of the participants who completed the survey are shown in Table[2](https://arxiv.org/html/2608.20649#S4.T2)\.

Table 2\.Participant Demographics \(N = 39\)CategoryResponseN%GenderWomen1641\.0Men2256\.4Other / prefer not to say12\.6Age18–341538\.535–441230\.845\+1025\.6Prefer not to say25\.1Academic RankFaculty2153\.8PhD Student1230\.8Industry25\.1Postdoc12\.6Other37\.7Primary DisciplineComputational Social Sciences1743\.6Computer Science1025\.6Sociology410\.3Communications and Media Studies410\.3Political Science25\.1Other25\.1

## 5\.Results and Analysis

### 5\.1\.Category\-Level Tradeoffs

To better understand how experts perceive the misinformation interventions landscape, we first asked participants to rate the overall effectiveness and user acceptance of each of the 8 intervention categories among the American public\. Figure[2](https://arxiv.org/html/2608.20649#S5.F2)shows the fraction of participants who believed each category of countermeasures to be effective or acceptable to the public\. Content moderation, account moderation, and media literacy / education are regarded as the most effective, with over 70% of respondents agreeing they would be effective\. However, moderation, especially account moderation, is viewed by experts as one of the least acceptable to the public\.

![Stacked bar plot showing the percentage of responses on the effectiveness and acceptance of the general intervention categories.](https://arxiv.org/html/2608.20649v1/figures/general_eff_acc.png)Figure 2\.Stacked bar plot showing the percentage of responses on the effectiveness and acceptance of the general intervention categories\.Stacked bar plot showing the percentage of responses on the effectiveness and acceptance of the general intervention categories\.The sub\-figure on the left shows the responses for effectiveness, with the percentage of experts who at least somewhat agree that category of interventions would be effective highlighted\. Content moderation, account moderation, and media literacy score above 70\\%, content labeling and distribution above 60\\%, user\-based measures at around 50\\% and genAI at 28\\%\. The sub\-figure on the right shows the responses for acceptance, with the percentage of experts who at least somewhat agree that category of interventions would be acceptable to users highlighted\. Media literacy, user\-based measures, and content labeling are above 70\\%\. Content moderation, distribution, and institutional measures are around 50\\% , and account moderation and genAI are just below 40\\%\.To further investigate the apparent tension between effectiveness and acceptance, Figure[3](https://arxiv.org/html/2608.20649#S5.F3)illustrates the average Likert scores and 95% confidence intervals for the effectiveness and acceptance ratings\. In this figure, we observe that media literacy / education and external institutional measures rise in the effectiveness rankings, bolstered by the proportion of participants who “strongly agree" regarding their effectiveness rather than more weakly agreeing\. Content and account moderation remain at high levels\. While Figure[2](https://arxiv.org/html/2608.20649#S5.F2)shows general agreement or disagreement among experts, this figure incorporates the strength of their opinions by using the Likert scores directly, making the differences in effectiveness and acceptance across some intervention categories more apparent\. Account moderation and user\-based measures exhibit the largest gap between effectiveness and acceptance, with effectiveness exceeding acceptance in account moderation, and acceptance exceeding effectiveness in user\-based measures \(both statistically significant atp<0\.05p<0\.05\)\.

![Average effectiveness and acceptance Likert scores per general intervention category.](https://arxiv.org/html/2608.20649v1/figures/general_error_plot.png)Figure 3\.Average effectiveness and acceptance Likert scores per general intervention category\.Average effectiveness and acceptance Likert scores per general intervention category\.Average effectiveness and acceptance Likert scores per general intervention category\. Account moderation and user\-based measures are the only categories where the 95\\% confidence intervals of the average effectiveness and acceptance Likert scores do not overlap, with account moderation having higher effectiveness scores and user\-based measures having higher acceptance scores\.These results align with a similar survey of experts, conducted by[7](https://arxiv.org/html/2608.20649#bib.bib5), which found that experts believed that media literacy, labeling of false content, and fact\-checking are the most effective interventions\. Each of these interventions received around 70% approval\. Meanwhile, accuracy prompts \(a form of content distribution\), source labeling, and inoculation ranked lower on the list\. These were the only interventions included in that survey in a similarly structured question\([7](https://arxiv.org/html/2608.20649#bib.bib5)\)\.[7](https://arxiv.org/html/2608.20649#bib.bib5)also asked experts if certain types of actionsshouldbe taken, and concluded that platform design changes, algorithmic changes, and general content moderation as the most popular \(over 75% agree\), and penalizing misinformation sharing and shadow banning \(a form of account moderation\) as the least popular\. In fact, shadow banning was the only suggested intervention where more experts disagreed with its implementation than agreed\. Similarly,[17](https://arxiv.org/html/2608.20649#bib.bib1)found in a survey of experts that they recommended implementing educational and institutional interventions \(such as media literacy, platform alterations, and journalist training\) over other types of interventions\. However, the authors note that these interventions tend to be among the least well\-studied or have mixed evidence on actual effectiveness\([17](https://arxiv.org/html/2608.20649#bib.bib1)\)\.

### 5\.2\.Metric Tradeoffs

Beyond effectiveness and acceptance of the high\-level intervention categories, we also asked the experts to rate each individual intervention across five criteria: political feasibility, effectiveness, user acceptance, cheapness \(inverted cost score\), ease of implementation \(inverted effort score\)\. Figure[4](https://arxiv.org/html/2608.20649#S5.F4)presents a heatmap that illustrates the overall average metric scores, averaged by category and ranked from highest to lowest\. Content labeling, user\-based measures, and content distribution topped the list among the five criteria\. Content labeling and user\-based measures scored well on effectiveness, acceptance, and feasibility\. Content distribution was the most balanced category with the most similar average metric values and the only category where all five metrics averaged above a 3 on the Likert scale\. Media literacy and external institutional measures were ranked last, primarily due to their high overall costs \(low affordability\) and required effort level\. Taken together, these results suggest that while some countermeasures are seen as effective, and some are even considered both effective and acceptable by the public, they are not always regarded as practical by experts\. Interventions that require more administrative effort or resources to implement tend to rank lower overall, despite their potential positive impact\.

![A heatmap showing the average score for each metric across general intervention categories.](https://arxiv.org/html/2608.20649v1/figures/all_interventions_heatmap.png)Figure 4\.A heatmap showing the average score for each metric across general intervention categories\.A heatmap showing the average score for each metric across general intervention categories\.A heatmap showing the average score for each metric across general intervention categories\. Most categories perform better on certain criteria than others\. The most well\-balanced category, and the only one that averages above a 3 on the Likert scale for each criterion, is content distribution\.
### 5\.3\.Top Interventions

To identify which individual interventions drive these category\-level tradeoffs, and what features they may have in common, we rank interventions by their overall scores \(average of the five metrics\) in Table[3](https://arxiv.org/html/2608.20649#S5.T3)\. In general, interventions in the content labeling, user\-based measures, and content distribution interventions dominate the top 10\. Only two interventions from moderation categories \(user control and demonetization\) rank in the top 10 overall, and these moderation interventions are among the least restrictive or provide users with a high level of agency compared to other moderation interventions\.

Table 3\.Interventions ranked by adjusted average score\. The top\-ranked intervention in each category is bolded\.Analyzing only the highest overall scores among individual interventions provides a limited view and fails to highlight the trade\-offs among the different metrics\. When diving deeper into the top ten interventions, we find that only three are also in the top ten for effectiveness, five for acceptance, effort, and cost, and seven for political feasibility; the interventions rate differently for different metrics\. Overall, most of the top interventions rank in the top ten for only two or three metrics\. To better analyze the trade\-offs, we used base R’s implementation of the Ward’s D2 clustering method to group interventions based on similar criterion scores\([76](https://arxiv.org/html/2608.20649#bib.bib115);[90](https://arxiv.org/html/2608.20649#bib.bib116)\)\. We identified four clusters that summarize recurring trade\-off patterns\. Table[4](https://arxiv.org/html/2608.20649#S5.T4)presents the average criteria values for each intervention in each cluster:

1. \(1\)Balanced High Performers \(well\-rounded, high performing interventions\) The second largest cluster included 12 interventions\. These were mostly platform\-implemented interventions that scored relatively well across all five metrics\. It included many of the less intrusive or higher agency affording interventions like demonetization of user accounts \(rather than suspensions or shadowbanning\), limits on forwarding or resharing content, and several labeling and user\-based measures like fact\-checking labels and improved user reporting\.
2. \(2\)Effective but Contested \(effective and low\-cost, but less popular interventions\) The largest group of interventions \(15\) consisted of platform\-implemented controls that were rated as being low cost to implement and effective to deploy\. However, they have relatively low user acceptance\. Generally, this category includes many of the more restrictive or intrusive account moderation, content moderation, and content distribution interventions, such as account suspensions, downranking, and banning political ads\.
3. \(3\)Effective but Costly \(effective and acceptable, but resource\-intensive interventions\) This category included 9 interventions and mostly consisted of almost all of the external institutional measures \(such as investing in journalism, pursuing privacy legislation or anti\-trust action\) and the media literacy/educational measures\. These interventions had by far the highest costs and effort associated with implementing and maintaining these programs, and ranked on the lower end of the spectrum for political feasibility relative to other clusters\.
4. \(4\)Easy Starters \(politically feasible and popular, but low\-impact interventions\) The smallest cluster included only 4 interventions, and it included the measures with by far the highest overall user acceptance scores and highest likelihood of implementation \(such as crowdsourcing context labels, regulating deepfakes, and regulating AI in ads\)\. However, in general, they were viewed as less effective than most other measures\.

CategoryInterventionNFeasibilityEffectivenessAcceptanceCheapnessEaseOverallBalanced high performersAccount ModerationDemonetization123\.58 \(1\.31\)4\.00 \(0\.60\)3\.75 \(1\.06\)2\.92 \(1\.51\)2\.92 \(1\.24\)3\.43 \(0\.52\)Content ModerationAlgorithmic Changes113\.64 \(1\.03\)4\.09 \(0\.54\)3\.55 \(1\.04\)3\.00 \(1\.26\)2\.55 \(0\.93\)3\.36 \(0\.54\)Content DistributionAccuracy Prompts83\.38 \(1\.51\)3\.62 \(0\.92\)3\.50 \(1\.07\)2\.88 \(1\.13\)3\.38 \(1\.06\)3\.35 \(0\.75\)Content DistributionLimit Forwarding134\.08 \(0\.86\)3\.69 \(0\.63\)3\.54 \(1\.05\)4\.31 \(0\.85\)3\.92 \(1\.04\)3\.91 \(0\.49\)Content DistributionLimit Resharing94\.00 \(1\.22\)3\.78 \(0\.44\)3\.00 \(1\.22\)4\.56 \(0\.53\)4\.22 \(0\.67\)3\.91 \(0\.45\)Content LabelingFact\-check Labels153\.53 \(1\.06\)3\.87 \(0\.74\)3\.80 \(0\.94\)2\.79 \(1\.12\)2\.21 \(1\.12\)3\.26 \(0\.59\)Content LabelingGovernment Labels54\.00 \(0\.00\)3\.80 \(1\.30\)4\.00 \(0\.00\)4\.40 \(0\.55\)4\.00 \(0\.00\)4\.04 \(0\.26\)Content LabelingAI Disclosure143\.71 \(1\.59\)3\.71 \(0\.99\)4\.07 \(0\.83\)3\.31 \(1\.25\)3\.31 \(1\.18\)3\.64 \(0\.75\)User\-based MeasuresReporting104\.40 \(0\.52\)3\.80 \(0\.79\)4\.00 \(0\.94\)3\.00 \(1\.49\)2\.40 \(1\.07\)3\.52 \(0\.45\)User\-based MeasuresSocial Norms133\.62 \(0\.77\)3\.62 \(0\.96\)3\.83 \(0\.94\)3\.33 \(1\.23\)2\.67 \(1\.37\)3\.40 \(0\.60\)User\-based MeasuresAlter Platform Metrics124\.08 \(0\.79\)3\.92 \(1\.00\)3\.75 \(1\.36\)3\.00 \(1\.41\)2\.50 \(1\.09\)3\.45 \(0\.62\)Institutional MeasuresTargeted Ads133\.38 \(1\.19\)3\.38 \(0\.65\)3\.85 \(0\.80\)3\.08 \(1\.38\)3\.33 \(1\.30\)3\.44 \(0\.64\)Cluster 1 mean3\.783\.773\.723\.383\.123\.56Effective and low\-cost, but less popular interventionsAccount ModerationAccount Suspensions152\.80 \(1\.52\)3\.53 \(0\.99\)2\.86 \(1\.29\)3\.47 \(1\.13\)2\.40 \(0\.83\)3\.02 \(0\.64\)Account ModerationDeplatforming122\.58 \(1\.31\)3\.83 \(0\.83\)2\.92 \(1\.00\)3\.25 \(1\.36\)2\.33 \(1\.07\)2\.98 \(0\.38\)Account ModerationShadow Banning93\.44 \(1\.24\)3\.33 \(1\.32\)2\.89 \(1\.05\)3\.50 \(1\.69\)3\.38 \(1\.51\)3\.27 \(0\.90\)Content ModerationContent Removal93\.11 \(1\.17\)3\.89 \(0\.93\)3\.33 \(0\.87\)3\.78 \(0\.97\)2\.44 \(1\.24\)3\.31 \(0\.66\)Content ModerationDownranking133\.15 \(1\.46\)3\.92 \(0\.76\)3\.23 \(1\.17\)3\.15 \(1\.28\)3\.08 \(1\.04\)3\.31 \(0\.69\)Content ModerationVirality Circuit Breakers123\.58 \(1\.08\)3\.58 \(1\.38\)3\.08 \(0\.90\)3\.00 \(1\.28\)2\.92 \(1\.00\)3\.23 \(0\.65\)Content DistributionRedirection102\.60 \(1\.35\)3\.10 \(0\.99\)3\.10 \(1\.10\)3\.90 \(0\.88\)3\.10 \(1\.10\)3\.16 \(0\.45\)Content DistributionFriction123\.75 \(0\.97\)3\.50 \(1\.00\)3\.33 \(1\.07\)3\.83 \(1\.47\)3\.33 \(1\.15\)3\.55 \(0\.67\)Content DistributionPlatform Alterations103\.60 \(1\.07\)3\.80 \(1\.32\)3\.00 \(0\.94\)3\.10 \(1\.37\)2\.50 \(0\.97\)3\.20 \(0\.65\)Content DistributionBan Political Ads152\.20 \(1\.32\)3\.00 \(0\.93\)3\.21 \(1\.37\)3\.27 \(1\.49\)3\.60 \(1\.30\)3\.05 \(0\.72\)Generative AIGen AI Chatbots183\.39 \(0\.92\)3\.24 \(0\.75\)2\.71 \(0\.92\)2\.65 \(1\.17\)2\.82 \(1\.01\)2\.96 \(0\.59\)Content LabelingWarning Labels123\.42 \(1\.08\)3\.75 \(0\.97\)3\.17 \(1\.03\)3\.17 \(1\.40\)2\.42 \(1\.08\)3\.18 \(0\.67\)Content LabelingSource Labeling93\.11 \(1\.17\)3\.67 \(0\.87\)3\.67 \(0\.87\)3\.56 \(1\.24\)2\.33 \(1\.00\)3\.27 \(0\.66\)Generative AIGen AI Content123\.33 \(0\.65\)3\.00 \(0\.85\)3\.09 \(0\.83\)2\.64 \(0\.81\)2\.73 \(0\.79\)2\.96 \(0\.50\)Institutional MeasuresTaxes / Fines132\.15 \(0\.99\)3\.08 \(1\.26\)3\.92 \(1\.32\)2\.83 \(1\.64\)2\.25 \(1\.48\)2\.88 \(0\.77\)Cluster 2 mean3\.083\.483\.173\.272\.783\.16Effective and acceptable, but resource\-intensiveContent DistributionFact\-Check Ads93\.33 \(1\.00\)3\.44 \(0\.73\)3\.75 \(0\.71\)2\.00 \(0\.93\)2\.00 \(0\.93\)2\.95 \(0\.62\)Media LiteracyDigital Media Literacy93\.78 \(0\.83\)3\.78 \(1\.20\)4\.11 \(0\.93\)2\.00 \(1\.12\)2\.00 \(1\.12\)3\.13 \(0\.51\)Media LiteracyInoculation103\.60 \(0\.84\)3\.40 \(0\.84\)3\.90 \(0\.88\)2\.50 \(0\.97\)2\.10 \(0\.99\)3\.10 \(0\.56\)Institutional MeasuresMedia Support103\.30 \(1\.34\)3\.40 \(1\.07\)3\.78 \(0\.67\)1\.80 \(0\.92\)2\.00 \(0\.94\)2\.83 \(0\.66\)Institutional MeasuresJournalism Support133\.46 \(1\.20\)4\.00 \(0\.71\)3\.77 \(1\.01\)1\.38 \(0\.51\)1\.85 \(0\.80\)2\.89 \(0\.42\)Institutional MeasuresData Sharing113\.00 \(1\.41\)4\.09 \(0\.83\)4\.18 \(0\.75\)2\.73 \(1\.01\)2\.00 \(1\.41\)3\.20 \(0\.86\)Institutional MeasuresGovernment Regulation112\.91 \(1\.14\)3\.64 \(1\.12\)3\.55 \(0\.82\)1\.82 \(0\.60\)1\.55 \(1\.04\)2\.69 \(0\.40\)Institutional MeasuresPrivacy Legislation122\.75 \(1\.48\)3\.25 \(1\.36\)3\.80 \(0\.92\)1\.36 \(0\.50\)1\.09 \(0\.30\)2\.51 \(0\.86\)Institutional MeasuresAnti\-Trust Action102\.20 \(1\.40\)3\.60 \(1\.07\)4\.20 \(0\.63\)1\.40 \(0\.70\)1\.70 \(1\.06\)2\.62 \(0\.55\)Cluster 3 mean3\.153\.623\.891\.891\.812\.88Easy startersContent ModerationUser Control144\.21 \(1\.12\)3\.00 \(1\.18\)4\.36 \(0\.84\)3\.38 \(1\.39\)2\.31 \(0\.95\)3\.47 \(0\.70\)Content DistributionAI in Ads183\.61 \(1\.33\)2\.94 \(1\.00\)3\.89 \(0\.96\)2\.83 \(1\.34\)3\.00 \(1\.24\)3\.26 \(0\.68\)Content ModerationBan Deepfakes114\.00 \(0\.89\)2\.55 \(1\.29\)4\.10 \(0\.99\)3\.00 \(1\.25\)2\.40 \(1\.17\)3\.26 \(0\.68\)Content LabelingCrowdsourcing144\.29 \(0\.61\)3\.21 \(0\.80\)4\.08 \(0\.64\)4\.14 \(1\.17\)3\.07 \(1\.27\)3\.76 \(0\.60\)Cluster 4 mean4\.032\.934\.113\.342\.693\.44Table 4\.Interventions sorted by cluster and category with mean scores and standard deviations \(in parentheses\)\. Cluster mean rows show the average across interventions in that cluster\.

## 6\.Discussion

In this study, 39 experts evaluated misinformation interventions based on five evaluative criteria: political feasibility, effectiveness, acceptance, cost, and effort\. An intervention is never perfect; it is a trade\-off between these evaluative criteria\. The trade\-offs between evaluative criteria for interventions are systematic and not random\. Across both category\-level and individual\-level interventions, the data shows a recurring pattern, where interventions that are restrictive or opaque face acceptance and feasibility barriers regardless of their effectiveness\. Informational and user\-empowering interventions face the opposite problem of high acceptance but uncertain effectiveness\. This finding should inform how platforms build their intervention portfolios, that they should diversify across multiple intervention types rather than maximizing on any single metric\.

The main findings suggest that, like in many other public policy domains, people prefer informational interventions that afford them agency over more intrusive or restrictive ones\([28](https://arxiv.org/html/2608.20649#bib.bib10);[44](https://arxiv.org/html/2608.20649#bib.bib11)\)\. On average, content labeling, user\-based measures, and content distribution interventions scored the highest among expert raters across the five metrics\. Among these top three categories, content labeling interventions primarily target content after it has already spread, focusing on the belief phase of the misinformation pipeline\. Content distribution interventions generally target the spread of misinformation, either addressing the direct sharing component \(such as friction and accuracy prompts\) or the amplification of the content \(by limiting resharing or forwarding\)\. The top\-ranked user\-based measures included improving user reporting, which also usually targets the spread phase\. When determining which interventions to implement, these top categories unfortunately generally target only thespreadandbeliefin misinformation rather than thecreationof networks or content\. An ideal enforcement strategy would be to determine and implement the best interventions across the misinformation pipeline\.

However, two moderation strategies ranked among the top 10 highest\-ranked interventions: altering platform metrics to reward accuracy rather than engagement and using demonetization as an account moderation strategy\. These interventions target the incentives associated with creating misinformation content and can also affect the potential spread and amplification\. Experts in the elicitation survey ranked them among the most effective interventions \(5th and 3rd, respectively\), although they are perceived to have middling user acceptance and require a decent amount of effort or cost to implement\. The literature supports these results, suggesting that reducing incentives to create and share misinformation \(or conversely, increasing either monetary or social incentives to share accurate information\) can considerably improve the quality of content shared and improve user discernment\([54](https://arxiv.org/html/2608.20649#bib.bib25);[38](https://arxiv.org/html/2608.20649#bib.bib24)\)\.

Further, comparing our results with a similar survey of public opinion\([7](https://arxiv.org/html/2608.20649#bib.bib5)\), we note that expert judgment does not necessarily track public perception\. For example, members of the public rate friction\-based interventions, like delaying users from posting content they have not viewed, as very intrusive and among the least effective\([61](https://arxiv.org/html/2608.20649#bib.bib3)\), but experts rate them as moderately effective\. Notably, this expert\-public mismatch is not uniform\. For content moderation, experts and the public converge on the trade\-off structure itself, independently judging that both account and content moderation are effective yet unpopular\. This observation argues for participatory design for sociotechnical interventions, where expert elicitation are not treated as a direct stand in for public input, but user opinions are also folded into the decision process\. This would surface disagreements before deployment and create a more inclusive online space\.

Finally, these sets of interventions go beyond the scope of misinformation\. They target mechanisms of account and content visibility and distribution that recur across socio\-technical systems more broadly\. For example, account and content moderation are primary levers that platforms use to constrain the amplification of undesired narratives by automated and coordinated inauthentic accounts, regardless of whether the content is misinformation or not\([80](https://arxiv.org/html/2608.20649#bib.bib38)\)\.

Consider the case of coordinated inauthentic behavior, where networks of automated and human\-operated accounts amplify desired narratives\([79](https://arxiv.org/html/2608.20649#bib.bib43)\)\. The same five criteria apply, and they reproduce the same tension we find in misinformation\. Account\-level takedowns are effective, but opaque and contested\([80](https://arxiv.org/html/2608.20649#bib.bib38)\), placing them in the low\-acceptance, high\-effectiveness cluster; transparency measures such as behavioral labels are acceptable but effects may vary, echoing the easy\-starter cluster\. The lifecycle gap recurs here too\. Takedowns act only after a network has already amplified content, leaving account creation, the stage at which coordination is the easiest to disrupt, largely unaddressed, because it is difficult to predict if the account will later coordinate\. Generally, the lifecycle lens and the trade\-off structure carry over, though the interventions may need re\-fitting to the domain\.

### 6\.1\.Design Implications

Based on the main findings and the four calculated intervention clusters, we derive design implications and then propose a decision framework for practitioners\.

#### 6\.1\.1\.Intervention Cluster Implications

The four calculated intervention clusters point to specific design implications\. See Section[5\.3](https://arxiv.org/html/2608.20649#S5.SS3)and Table[4](https://arxiv.org/html/2608.20649#S5.T4)for details on each of the four clusters\.

##### \(1\) Balanced High Performers \- Design for user agency where the effectiveness case allows for it\.

Across the category\-level comparison in Figure[4](https://arxiv.org/html/2608.20649#S5.F4)and the intervention\-level analysis \(Table[4](https://arxiv.org/html/2608.20649#S5.T4)\), user agency is a recurring feature of the most acceptable platform interventions in the cluster of well\-balanced high performers\. The interventions that users find most acceptable preserve their ability to interpret, contest, or act on information, rather than removing the choice for them\. Among the top 10 overall interventions \(Table[3](https://arxiv.org/html/2608.20649#S5.T3)\), labeling government sources, crowdsourcing labels, AI disclosure, encouraging user reporting, and allowing users to control their news feeds directly all preserve a high level of information or agency for the user, rather than restrict it\.

##### \(2\) Effective but Contested \- Acceptance should be a necessary design target, not cost\.

Some of the most effective interventions, including many that are low\-cost and have a low administrative burden, are unfortunately not always acceptable to users\. As shown in Figure[2](https://arxiv.org/html/2608.20649#S5.F2), large majorities of our expert participants rated the account and content moderation categories as effective \(74% and 77%, respectively\), but they also show the largest drop in percentage that believe these interventions would be acceptable to the general public \(only 38% and 54%, respectively\)\. This gap is also acknowledged in public policy literature, where acceptance is driven by the perceived fairness, transparency and trust in the implementation, or the implementer, rather than the intervention’s actual effectiveness\([42](https://arxiv.org/html/2608.20649#bib.bib7)\)\. Therefore, implementers should focus on increasing user acceptance by improving trust in the systems and also improving the public messaging\. Prior HCI work also shows that users tend to distrust what they cannot see or appeal\([77](https://arxiv.org/html/2608.20649#bib.bib109);[80](https://arxiv.org/html/2608.20649#bib.bib38)\), so labeling systems should reduce opacity and include recourse and source transparency\.

##### \(3\) Effective but Costly \- Lower the cost of interventions that are already rated as highly effective and acceptable\.

The media literacy and external institutional measures in cluster 3 are rated as both highly effective and among the most acceptable, yet their overall scores on average are the lowest \(see Table[4](https://arxiv.org/html/2608.20649#S5.T4), Figure[4](https://arxiv.org/html/2608.20649#S5.F4)\) because they are often too resource\-intensive to be practiced\. In this case, the trade\-off is not values\-based \(as in moderation\), but the burden of implementation\. This points to a different design lever: reducing delivery cost rather than reducing scope\. Meta\-analytic work on media literacy interventions show that effectiveness is a measure that is sensitive to how and where an intervention is delivered\([47](https://arxiv.org/html/2608.20649#bib.bib15)\), suggesting that institutions without dedicated funding should look to amortize these interventions into existing infrastructure\. For example, literacy prompts could be embedded into product onboarding flows rather than a freestanding curricular\.

##### \(4\) Easy Starters \- Emerging interventions that are straightforward and low\-cost are a good first step\.

Cluster 4 includes many newer interventions, such as regulating AI in social media ads, controlling the use of deepfakes, crowdsourcing fact\-checking labels, and giving users control over their news feeds\. These interventions have less evidence of effectiveness than the more established interventions in clusters 1\-3, and their implementation would likely require creating and deploying new features\. As Table[4](https://arxiv.org/html/2608.20649#S5.T4)shows, these interventions score poorly on effectiveness and ease of implementation but rank highest among all clusters in political feasibility and user acceptance\. This may partly be because these interventions are either high\-agency\-preserving or focus on regulating or controlling AI, a technology which is currently relatively unpopular across demographic groups in the US\([73](https://arxiv.org/html/2608.20649#bib.bib113)\)\. Since they are also low\-cost and relatively simple to deploy, requiring only that platforms choose to do so, they present a low\-risk option that platforms could adopt\. Implementers can quickly adopt them, monitor user responses, and iterate to improve these interventions\. These changes could enhance their future effectiveness and acceptability\. We therefore recommend that practitioners view cluster 4 not as a final set of solutions but as a starting point: low\-cost interventions that can be deployed quickly, are transparent, and empower users\. This can help establish the legitimacy and trust needed between users and platforms to later implement more restrictive yet more effective interventions\.

#### 6\.1\.2\.General Implications

Beyond the cluster\-level patterns, a structural gap emerges when interventions are mapped onto the information lifecycle \(Figure[1](https://arxiv.org/html/2608.20649#S3.F1)\)\. The most viable interventions, i\.e\. those in Cluster 1, are concentrated at the spread and belief phases, leaving the creation phase largely unaddressed\. Of the top 10 interventions overall \(Table[3](https://arxiv.org/html/2608.20649#S5.T3)\), only two \(altering platform metrics, demonetization\), directly target the incentives that drive content creation\. However, both have middling acceptance scores \(both average a 3\.75 on a 5\-pt Likert scale, and rank 19th/ 20th out of 40 interventions on acceptance respectively\)\. This aligns with research showing that misinformation persists because of platform incentive structures that favor engagement over accuracy at the point of content creation\([38](https://arxiv.org/html/2608.20649#bib.bib24)\)\. This is not merely an artifact of what has been studied, instead, it reflects a structural feature of the field: the most mature, acceptable and deployable tools act downstream of the incentive structures that generate the content in the first place, and are unable to address the root cause of content creation\([38](https://arxiv.org/html/2608.20649#bib.bib24)\)\.

This pipeline coverage gap has direct implications for platform architecture\. Closing it will require designing creation\-phase interventions that can occupy the same viable space as the spread and belief phase interventions in Cluster 1 – transparency, agency\-preserving and low\-burden interventions\. For the existing creation\-phase interventions that come the closest \(demonetization and altering platform metrics\), they are already pointing in this direction: these interventions work by restructuring incentives rather than restricting behavior, which likely explains why they are better received than more overtly restrictive creation\-phase measures like account suspension and deplatforming\.

It is critical that all stages of the information lifecycle are addressed with interventions\. Future intervention design should take this as a signal: creation\-phase interventions framed as restructuring incentives, rather than restricting participation, may be more likely to achieve the acceptance needed for sustained deployment\. More broadly, these findings argue against single\-intervention strategies, and that combining interventions across different mechanisms reduces the spread of viral misinformation\([13](https://arxiv.org/html/2608.20649#bib.bib108)\)\. Because the most viable interventions cluster at the specific pipeline stages, and trade\-off against each other on the five criteria, an effective implementation strategy requires a portfolio that deliberately spans the lifecycle\. It should pair high\-acceptance spread and belief\-phase interventions with creation\-phase measures whose acceptance can be incrementally built\. In the next section, we propose a decision framework that is intended to support this kind of portfolio reasoning\.

#### 6\.1\.3\.A Decision Framework

The four empirically derived clusters translate directly into a decision framework for practitioners\. Organizations deploying socio\-technical interventions typically operate under a specific set of constraints \(i\.e\., budget, political exposure, timeline, or public trust\), and the clusters align cleanly with the four constraint profiles\. Table[5](https://arxiv.org/html/2608.20649#S6.T5)maps each cluster to the conditions under which it is the appropriate starting point, with representative interventions and design considerations\. This proposed decision framework should be used as a navigation tool\. It begins with where the practitioner is constrained, which identifies the starting cluster of interventions, then, what design investments are needed to unlock the next cluster\.

A key implication of this framework is that constraint profiles change over time, and the clusters are designed to be used sequentially as well as independently\. Cluster 4 interventions \(that are low\-cost, high user acceptance and politically feasible\), are most appropriate as a first step to build the legitimacy of the interventions, establishing the trust between users and platforms\. This is a precondition for later deploying more effective but more contested interventions that are in Clusters 1 and 2\. Cluster 3 interventions are both highly effective and well\-received but are resource\-intensive, and therefore should be pursued when coalition partners or dedicated funding can amortize the delivery cost\. For example, embedding media literacy prompts into product onboarding flows rather than standalone curricula, is one way to reduce the structural barrier without reducing the intervention scope\.

Critically, no single cluster is sufficient on its own\. The pipeline coverage gap identified in Section[6\.1\.2](https://arxiv.org/html/2608.20649#S6.SS1.SSS2)means that an intervention portfolio needs to balance across multiple clusters\. For example, sole reliance on Cluster 1 interventions will systematically under\-address account and content creation\. Therefore, practitioners should pair any Cluster 1 deployment with at least one creation\-phase measure, most likely from Cluster 2\. They should also invest in the transparency and recourse mechanisms that can shift the measure’s acceptance over time\.

Table 5\.A practitioner decision framework for selecting misinformation interventions, mapped to the four empirically derived clusters in Table[4](https://arxiv.org/html/2608.20649#S5.T4)\.†marks creation\-phase interventions\.

### 6\.2\.Limitations

While this study is among the first expert surveys on misinformation interventions and the only one so far to analyze multiple evaluative metrics at once, several limitations are noted\. First, our sample was limited in both size \(n = 39\) and geographical reach\. The researchers surveyed were based predominantly at aNorthern American academic institution, with researchers from \[\[Host university name withheld\]\]in particular representing a high fraction of respondents\. Second, while the intervention list had forty interventions, respondents only each rated twelve of the interventions to limit the survey length and improve response rates\. This limited the sample size for each specific intervention, although all respondents rated the overall effectiveness and acceptance of the general countermeasures categories\. Future work should consider a larger sample size or a more diverse set of respondents\. We additionally weighted all five evaluative criteria equally, although some are plausibly correlated\. An extension of this work could analyze the relationship between the metrics and how that may affect overall ratings\. Further, there may be metrics not included in this work that may be relevant to practitioners or policymakers that could be valuable to investigate in future research\. Finally, this survey captures expert judgment only, and not public perception of interventions\. While our discussion compares our expert ratings against companion user findings from literature, that comparison spans separate data collection efforts\. Future work should expand the survey to the general public for a paired measurement\.

## 7\.Conclusion

This work set out to give designers and policymakers a principled basis for comparing sociotechnical interventions, rather than evaluating them one at a time on effectiveness alone\. We defined eight categories of interventions and forty operationalized interventions that can be deployed along the information lifecycle\. Instantiating this through the case of misinformation, we surveyedN=39N=39researchers on the forty interventions across five evaluative criteria\. The five criteria are: political feasibility, effectiveness, acceptance, cost, and effort\. Our results find that interventions that experts judge to be the most effective are often not those judged to be most acceptable or feasible\. User\-based measures, content labeling and content distribution measures scored the highest in multiple evaluative metrics\. Additionally, some provably effective interventions, such as content and account moderation techniques, lack user support and acceptance due to their perceived unfairness or intrusiveness\.

An intervention that experts rate highly effective can still fail in deployment if it is unacceptable, too costly or politically infeasible\. The framework we presented is a first step towards a comparative, multi\-criteria evaluation of sociotechnical interventions\. The framework is intended to support principled design decisions by policymakers and platform governance teams through identifying trade\-offs\. This framework also extends naturally beyond misinformation, including to interventions targeting the automated coordinated accounts that platforms moderate independently of content\. The five criteria, the lifecycle lens and the cluster logic separates viable\-but\-limited interventions from effective\-but\-contested ones, providing a repeatable method for designing and evaluating such socio\-technical interventions in different scenarios\.

###### Acknowledgements\.

This work was supported by the Knight Foundation, the Office of Naval Research’s MURI: Persuasion, Identity, & Morality in Social\-Cyber Environments grant N00014\-21\-12749, Carnegie Mellon University’s Graduate small Project Help \(GuSH\), the Center for Computational Analysis of Social and Organizational Systems \(CASOS\), and the Center for Informed Democracy and Social\-cybersecurity \(IDeaS\)\. The views and conclusions contained in this document are those of the authors alone\. The funders have no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript\.

## References

- Accountable Tech \(2024\)Accountable TechDemocracy by Design: Social Media’s Policy Scores\.Technical reportAccountable Tech\(en\-US\)\.External Links:[Link](https://accountabletech.org/research/democracy-by-design-social-medias-policy-scores/)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I2.i4.p1.1),[1st item](https://arxiv.org/html/2608.20649#A1.I3.i1.p1.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p2.1)\.
- Acquistiet al\.\(2017\)A\. Acquisti, I\. Adjerid, R\. Balebako, L\. Brandimarte, L\. F\. Cranor, S\. Komanduri, P\. G\. Leon, N\. Sadeh, F\. Schaub, M\. Sleeper,et al\.Nudges for privacy and security: understanding and assisting users’ choices online\.ACM Computing Surveys \(CSUR\)50\(3\),pp\. 1–41\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p4.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1)\.
- Aghajariet al\.\(2023\)Z\. Aghajari, E\. P\. S\. Baumer, and D\. DiFranzoReviewing Interventions to Address Misinformation: The Need to Expand Our Vision Beyond an Individualistic Focus\.InProceedings of the ACM on Human\-Computer Interaction,CSCW, Vol\.7,pp\. 87:1–87:34\.External Links:[Link](https://doi.org/10.1145/3579520),[Document](https://dx.doi.org/10.1145/3579520)Cited by:[Table 6](https://arxiv.org/html/2608.20649#A1.T6.2.1.3.1.1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p3.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p6.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p1.1)\.
- Aikinet al\.\(2017\)K\. J\. Aikin, B\. G\. Southwell, R\. S\. Paquin, D\. J\. Rupert, A\. C\. O’Donoghue, K\. R\. Betts, and P\. K\. LeeCorrection of misleading information in prescription drug television advertising: The roles of advertisement similarity and time delay\.Research in Social and Administrative Pharmacy13\(2\),pp\. 378–388\.External Links:ISSN 1551\-7411,[Link](https://www.sciencedirect.com/science/article/pii/S1551741116300146),[Document](https://dx.doi.org/10.1016/j.sapharm.2016.04.004)Cited by:[5th item](https://arxiv.org/html/2608.20649#A1.I3.i5.p1.1)\.
- Allenet al\.\(2021\)J\. Allen, A\. A\. Arechar, G\. Pennycook, and D\. G\. RandScaling up fact\-checking using the wisdom of crowds\.Science Advances7\(36\),pp\. eabf4393\(en\)\.External Links:[Link](https://www.science.org/doi/full/10.1126/sciadv.abf4393),[Document](https://dx.doi.org/10.1126/sciadv.abf4393)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I5.i1.p1.1)\.
- Allenet al\.\(2022\)J\. Allen, C\. Martel, and D\. G\. RandBirds of a feather don’t fact\-check each other: Partisanship and the evaluation of news in Twitter’s Birdwatch crowdsourced fact\-checking program\.InCHI Conference on Human Factors in Computing Systems,New Orleans LA USA,pp\. 1–19\(en\)\.External Links:ISBN 978\-1\-4503\-9157\-3,[Link](https://dl.acm.org/doi/10.1145/3491102.3502040),[Document](https://dx.doi.org/10.1145/3491102.3502040)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I5.i1.p1.1),[2nd item](https://arxiv.org/html/2608.20649#A1.I5.i2.p1.1)\.
- Altayet al\.\(2023\)S\. Altay, M\. Berriche, H\. Heuer, J\. Farkas, and S\. RathjeA survey of expert views on misinformation: Definitions, determinants, solutions, and future of the field\.Harvard Kennedy School Misinformation Review4\(en\-US\)\.External Links:[Link](https://misinforeview.hks.harvard.edu/article/a-survey-of-expert-views-on-misinformation-definitions-determinants-solutions-and-future-of-the-field/),[Document](https://dx.doi.org/10.37016/mr-2020-119)Cited by:[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p4.1),[§5\.1](https://arxiv.org/html/2608.20649#S5.SS1.p3.1),[§6](https://arxiv.org/html/2608.20649#S6.p4.1)\.
- Arifet al\.\(2017\)A\. Arif, J\. J\. Robinson, S\. A\. Stanek, E\. S\. Fichet, P\. Townsend, Z\. Worku, and K\. StarbirdA Closer Look at the Self\-Correcting Crowd: Examining Corrections in Online Rumors\.InProceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing,CSCW,New York, NY, USA,pp\. 155–168\.External Links:ISBN 978\-1\-4503\-4335\-0,[Link](https://dl.acm.org/doi/10.1145/2998181.2998294),[Document](https://dx.doi.org/10.1145/2998181.2998294)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I6.i3.p1.1)\.
- Assenmacheret al\.\(2022\)D\. Assenmacher, D\. Weber, M\. Preuss, A\. Calero Valdez, A\. Bradshaw, B\. Ross, S\. Cresci, H\. Trautmann, F\. Neumann, and C\. GrimmeBenchmarking Crisis in Social Media Analytics: A Solution for the Data\-Sharing Problem\.Social Science Computer Review40\(6\),pp\. 1496–1522\.External Links:ISSN 0894\-4393,[Link](https://doi.org/10.1177/08944393211012268),[Document](https://dx.doi.org/10.1177/08944393211012268)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I8.i3.p1.1)\.
- Avramet al\.\(2020\)M\. Avram, N\. Micallef, S\. Patil, and F\. MenczerExposure to social engagement metrics increases vulnerability to misinformation\.Harvard Kennedy School Misinformation Review1\(5\) \(en\-US\)\.External Links:[Link](https://misinforeview.hks.harvard.edu/article/exposure-to-social-engagement-metrics-increases-vulnerability-to-misinformation/),[Document](https://dx.doi.org/10.37016/mr-2020-033)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I6.i1.p1.1)\.
- Badrinathan and Chauchard \(2023\)S\. Badrinathan and S\. Chauchard“I Don’t Think That’s True, Bro\!” Social Corrections of Misinformation in India\.The International Journal of Press/Politics29,pp\. 394–416\(en\)\.External Links:ISSN 1940\-1612,[Document](https://dx.doi.org/10.1177/19401612231158770)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I6.i2.p1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p6.1)\.
- Bagoet al\.\(2020\)B\. Bago, D\. G\. Rand, and G\. PennycookFake news, fast and slow: Deliberation reduces belief in false \(but not true\) news headlines\.Journal of Experimental Psychology: General149\(8\),pp\. 1608–1613\.External Links:ISSN 1939\-2222,[Document](https://dx.doi.org/10.1037/xge0000729)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I3.i2.p1.1)\.
- Bak\-Colemanet al\.\(2022\)J\. B\. Bak\-Coleman, I\. Kennedy, M\. Wack, A\. Beers, J\. S\. Schafer, E\. S\. Spiro, K\. Starbird, and J\. D\. WestCombining interventions to reduce the spread of viral misinformation\.Nature Human Behaviour6\(10\),pp\. 1–9\(en\)\.External Links:ISSN 2397\-3374,[Document](https://dx.doi.org/10.1038/s41562-022-01388-6)Cited by:[§6\.1\.2](https://arxiv.org/html/2608.20649#S6.SS1.SSS2.p3.1)\.
- Barbashin \(2018\)A\. BarbashinImproving the Western Strategy to Combat Kremlin Propaganda and Disinformation\.Technical reportAtlantic Council: Eurasia Center\(en\)\.External Links:[Link](https://www.atlanticcouncil.org/wp-content/uploads/2018/06/Improving_the_Western_Strategy.pdf)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I8.i2.p1.1)\.
- Barrett \(2019\)P\. BarrettTackling Domestic Disinformation: What the Social Media Companies Need to Do\.Technical reportNYU Stern Center for Business and Human Rights\(en\)\.External Links:[Link](https://issuu.com/nyusterncenterforbusinessandhumanri/docs/nyu_domestic_disinformation_digital)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I2.i4.p1.1),[5th item](https://arxiv.org/html/2608.20649#A1.I3.i5.p1.1),[3rd item](https://arxiv.org/html/2608.20649#A1.I8.i3.p1.1)\.
- Bishop and Gray \(2017\)L\. Bishop and D\. GrayEthical Challenges of Publishing and Sharing Social Media Research Data\.InThe Ethics of Online Research,K\. Woodfield \(Ed\.\),Advances in Research Ethics and Integrity, Vol\.2,pp\. 159–187\.Note:https://doi\.org/10\.1108/S2398\-601820180000002007External Links:ISBN 978\-1\-78714\-486\-6 978\-1\-78714\-485\-9Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I8.i3.p1.1)\.
- Blairet al\.\(2024\)R\. A\. Blair, J\. Gottlieb, B\. Nyhan, L\. Paler, P\. Argote, and C\. J\. StainfieldInterventions to counter misinformation: Lessons from the Global North and applications to the Global South\.Current Opinion in Psychology55,pp\. 101732\.External Links:ISSN 2352\-250X,[Link](https://www.sciencedirect.com/science/article/abs/pii/S2352250X2300177X),[Document](https://dx.doi.org/10.1016/j.copsyc.2023.101732)Cited by:[Table 6](https://arxiv.org/html/2608.20649#A1.T6.2.1.4.1.1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p4.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p6.1),[§2\.1](https://arxiv.org/html/2608.20649#S2.SS1.p1.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p1.1),[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p1.1),[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p4.1),[§5\.1](https://arxiv.org/html/2608.20649#S5.SS1.p3.1)\.
- Bode and Vraga \(2015\)L\. Bode and E\. K\. VragaIn Related News, That Was Wrong: The Correction of Misinformation Through Related Stories Functionality in Social Media\.Journal of Communication65\(4\),pp\. 619–638\.External Links:ISSN 0021\-9916,[Document](https://dx.doi.org/10.1111/jcom.12166)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I3.i4.p1.1)\.
- Bode and Vraga \(2018\)L\. Bode and E\. K\. VragaSee Something, Say Something: Correction of Global Health Misinformation on Social Media\.Health Communication33\(9\),pp\. 1131–1140\.External Links:ISSN 1041\-0236,[Document](https://dx.doi.org/10.1080/10410236.2017.1331312)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I2.i4.p1.1),[2nd item](https://arxiv.org/html/2608.20649#A1.I6.i2.p1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p6.1)\.
- Boret al\.\(2020\)A\. Bor, M\. Osmundsen, S\. H\. R\. Rasmussen, A\. Bechmann, and M\. B\. Petersen“Fact\-checking" videos reduce belief in misinformation and improve the quality of news shared on Twitter\.OSF\(en\-us\)\.External Links:[Link](https://osf.io/a7huq),[Document](https://dx.doi.org/10.31234/osf.io/a7huq)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I2.i2.p1.1)\.
- Bradshaw and Neudert \(2021\)S\. Bradshaw and L\. NeudertThe Road Ahead: Mapping Civil Society Responses to Disinformation\.Technical reportNational Endowment for Democracy\.External Links:[Link](https://www.ned.org/wp-content/uploads/2021/01/The-Road-Ahead-Mapping-Civil-Society-Responses-to-Disinformation-Bradshaw-Neudert-Jan-2021-2.pdf)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I8.i2.p1.1),[4th item](https://arxiv.org/html/2608.20649#A1.I8.i4.p1.1),[§A\.8](https://arxiv.org/html/2608.20649#A1.SS8.p1.1)\.
- Carley \(2020\)K\. M\. CarleySocial cybersecurity: an emerging science\.Computational and Mathematical Organization Theory26\(4\),pp\. 365–381\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p1.1)\.
- Chanet al\.\(2017\)M\. S\. Chan, C\. R\. Jones, K\. Hall Jamieson, and D\. AlbarracínDebunking: A Meta\-Analysis of the Psychological Efficacy of Messages Countering Misinformation\.Psychological Science28\(11\),pp\. 1531–1546\(en\)\.External Links:ISSN 0956\-7976,[Document](https://dx.doi.org/10.1177/0956797617714579)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I2.i3.p1.1)\.
- Chiou and Tucker \(2018\)L\. Chiou and C\. TuckerFake News and Advertising on Social Media: A Study of the Anti\-Vaccination Movement\.Working Paper,Working Paper Series,National Bureau of Economic Research\.External Links:[Link](https://www.nber.org/papers/w25223),[Document](https://dx.doi.org/10.3386/w25223)Cited by:[5th item](https://arxiv.org/html/2608.20649#A1.I3.i5.p1.1)\.
- Ciampaglia \(2018\)G\. L\. CiampagliaThe Digital Misinformation Pipeline\.InPositive Learning in the Age of Information: A Blessing or a Curse?,O\. Zlatkin\-Troitschanskaia, G\. Wittum, and A\. Dengel \(Eds\.\),pp\. 413–421\(en\)\.External Links:ISBN 978\-3\-658\-19567\-0,[Link](https://doi.org/10.1007/978-3-658-19567-0_25),[Document](https://dx.doi.org/10.1007/978-3-658-19567-0%5F25)Cited by:[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p2.1)\.
- Costelloet al\.\(2024\)T\. H\. Costello, G\. Pennycook, and D\. G\. RandDurably reducing conspiracy beliefs through dialogues with AI\.Science385\(6714\),pp\. eadq1814\.External Links:[Link](https://www.science.org/doi/10.1126/science.adq1814),[Document](https://dx.doi.org/10.1126/science.adq1814)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I4.i2.p1.1)\.
- Courchesneet al\.\(2021\)L\. Courchesne, J\. Ilhardt, and J\. N\. ShapiroReview of social science research on the impact of countermeasures against influence operations\.Harvard Kennedy School Misinformation Review2\.External Links:[Document](https://dx.doi.org/10.37016/mr-2020-79)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I3.i4.p1.1),[5th item](https://arxiv.org/html/2608.20649#A1.I3.i5.p1.1),[5th item](https://arxiv.org/html/2608.20649#A1.I5.i5.p1.1),[§A\.1](https://arxiv.org/html/2608.20649#A1.SS1.p1.1),[Table 6](https://arxiv.org/html/2608.20649#A1.T6.2.1.2.1.1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p2.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p6.1),[§1](https://arxiv.org/html/2608.20649#S1.p1.1),[§2\.1](https://arxiv.org/html/2608.20649#S2.SS1.p1.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p1.1)\.
- Diepeveenet al\.\(2013\)S\. Diepeveen, T\. Ling, M\. Suhrcke, M\. Roland, and T\. M\. MarteauPublic acceptability of government intervention to change health\-related behaviours: a systematic review and narrative synthesis\.BMC Public Health13\(1\),pp\. 756\.External Links:ISSN 1471\-2458,[Link](https://doi.org/10.1186/1471-2458-13-756),[Document](https://dx.doi.org/10.1186/1471-2458-13-756)Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p2.1),[§6](https://arxiv.org/html/2608.20649#S6.p2.1)\.
- S\. E\. Dudley and J\. Brito \(Eds\.\) \(2012\)S\. E\. Dudley and J\. Brito \(Eds\.\)Regulation: A Primer\.2nd ed edition,Mercatus Center at George Mason University,Arlington, Va\.External Links:ISBN 978\-0\-9836077\-3\-1 978\-0\-9836077\-4\-8Cited by:[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p2.1)\.
- Ecker and Antonio \(2021\)U\. K\. H\. Ecker and L\. M\. AntonioCan you believe it? An investigation into the impact of retraction source credibility on the continued influence effect\.Memory & Cognition49\(4\),pp\. 631–644\(en\)\.External Links:ISSN 1532\-5946,[Link](https://doi.org/10.3758/s13421-020-01129-y),[Document](https://dx.doi.org/10.3758/s13421-020-01129-y)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I6.i3.p1.1)\.
- Eckeret al\.\(2023\)U\. K\. H\. Ecker, J\. A\. Sanderson, P\. McIlhiney, J\. J\. Rowsell, H\. L\. Quekett, G\. D\. Brown, and S\. LewandowskyCombining refutations and social norms increases belief change\.Quarterly Journal of Experimental Psychology \(2006\)76\(6\),pp\. 1275–1297\(eng\)\.External Links:ISSN 1747\-0226,[Link](https://journals.sagepub.com/doi/10.1177/17470218221111750),[Document](https://dx.doi.org/10.1177/17470218221111750)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I6.i1.p1.1)\.
- Epsteinet al\.\(2021\)Z\. Epstein, A\. J\. Berinsky, R\. Cole, A\. Gully, G\. Pennycook, and D\. G\. RandDeveloping an accuracy\-prompt toolkit to reduce COVID\-19 misinformation online\.Harvard Kennedy School Misinformation Review2\(3\) \(en\-US\)\.External Links:[Link](https://misinforeview.hks.harvard.edu/article/developing-an-accuracy-prompt-toolkit-to-reduce-covid-19-misinformation-online/),[Document](https://dx.doi.org/10.37016/mr-2020-71)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I3.i3.p1.1)\.
- Friedman and Hendry \(2019\)B\. Friedman and D\. G\. HendryValue sensitive design: shaping technology with moral imagination\.Mit Press\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p1.1)\.
- Gaoet al\.\(2018\)M\. Gao, Z\. Xiao, K\. Karahalios, and W\. FuTo Label or Not to Label: The Effect of Stance and Credibility Labels on Readers’ Selection and Perception of News Articles\.InProceedings of the ACM on Human\-Computer Interaction,CSCW, Vol\.2,pp\. 55:1–55:16\.External Links:[Link](https://dl.acm.org/doi/10.1145/3274324),[Document](https://dx.doi.org/10.1145/3274324)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I5.i3.p1.1)\.
- Gillespie \(2018\)T\. GillespieCustodians of the internet: platforms, content moderation, and the hidden decisions that shape social media\.Yale University Press\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p2.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1)\.
- Gillespie \(2022\)T\. GillespieDo Not Recommend? Reduction as a Form of Content Moderation\.Social Media \+ Society8\(3\),pp\. 20563051221117552\.External Links:ISSN 2056\-3051,[Link](https://doi.org/10.1177/20563051221117552),[Document](https://dx.doi.org/10.1177/20563051221117552)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I2.i4.p1.1),[6th item](https://arxiv.org/html/2608.20649#A1.I3.i6.p1.1)\.
- Gimpelet al\.\(2021\)H\. Gimpel, S\. Heger, C\. Olenberger, and L\. UtzThe Effectiveness of Social Norms in Fighting Fake News on Social Media\.Journal of Management Information Systems38\(1\),pp\. 196–221\(EN\)\.External Links:ISSN 0742\-1222,[Document](https://dx.doi.org/10.1080/07421222.2021.1870389)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I6.i1.p1.1)\.
- Globiget al\.\(2023\)L\. K\. Globig, N\. Holtz, and T\. SharotChanging the incentive structure of social media platforms to halt the spread of misinformation\.eLife12,pp\. e85767\.External Links:ISSN 2050\-084X,[Link](https://doi.org/10.7554/eLife.85767),[Document](https://dx.doi.org/10.7554/eLife.85767)Cited by:[§6\.1\.2](https://arxiv.org/html/2608.20649#S6.SS1.SSS2.p1.1),[§6](https://arxiv.org/html/2608.20649#S6.p3.1)\.
- Gorwaet al\.\(2020\)R\. Gorwa, R\. Binns, and C\. KatzenbachAlgorithmic content moderation: Technical and political challenges in the automation of platform governance\.Big Data & Society7\(1\),pp\. 2053951719897945\.External Links:ISSN 2053\-9517,[Link](https://doi.org/10.1177/2053951719897945),[Document](https://dx.doi.org/10.1177/2053951719897945)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I2.i4.p1.1)\.
- Gorwa \(2019\)R\. GorwaWhat is platform governance?\.Information, Communication & Society22\(6\),pp\. 854–871\.External Links:ISSN 1369\-118X,[Link](https://doi.org/10.1080/1369118X.2019.1573914),[Document](https://dx.doi.org/10.1080/1369118X.2019.1573914)Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p1.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p2.1)\.
- Greenet al\.\(2022\)Y\. Green, A\. Gully, Y\. Roth, A\. Roy, J\. A\. Tucker, and A\. WanlessEvidence\-Based Misinformation Interventions: Challenges and Opportunities for Measurement and Collaboration\.Technical reportCarnegie Endowment for International Peace\(en\)\.External Links:[Link](https://carnegieendowment.org/2023/01/09/evidence-based-misinformation-interventions-challenges-and-opportunities-for-measurement-and-collaboration-pub-88661)Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p1.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1)\.
- Grelle and Hofmann \(2024\)S\. Grelle and W\. HofmannWhen and Why Do People Accept Public\-Policy Interventions? An Integrative Public\-Policy\-Acceptance Framework\.Perspectives on Psychological Science19\(1\),pp\. 258–279\(en\)\.External Links:ISSN 1745\-6916,[Link](https://doi.org/10.1177/17456916231180580),[Document](https://dx.doi.org/10.1177/17456916231180580)Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p2.1),[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p1.1),[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p2.1),[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p3.1),[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p3.1),[§6\.1\.1](https://arxiv.org/html/2608.20649#S6.SS1.SSS1.Px2.p1.1)\.
- Guesset al\.\(2020\)A\. M\. Guess, M\. Lerner, B\. Lyons, J\. M\. Montgomery, B\. Nyhan, J\. Reifler, and N\. SircarA digital media literacy intervention increases discernment between mainstream and false news in the United States and India\.Proceedings of the National Academy of Sciences117\(27\),pp\. 15536–15545\.External Links:[Document](https://dx.doi.org/10.1073/pnas.1920498117)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I7.i1.p1.1),[§A\.7](https://arxiv.org/html/2608.20649#A1.SS7.p1.1)\.
- Hagmannet al\.\(2018\)D\. Hagmann, M\. Siegrist, and C\. HartmannTaxes, labels, or nudges? Public acceptance of various interventions designed to reduce sugar intake\.Food Policy79,pp\. 156–165\.External Links:ISSN 0306\-9192,[Link](https://www.sciencedirect.com/science/article/pii/S0306919217310096),[Document](https://dx.doi.org/10.1016/j.foodpol.2018.06.008)Cited by:[§6](https://arxiv.org/html/2608.20649#S6.p2.1)\.
- Helmus and Chandra \(2024\)T\. C\. Helmus and B\. ChandraGenerative Artificial Intelligence Threats to Information Integrity and Potential Policy Responses\.Technical reportRAND Corporation\(en\)\.External Links:[Link](https://www.rand.org/pubs/perspectives/PEA3089-1.html)Cited by:[§A\.4](https://arxiv.org/html/2608.20649#A1.SS4.p1.2)\.
- Helmus and Kepe \(2021\)T\. C\. Helmus and M\. KepeA Compendium of Recommendations for Countering Russian and Other State\-Sponsored Propaganda\.Technical reportRAND Corporation\(en\)\.External Links:[Link](https://www.rand.org/pubs/research_reports/RRA894-1.html)Cited by:[5th item](https://arxiv.org/html/2608.20649#A1.I3.i5.p1.1)\.
- Huanget al\.\(2024\)G\. Huang, W\. Jia, and W\. YuMedia Literacy Interventions Improve Resilience to Misinformation: A Meta\-Analytic Investigation of Overall Effect and Moderating Factors\.Communication Research,pp\. 00936502241288103\.External Links:ISSN 0093\-6502,[Link](https://doi.org/10.1177/00936502241288103),[Document](https://dx.doi.org/10.1177/00936502241288103)Cited by:[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1),[§6\.1\.1](https://arxiv.org/html/2608.20649#S6.SS1.SSS1.Px3.p1.1)\.
- Jeonget al\.\(2012\)S\. Jeong, H\. Cho, and Y\. HwangMedia Literacy Interventions: A Meta\-Analytic Review\.The Journal of Communication62\(3\),pp\. 454–472\(eng\)\.External Links:ISSN 0021\-9916,[Document](https://dx.doi.org/10.1111/j.1460-2466.2012.01643.x)Cited by:[§A\.7](https://arxiv.org/html/2608.20649#A1.SS7.p1.1)\.
- Jhaver and Zhang \(2023\)S\. Jhaver and A\. X\. ZhangDo users want platform moderation or individual control? Examining the role of third\-person effects and free speech support in shaping moderation preferences\.New Media & Society27\.External Links:ISSN 1461\-4448,[Link](https://doi.org/10.1177/14614448231217993),[Document](https://dx.doi.org/10.1177/14614448231217993)Cited by:[5th item](https://arxiv.org/html/2608.20649#A1.I2.i5.p1.1)\.
- Jianget al\.\(2023\)J\. A\. Jiang, P\. Nie, J\. R\. Brubaker, and C\. FieslerA Trade\-off\-centered Framework of Content Moderation\.ACM Transactions on Computer\-Human Interaction30\.External Links:ISSN 1073\-0516,[Link](https://dl.acm.org/doi/10.1145/3534929),[Document](https://dx.doi.org/10.1145/3534929)Cited by:[§A\.2](https://arxiv.org/html/2608.20649#A1.SS2.p1.1),[§1](https://arxiv.org/html/2608.20649#S1.p4.1),[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p3.1)\.
- Johnset al\.\(2024\)A\. Johns, F\. Bailo, E\. Booth, and M\. RizoiuLabelling, shadow bans and community resistance: did Meta’s strategy to suppress rather than remove COVID misinformation and conspiracy theory on Facebook slow the spread?\.Media International Australia197\(1\),pp\. 222–241\(en\)\.External Links:ISSN 1329\-878X, 2200\-467X,[Link](https://journals.sagepub.com/doi/10.1177/1329878X241236984),[Document](https://dx.doi.org/10.1177/1329878X241236984)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I1.i2.p1.1)\.
- Jones\-Janget al\.\(2021\)S\. M\. Jones\-Jang, T\. Mortensen, and J\. LiuDoes Media Literacy Help Identification of Fake News? Information Literacy Helps, but Other Literacies Don’t\.American Behavioral Scientist65\(2\),pp\. 371–388\(en\)\.External Links:ISSN 0002\-7642,[Link](https://doi.org/10.1177/0002764219869406),[Document](https://dx.doi.org/10.1177/0002764219869406)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I7.i1.p1.1)\.
- Kaplan \(2025\)J\. KaplanMore Speech and Fewer Mistakes\.\(en\-US\)\.External Links:[Link](https://about.fb.com/news/2025/01/meta-more-speech-fewer-mistakes/)Cited by:[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p3.1)\.
- Kapooret al\.\(2023\)H\. Kapoor, S\. Rezaei, S\. Gurjar, A\. Tagat, D\. George, Y\. Budhwar, and A\. PuthillamDoes incentivization promote sharing “true” content online?\.Harvard Kennedy School Misinformation Review4\(4\) \(en\-US\)\.External Links:[Link](https://misinforeview.hks.harvard.edu/article/does-incentivization-promote-sharing-true-content-online/),[Document](https://dx.doi.org/10.37016/mr-2020-120)Cited by:[§6](https://arxiv.org/html/2608.20649#S6.p3.1)\.
- Katsaroset al\.\(2022\)M\. Katsaros, K\. Yang, and L\. FratamicoReconsidering Tweets: Intervening during Tweet Creation Decreases Offensive Content\.InProceedings of the International AAAI Conference on Web and Social Media,ICWSM, Vol\.16,pp\. 477–487\(en\)\.External Links:[Link](https://ojs.aaai.org/index.php/ICWSM/article/view/19308),[Document](https://dx.doi.org/10.1609/icwsm.v16i1.19308)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I3.i2.p1.1)\.
- Khanet al\.\(2021\)T\. Khan, A\. Michalas, and A\. AkhunzadaFake news outbreak 2021: Can we stop the viral spread?\.Journal of Network and Computer Applications190,pp\. 103112\(en\)\.External Links:ISSN 1084\-8045,[Link](https://www.sciencedirect.com/science/article/pii/S1084804521001326),[Document](https://dx.doi.org/10.1016/j.jnca.2021.103112)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I2.i1.p1.1)\.
- Kimet al\.\(2022\)J\. Kim, C\. McDonald, P\. Meosky, M\. Katsaros, and T\. TylerPromoting Online Civility Through Platform Architecture\.Journal of Online Trust and Safety1\(4\) \(en\)\.External Links:ISSN 2770\-3142,[Link](https://tsjournal.org/index.php/jots/article/view/54),[Document](https://dx.doi.org/10.54501/jots.v1i4.54)Cited by:[6th item](https://arxiv.org/html/2608.20649#A1.I3.i6.p1.1)\.
- King and Carley \(2025\)C\. King and K\. M\. CarleyPromoting social corrections: a media literacy intervention for misinformation on social media\.InInternational Conference on Social Computing, Behavioral\-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation,R\. Thomson, S\. Renshaw, S\. Al\-khateeb, A\. Burger, P\. Park, and A\. A\. Pyke \(Eds\.\),pp\. 223–232\.Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p2.1)\.
- Kinget al\.\(2025a\)C\. King, P\. Carragher, and K\. M\. CarleyMapping the Scientific Literature on Misinformation Interventions: A Bibliometric Review\.Workshop Proceedings of the 19th International AAAI Conference on Web and Social Media2025,pp\. 10\(en\-US\)\.External Links:[Link](https://workshop-proceedings.icwsm.org/abstract.php?id=2025_10),[Document](https://dx.doi.org/10.36190/2025.10)Cited by:[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p1.1)\.
- Kinget al\.\(2025b\)C\. King, S\. C\. Phillips, and K\. M\. CarleyA path forward on online misinformation mitigation based on current user behavior\.Scientific Reports15\(1\),pp\. 9475\.Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p2.1)\.
- Kinget al\.\(2026\)C\. King, S\. C\. Phillips, and K\. M\. CarleyPublic Support for Misinformation Interventions Depends On Perceived Fairness, Effectiveness, and Intrusiveness\.Journal of Online Trust and Safety3\(2\)\.External Links:ISSN 2770\-3142,[Link](https://tsjournal.org/index.php/jots/article/view/267),[Document](https://dx.doi.org/10.54501/jots.v3i2.267)Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p1.1),[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p1.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p2.1),[Figure 1](https://arxiv.org/html/2608.20649#S3.F1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p2.1),[Table 1](https://arxiv.org/html/2608.20649#S3.T1),[§6](https://arxiv.org/html/2608.20649#S6.p4.1)\.
- King \(2025\)C\. KingEffective and practical strategies for combatting misinformation\.PhD Dissertation,Carnegie Mellon University,Pittsburgh, PA, USA\.External Links:[Document](https://dx.doi.org/10.1184/R1/29316179)Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p1.1)\.
- Kirchner and Reuter \(2020\)J\. Kirchner and C\. ReuterCountering Fake News: A Comparison of Possible Solutions Regarding User Acceptance and Effectiveness\.Proceedings of the ACM on Human\-Computer Interaction4\(CSCW2\),pp\. 1–27\(en\)\.External Links:ISSN 2573\-0142,[Link](https://dl.acm.org/doi/10.1145/3415211),[Document](https://dx.doi.org/10.1145/3415211)Cited by:[6th item](https://arxiv.org/html/2608.20649#A1.I3.i6.p1.1)\.
- Kozyrevaet al\.\(2024\)A\. Kozyreva, P\. Lorenz\-Spreen, S\. M\. Herzog, U\. K\. H\. Ecker, S\. Lewandowsky, R\. Hertwig, A\. Ali, J\. Bak\-Coleman, S\. Barzilai, M\. Basol, A\. J\. Berinsky, C\. Betsch, J\. Cook, L\. K\. Fazio, M\. Geers, A\. M\. Guess, H\. Huang, H\. Larreguy, R\. Maertens, F\. Panizza, G\. Pennycook, D\. G\. Rand, S\. Rathje, J\. Reifler, P\. Schmid, M\. D\. Smith, B\. Swire\-Thompson, P\. Szewach, S\. van der Linden, and S\. WineburgToolbox of individual\-level interventions against online misinformation\.Nature Human Behaviour8,pp\. 1044––1052\.External Links:ISSN 2397\-3374,[Link](https://www.nature.com/articles/s41562-024-01881-0),[Document](https://dx.doi.org/10.1038/s41562-024-01881-0)Cited by:[Table 6](https://arxiv.org/html/2608.20649#A1.T6.2.1.5.1.1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p4.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p5.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p6.1),[§2\.1](https://arxiv.org/html/2608.20649#S2.SS1.p1.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p1.1),[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p1.1)\.
- Kraft and Furlong \(2017\)M\. E\. Kraft and S\. R\. FurlongPublic Policy: Politics, Analysis, and Alternatives\.6th edition,CQ Press\(en\)\.External Links:ISBN 978\-1\-5063\-5814\-7Cited by:[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p2.1)\.
- Lewandowskyet al\.\(2020\)S\. Lewandowsky, J\. Cook, U\. Ecker, D\. Albarracín, M\. A\. Amazeen, P\. Kendeou, D\. Lombardi, E\. J\. Newman, G\. Pennycook, E\. Porter, D\. G\. Rand, D\. N\. Rapp, J\. Reifler, J\. Roozenbeek, P\. Schmid, C\. M\. Seifert, G\. M\. Sinatra, B\. Swire\-Thompson, S\. van der Linden, E\. K\. Vraga, T\. J\. Wood, and M\. S\. ZaragozaDebunking Handbook 2020\.Databrary\(en\)\.External Links:[Link](https://sks.to/db2020),[Document](https://dx.doi.org/10.17910/b7.1182)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I2.i3.p1.1),[Appendix A](https://arxiv.org/html/2608.20649#A1.p6.1)\.
- Lewandowsky and van der Linden \(2021\)S\. Lewandowsky and S\. van der LindenCountering Misinformation and Fake News Through Inoculation and Prebunking\.European Review of Social Psychology0\(0\),pp\. 1–38\.External Links:ISSN 1046\-3283,[Document](https://dx.doi.org/10.1080/10463283.2021.1876983)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I7.i2.p1.1)\.
- Liuet al\.\(2022\)Y\. Liu, P\. Yildirim, and Z\. J\. ZhangImplications of Revenue Models and Technology for Content Moderation Strategies\.Marketing Science41\(4\),pp\. 831–847\.External Links:ISSN 0732\-2399,[Link](https://pubsonline.informs.org/doi/abs/10.1287/mksc.2022.1361),[Document](https://dx.doi.org/10.1287/mksc.2022.1361)Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p1.1)\.
- Luet al\.\(2022\)Z\. Lu, P\. Li, W\. Wang, and M\. YinThe Effects of AI\-based Credibility Indicators on the Detection and Spread of Misinformation under Social Influence\.InProceedings of the ACM on Human\-Computer Interaction,CSCW, Vol\.6,pp\. 461:1–461:27\.External Links:[Link](https://dl.acm.org/doi/10.1145/3555562),[Document](https://dx.doi.org/10.1145/3555562)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I2.i1.p1.1)\.
- Maertenset al\.\(2021\)R\. Maertens, J\. Roozenbeek, M\. Basol, and S\. van der LindenLong\-term effectiveness of inoculation against misinformation: Three longitudinal experiments\.Journal of Experimental Psychology: Applied27\(1\),pp\. 1–16\.External Links:ISSN 1939\-2192,[Document](https://dx.doi.org/10.1037/xap0000315)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I7.i3.p1.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1)\.
- Martínez\-Lópezet al\.\(2022\)F\. J\. Martínez\-López, Y\. Li, and S\. M\. YoungSocial Media Monetization and Demonetization: Risks, Challenges, and Potential Solutions\.InSocial Media Monetization: Platforms, Strategic Models and Critical Success Factors,F\. J\. Martínez\-López, Y\. Li, and S\. M\. Young \(Eds\.\),pp\. 185–214\(en\)\.External Links:ISBN 978\-3\-031\-14575\-9,[Link](https://doi.org/10.1007/978-3-031-14575-9_13),[Document](https://dx.doi.org/10.1007/978-3-031-14575-9%5F13)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I1.i3.p1.1)\.
- Mathuret al\.\(2019\)A\. Mathur, G\. Acar, M\. J\. Friedman, E\. Lucherini, J\. Mayer, M\. Chetty, and A\. NarayananDark patterns at scale: findings from a crawl of 11k shopping websites\.Proceedings of the ACM on human\-computer interaction3\(CSCW\),pp\. 1–32\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p4.1)\.
- McClainet al\.\(2025\)C\. McClain, B\. Kennedy, J\. Gottfried, M\. Anderson, and G\. PasquiniHow the U\.S\. Public and AI Experts View Artificial Intelligence\.Technical reportPew Research Center\(en\-US\)\.External Links:[Link](https://www.pewresearch.org/internet/2025/04/03/how-the-us-public-and-ai-experts-view-artificial-intelligence/)Cited by:[§6\.1\.1](https://arxiv.org/html/2608.20649#S6.SS1.SSS1.Px4.p1.1)\.
- Modirrousta\-Galian and Higham \(2023\)A\. Modirrousta\-Galian and P\. A\. HighamGamified inoculation interventions do not improve discrimination between true and fake news: Reanalyzing existing research with receiver operating characteristic analysis\.Journal of Experimental Psychology: General152\(9\),pp\. 2411–2437\.External Links:ISSN 1939\-2222,[Document](https://dx.doi.org/10.1037/xge0001395)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I7.i3.p1.1)\.
- Morrowet al\.\(2022\)G\. Morrow, B\. Swire\-Thompson, J\. M\. Polny, M\. Kopec, and J\. P\. WihbeyThe emerging science of content labeling: Contextualizing social media content moderation\.Journal of the Association for Information Science and Technology73\(10\),pp\. 1365–1386\(en\)\.External Links:ISSN 2330\-1643,[Link](https://onlinelibrary.wiley.com/doi/abs/10.1002/asi.24637),[Document](https://dx.doi.org/10.1002/asi.24637)Cited by:[§A\.5](https://arxiv.org/html/2608.20649#A1.SS5.p1.1)\.
- Murtagh and Legendre \(2014\)F\. Murtagh and P\. LegendreWard’s hierarchical agglomerative clustering method: which algorithms implement ward’s criterion?\.Journal of Classification31\(3\),pp\. 274–295\.Cited by:[§5\.3](https://arxiv.org/html/2608.20649#S5.SS3.p2.1)\.
- Myers West \(2018\)S\. Myers WestCensored, suspended, shadowbanned: user interpretations of content moderation on social media platforms\.New Media & Society20\(11\),pp\. 4366–4383\.Cited by:[§6\.1\.1](https://arxiv.org/html/2608.20649#S6.SS1.SSS1.Px2.p1.1)\.
- Nassetta and Gross \(2020\)J\. Nassetta and K\. GrossState media warning labels can counteract the effects of foreign disinformation\.Harvard Kennedy School Misinformation Review1\(Special Issue on US Elections and Disinformation\) \(en\-US\)\.External Links:[Link](https://misinforeview.hks.harvard.edu/article/state-media-warning-labels-can-counteract-the-effects-of-foreign-misinformation/),[Document](https://dx.doi.org/10.37016/mr-2020-45)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I5.i3.p1.1)\.
- Ng and Carley \(2022\)L\. H\. X\. Ng and K\. M\. CarleyOnline coordination: methods and comparative case studies of coordinated groups across four events in the united states\.InProceedings of the 14th ACM Web Science Conference 2022,pp\. 12–21\.Cited by:[§6](https://arxiv.org/html/2608.20649#S6.p6.1)\.
- Nget al\.\(2026\)L\. H\. X\. Ng, E\. Pan, M\. Yoder, and K\. CarleyFATe of bots: ethical considerations of social bot detection\.ACM Journal on Responsible Computing3\(2\),pp\. 1–35\.Cited by:[§6\.1\.1](https://arxiv.org/html/2608.20649#S6.SS1.SSS1.Px2.p1.1),[§6](https://arxiv.org/html/2608.20649#S6.p5.1),[§6](https://arxiv.org/html/2608.20649#S6.p6.1)\.
- Ng and Taeihagh \(2021\)L\. H\. Ng and A\. TaeihaghHow does fake news spread? understanding pathways of disinformation spread through apis\.Policy & Internet13\(4\),pp\. 560–585\.Cited by:[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p2.1)\.
- Niklewicz \(2017\)K\. NiklewiczWeeding Out Fake News: An Approach to Social Media Regulation\.European View16\(2\),pp\. 335–335\(en\)\.External Links:ISSN 1781\-6858, 1865\-5831,[Document](https://dx.doi.org/10.1007/s12290-017-0468-0)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I6.i4.p1.1),[1st item](https://arxiv.org/html/2608.20649#A1.I8.i1.p1.1)\.
- O’Haganet al\.\(2006\)A\. O’Hagan, C\. E\. Buck, A\. Daneshkhah, J\. R\. Eiser, P\. H\. Garthwaite, D\. J\. Jenkinson, J\. E\. Oakley, and T\. RakowUncertain judgements: eliciting experts’ probabilities\.John Wiley & Sons\.Cited by:[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p4.1)\.
- Papakyriakopoulos and Goodman \(2022\)O\. Papakyriakopoulos and E\. GoodmanThe Impact of Twitter Labels on Misinformation Spread and User Engagement: Lessons from Trump’s Election Tweets\.InProceedings of the ACM Web Conference 2022,WWW ’22,New York, NY, USA,pp\. 2541–2551\.External Links:ISBN 978\-1\-4503\-9096\-5,[Link](https://doi.org/10.1145/3485447.3512126),[Document](https://dx.doi.org/10.1145/3485447.3512126)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I5.i4.p1.1)\.
- Pennycooket al\.\(2021\)G\. Pennycook, Z\. Epstein, M\. Mosleh, A\. A\. Arechar, D\. Eckles, and D\. G\. RandShifting attention to accuracy can reduce misinformation online\.Nature592,pp\. 590––595\(en\)\.External Links:ISSN 1476\-4687,[Link](https://www.nature.com/articles/s41586-021-03344-2),[Document](https://dx.doi.org/10.1038/s41586-021-03344-2)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I3.i3.p1.1)\.
- Porter and Wood \(2021\)E\. Porter and T\. J\. WoodThe global effectiveness of fact\-checking: Evidence from simultaneous experiments in Argentina, Nigeria, South Africa, and the United Kingdom\.Proceedings of the National Academy of Sciences118\(37\),pp\. e2104235118\.External Links:ISSN 0027\-8424, 1091\-6490,[Link](https://pnas.org/doi/full/10.1073/pnas.2104235118),[Document](https://dx.doi.org/10.1073/pnas.2104235118)Cited by:[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1)\.
- Porter \(2020\)J\. PorterWhatsApp says its forwarding limits have cut the spread of viral messages by 70 percent\.\(en\-US\)\.External Links:[Link](https://www.theverge.com/2020/4/27/21238082/whatsapp-forward-message-limits-viral-misinformation-decline)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I3.i1.p1.1)\.
- Porterfield \(2020\)C\. PorterfieldTwitter Begins Asking Users To Actually Read Articles Before Sharing Them\.\(en\)\.External Links:[Link](https://www.forbes.com/sites/carlieporterfield/2020/06/10/twitter-begins-asking-users-to-actually-read-articles-before-sharing-them/)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I3.i2.p1.1)\.
- Povich \(2021\)E\. S\. PovichInternet Ads Are a Popular Tax Target for Both Parties\.\(en\)\.External Links:[Link](https://pew.org/3psT1Rq)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I8.i1.p1.1)\.
- R Core Team \(2025\)R Core TeamR: a language and environment for statistical computing\.R Foundation for Statistical Computing,Vienna, Austria\.External Links:[Link](https://www.r-project.org/)Cited by:[§5\.3](https://arxiv.org/html/2608.20649#S5.SS3.p2.1)\.
- Rand and Martel \(2025\)D\. Rand and C\. MartelAmericans actually do want expert content moderation\.External Links:[Link](https://thehill.com/opinion/technology/5109667-sorry-zuckerberg-americans-actually-do-want-expert-content-moderation/)Cited by:[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p3.1)\.
- Rauchfleisch and Kaiser \(2024\)A\. Rauchfleisch and J\. KaiserThe impact of deplatforming the far right: an analysis of YouTube and BitChute\.Information, Communication & Society27\(7\),pp\. 1478–1496\(en\)\.External Links:ISSN 1369\-118X, 1468\-4462,[Link](https://www.tandfonline.com/doi/full/10.1080/1369118X.2024.2346524),[Document](https://dx.doi.org/10.1080/1369118X.2024.2346524)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I1.i1.p1.1)\.
- Reuteret al\.\(2025\)C\. Reuter, A\. Lee Hughes, and C\. BuntainCombating information warfare: state and trends in user\-centred countermeasures against fake news and misinformation\.Behaviour & Information Technology44\(13\),pp\. 3348–3361\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p2.1)\.
- Reynoldset al\.\(2020\)J\. P\. Reynolds, K\. Stautz, M\. Pilling, S\. van der Linden, and T\. M\. MarteauCommunicating the effectiveness and ineffectiveness of government policies and their impact on public support: a systematic review with meta\-analysis\.Royal Society Open Science7\(1\),pp\. 190522\.External Links:[Link](https://doi.org/10.1098/rsos.190522),[Document](https://dx.doi.org/10.1098/rsos.190522)Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p3.1)\.
- Rochefort \(2020\)A\. RochefortRegulating Social Media Platforms: A Comparative Policy Analysis\.Communication Law and Policy25\(2\),pp\. 225–260\.External Links:ISSN 1081\-1680, 1532\-6926,[Link](https://www.tandfonline.com/doi/full/10.1080/10811680.2020.1735194),[Document](https://dx.doi.org/10.1080/10811680.2020.1735194)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I8.i1.p1.1),[§1](https://arxiv.org/html/2608.20649#S1.p2.1),[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p2.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p2.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p3.1),[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p1.1)\.
- Roozenbeeket al\.\(2020\)J\. Roozenbeek, S\. v\. d\. Linden, and T\. NygrenPrebunking interventions based on “inoculation” theory can reduce susceptibility to misinformation across cultures\.Harvard Kennedy School Misinformation Review1\(2\) \(en\-US\)\.External Links:[Document](https://dx.doi.org/10.37016//mr-2020-008)Cited by:[3rd item](https://arxiv.org/html/2608.20649#A1.I7.i3.p1.1)\.
- Schaffneret al\.\(2024\)B\. Schaffner, A\. N\. Bhagoji, S\. Cheng, J\. Mei, J\. L\. Shen, G\. Wang, M\. Chetty, N\. Feamster, G\. Lakier, and C\. Tan“Community Guidelines Make this the Best Party on the Internet": An In\-Depth Study of Online Platforms’ Content Moderation Policies\.InProceedings of the CHI Conference on Human Factors in Computing Systems,CHI ’24\.External Links:[Link](https://doi.org/10.1145/3613904.3642333),[Document](https://dx.doi.org/10.1145/3613904.3642333),ISBN 9798400703300Cited by:[§A\.2](https://arxiv.org/html/2608.20649#A1.SS2.p1.1)\.
- Schoenebecket al\.\(2023\)S\. Schoenebeck, A\. Batool, G\. Do, S\. Darling, G\. Grill, D\. Wilkinson, M\. Khan, K\. Toyama, and L\. AshwellOnline harassment in majority contexts: examining harms and remedies across countries\.InProceedings of the 2023 CHI conference on human factors in computing systems,pp\. 1–16\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p1.1)\.
- Selbstet al\.\(2019\)A\. D\. Selbst, D\. Boyd, S\. A\. Friedler, S\. Venkatasubramanian, and J\. VertesiFairness and abstraction in sociotechnical systems\.InProceedings of the conference on fairness, accountability, and transparency,pp\. 59–68\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p1.1)\.
- Sharevskiet al\.\(2022\)F\. Sharevski, R\. Alsaadi, P\. Jachim, and E\. PieroniMisinformation warnings: Twitter’s soft moderation effects on COVID\-19 vaccine belief echoes\.Computers & Security114,pp\. 102577\.External Links:ISSN 0167\-4048,[Link](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8675217/),[Document](https://dx.doi.org/10.1016/j.cose.2021.102577)Cited by:[6th item](https://arxiv.org/html/2608.20649#A1.I3.i6.p1.1),[4th item](https://arxiv.org/html/2608.20649#A1.I5.i4.p1.1)\.
- Smith and Seitz \(2019\)C\. N\. Smith and H\. H\. SeitzCorrecting Misinformation About Neuroscience via Social Media\.Science Communication41\(6\),pp\. 790–819\(en\)\.External Links:ISSN 1075\-5470,[Link](https://doi.org/10.1177/1075547019890073),[Document](https://dx.doi.org/10.1177/1075547019890073)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I3.i4.p1.1)\.
- Tandocet al\.\(2020\)E\. C\. Tandoc, D\. Lim, and R\. LingDiffusion of disinformation: How social media users respond to fake news and why\.Journalism21\(3\),pp\. 381–398\(en\)\.External Links:ISSN 1464\-8849,[Document](https://dx.doi.org/10.1177/1464884919868325)Cited by:[5th item](https://arxiv.org/html/2608.20649#A1.I6.i5.p1.1),[§A\.6](https://arxiv.org/html/2608.20649#A1.SS6.p1.1),[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p2.1)\.
- Thomas and Wahedi \(2023\)D\. R\. Thomas and L\. A\. WahediDisrupting hate: The effect of deplatforming hate organizations on their online audience\.Proceedings of the National Academy of Sciences120\(24\),pp\. e2214080120\.External Links:[Link](https://www.pnas.org/doi/10.1073/pnas.2214080120),[Document](https://dx.doi.org/10.1073/pnas.2214080120)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I1.i1.p1.1)\.
- Toff and Mathews \(2021\)B\. Toff and N\. MathewsIs Social Media Killing Local News? An Examination of Engagement and Ownership Patterns in U\.S\. Community News on Facebook\.Digital Journalism0\(0\),pp\. 1–20\.External Links:ISSN 2167\-0811,[Link](https://doi.org/10.1080/21670811.2021.1977668),[Document](https://dx.doi.org/10.1080/21670811.2021.1977668)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I8.i2.p1.1)\.
- Tullyet al\.\(2020\)M\. Tully, E\. K\. Vraga, and L\. BodeDesigning and Testing News Literacy Messages for Social Media\.Mass Communication and Society23\(1\),pp\. 22–46\.External Links:ISSN 1520\-5436,[Link](https://doi.org/10.1080/15205436.2019.1604970),[Document](https://dx.doi.org/10.1080/15205436.2019.1604970)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I7.i1.p1.1)\.
- Varela da Costa and Mira da Silva \(2025\)J\. Varela da Costa and M\. Mira da SilvaCountermeasures against fake news: a delphi study\.Transforming Government: People, Process and Policy19\(2\),pp\. 392–413\.Cited by:[§1](https://arxiv.org/html/2608.20649#S1.p2.1)\.
- Venturaet al\.\(2023\)T\. Ventura, R\. Majumdar, J\. Nagler, and J\. A\. TuckerMisinformation Exposure Beyond Traditional Feeds: Evidence from a WhatsApp Deactivation Experiment in Brazil\.SSRN Scholarly Paper,Rochester, NY\(en\)\.External Links:[Link](https://papers.ssrn.com/abstract=4457400),[Document](https://dx.doi.org/10.2139/ssrn.4457400)Cited by:[6th item](https://arxiv.org/html/2608.20649#A1.I6.i6.p1.1)\.
- Vogels \(2022\)E\. A\. VogelsSupport for more regulation of tech companies has declined in U\.S\., especially among Republicans\.\(en\-US\)\.External Links:[Link](https://www.pewresearch.org/short-reads/2022/05/13/support-for-more-regulation-of-tech-companies-has-declined-in-u-s-especially-among-republicans/)Cited by:[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p2.1)\.
- Walter and Murphy \(2018\)N\. Walter and S\. T\. MurphyHow to unring the bell: A meta\-analytic approach to correction of misinformation\.Communication Monographs85\(3\),pp\. 423–441\.External Links:ISSN 0363\-7751,[Document](https://dx.doi.org/10.1080/03637751.2018.1467564)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I2.i2.p1.1)\.
- Wardle and Derakhshan \(2017\)C\. Wardle and H\. DerakhshanInformation Disorder: Toward an interdisciplinary framework for research and policy making\.Technical reportCouncil of Europe\(en\)\.External Links:[Link](https://rm.coe.int/information-disorder-toward-an-interdisciplinary-framework-for-researc/168076277c)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I8.i4.p1.1),[§3\.1](https://arxiv.org/html/2608.20649#S3.SS1.p2.1)\.
- West \(2018\)S\. M\. WestCensored, suspended, shadowbanned: User interpretations of content moderation on social media platforms\.New Media & Society20\(11\),pp\. 4366–4383\.External Links:ISSN 1461\-4448,[Link](https://doi.org/10.1177/1461444818773059),[Document](https://dx.doi.org/10.1177/1461444818773059)Cited by:[2nd item](https://arxiv.org/html/2608.20649#A1.I1.i2.p1.1),[§A\.1](https://arxiv.org/html/2608.20649#A1.SS1.p1.1)\.
- Wright \(2022\)L\. WrightAutomated platform governance through visibility and scale: on the transformational power of automoderator\.Social Media\+ Society8\(1\),pp\. 20563051221077020\.Cited by:[§2\.3](https://arxiv.org/html/2608.20649#S2.SS3.p1.1)\.
- Yadav \(2020\)K\. YadavCountering Influence Operations: A Review of Policy Proposals Since 2016\.Technical reportCarnegie Endowment for International Peace\(en\)\.External Links:[Link](https://carnegieendowment.org/2020/11/30/countering-influence-operations-review-of-policy-proposals-since-2016-pub-83333)Cited by:[1st item](https://arxiv.org/html/2608.20649#A1.I8.i1.p1.1)\.
- Yadav \(2021\)K\. YadavPlatform Interventions: How Social Media Counters Influence Operations\.Technical reportCarnegie Endowment for International Peace\.External Links:[Link](https://carnegieendowment.org/2021/01/25/platform-interventions-how-social-media-counters-influence-operations-pub-83698)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I3.i4.p1.1),[§A\.5](https://arxiv.org/html/2608.20649#A1.SS5.p1.1)\.
- Zhan \(2025\)M\. ZhanHere’s why Meta ended fact\-checking, according to experts\.External Links:[Link](https://abcnews.go.com/US/why-did-meta-remove-fact-checkers-experts-explain/story?id=117417445)Cited by:[§3\.2](https://arxiv.org/html/2608.20649#S3.SS2.p3.1)\.
- Zhouet al\.\(2024\)H\. Zhou, Y\. Lu, L\. Zhao, B\. Wang, and T\. LiEffective reporting system to encourage users’ reporting behavior in social media platforms: an empirical study based on structural empowerment theory\.Behaviour & Information Technology43\(14\),pp\. 3490–3509\.External Links:ISSN 0144\-929X,[Link](https://doi.org/10.1080/0144929X.2023.2281491),[Document](https://dx.doi.org/10.1080/0144929X.2023.2281491)Cited by:[4th item](https://arxiv.org/html/2608.20649#A1.I6.i4.p1.1),[§2\.2](https://arxiv.org/html/2608.20649#S2.SS2.p3.1)\.

## Appendix AIntervention Categorization

This appendix describes the literature of interventions, in which we used to develop our categorization taxonomy\. We selected four prominent review articles about misinformation interventions, all published in different fields and journals \([3](https://arxiv.org/html/2608.20649#bib.bib41)in an HCI conference,[17](https://arxiv.org/html/2608.20649#bib.bib1)in a psychology venue,[27](https://arxiv.org/html/2608.20649#bib.bib2)in an interdisciplinary social science journal, and[64](https://arxiv.org/html/2608.20649#bib.bib4)in Nature Human Behavior\)\. Table[6](https://arxiv.org/html/2608.20649#A1.T6)summarizes the intervention categories analyzed by each paper in alphabetical order\.

The[27](https://arxiv.org/html/2608.20649#bib.bib2)article classifies types of interventions based on those publicly announced as having been implemented by various social media platforms\. These 10 categories are advertisement policy, content labeling, content / account moderation, content reporting, content distribution / sharing, disinformation disclosure, disinformation literacy, redirection, security / verification, and other\. The article notes that fact\-checking was included in the disinformation disclosure category, so this is categorized as fact\-checking in Table[6](https://arxiv.org/html/2608.20649#A1.T6)\.

The[3](https://arxiv.org/html/2608.20649#bib.bib41)article categorizes interventions primarily based on the driver of the misinformation that each intervention targets: Content, Source, Individual Users, or Community\. For content\-based interventions, they discuss fact\-checking, warning labels, and platform alterations \(reducing the size of misinformation\)\. For source\-based interventions, they discuss source credibility labels and the crowdsourcing of those labels\. For user\-based strategies, they consider media literacy, accuracy prompts, and account removal\. Lastly, for community\-based interventions, they address social norms\.

Table 6\.Categorizations of misinformation interventions in the literature\.The[17](https://arxiv.org/html/2608.20649#bib.bib1)and[64](https://arxiv.org/html/2608.20649#bib.bib4)articles categorize interventions not just in terms of which parts of the platforms are affected, but also what part of the social media misinformation pipeline is targeted \(such as the creation, spread, belief phases as well as prevention \- see Figure[1](https://arxiv.org/html/2608.20649#S3.F1)\)\. More specifically, the[17](https://arxiv.org/html/2608.20649#bib.bib1)article categorizes 11 types of interventions into four main groups: Informational \(targets belief\), Educational \(targets prevention\), Sociopsychological \(targets spread\), and Institutional \(external\)\. Information interventions include inoculation / prebunking, debunking, credibility labels / tags, and contextual labels / tags\. These interventions aim to correct misinformation\. Under educational interventions, they define media literacy\. This category of interventions seeks to prevent belief in misinformation\. Sociopsychological interventions include accuracy prompts, friction, and social norms\. These interventions aim to discourage users from spreading misinformation\. Lastly, they define platform alterations, politician messaging, and journalist training under institutional interventions\.

The[64](https://arxiv.org/html/2608.20649#bib.bib4)\. article classifies nine types of interventions into three main categories\([64](https://arxiv.org/html/2608.20649#bib.bib4)\): Nudges \(targets spread\), Boosts and Educational Interventions \(targets prevention\), and Refutation Strategies \(targets belief\)\. The nudges category includes accuracy prompts, friction, and social norms\. These interventions address sharing behavior by discouraging users from further distributing misinformation\. The boosts/educational interventions category includes inoculation, lateral reading and verification strategies, and media literacy tips\. These interventions aim to increase user competence in evaluating content online and discerning misinformation or untrustworthy content from reputable information\. Lastly, the refutation strategies category includes debunking and rebuttals, warning and fact\-checking labels, and source credibility labels\.

Many of the review articles used a similar categorization of countermeasures; however, there is no common typology\([27](https://arxiv.org/html/2608.20649#bib.bib2);[3](https://arxiv.org/html/2608.20649#bib.bib41);[17](https://arxiv.org/html/2608.20649#bib.bib1);[64](https://arxiv.org/html/2608.20649#bib.bib4)\)\. Many of these defined categories overlap \(e\.g\., lateral reading skills vs\. media literacy\) or are sub\-categories of other categories\. For example, redirection is a form of content distribution\. Similarly, context labels and source credibility labels are types of content labeling\. Additionally, specific categories are absent, particularly user\-led or institutional interventions such as government regulation\. User\-based measures are an often overlooked aspect in the fight against misinformation\. Individual\-level debunking, especially from trusted messengers, is effective in various contexts\([11](https://arxiv.org/html/2608.20649#bib.bib84);[19](https://arxiv.org/html/2608.20649#bib.bib68);[66](https://arxiv.org/html/2608.20649#bib.bib66)\)\. In addition to user reporting and social norms, users can block others or engage in social corrections\. When misinformation is successfully posted on social media, other users serve as the first line of defense since they can flag or debunk it\. While social corrections can be considered a type of debunking or fact\-checking, the social context may be important to distinguish\.

By synthesizing the categorizations used by four review articles, we developed eight general categories of countermeasures, as shown in Table[1](https://arxiv.org/html/2608.20649#S3.T1)\. These categories are primarily classified by the part of the information lifecycle targeted and the part of the platform affected\. The following subsections describe these eight general categories in more detail, as well as the subcategories of intervention types often described in the literature for each general category\.

### A\.1\.Account Moderation

Account moderation refers to platform interventions directed at user accounts\. These interventions include actions such as account suspensions or account removals, limiting account visibility through shadowbanning, or demonetizing accounts, which restricts their ability to monetize content\([27](https://arxiv.org/html/2608.20649#bib.bib2);[111](https://arxiv.org/html/2608.20649#bib.bib74)\)\. Specific interventions include:

- •Account Suspension\- Account suspension involves temporarily suspending or permanently banning accounts that repeatedly violate platform policies, like repeatedly disseminating misinformation\. Deplatforming describes a coordinated effort to remove a high\-risk account across multiple platforms, rather than just on a single platform\([92](https://arxiv.org/html/2608.20649#bib.bib71);[103](https://arxiv.org/html/2608.20649#bib.bib72)\)\.
- •Shadow Banning\- Shadow banning suppresses an account’s reach by limiting its visibility to others, while leaving the account itself still active\. Users are often not notified that their content has been throttled\([51](https://arxiv.org/html/2608.20649#bib.bib73);[111](https://arxiv.org/html/2608.20649#bib.bib74)\)\.
- •Demonetization\- Demonetization refers to platforms removing or limiting an account’s monetization privileges\([71](https://arxiv.org/html/2608.20649#bib.bib75)\)\. Platforms typically apply such restrictions after repeated policy violations\.

### A\.2\.Content Moderation

Content moderation refers to a broad range of interventions regarding the way content is presented, distributed, or displayed on social media\. These interventions range from evaluating the accuracy of information to modifying its visibility or circulation, and may be implemented through human moderators, users or automated systems\([50](https://arxiv.org/html/2608.20649#bib.bib14);[97](https://arxiv.org/html/2608.20649#bib.bib60)\)\. Specific interventions include:

- •Misinformation Detection\- Misinformation detection refers to the use of computational methods to identify content that is potentially false, misleading, or otherwise unreliable\. These systems typically serve as an upstream component of content moderation, identifying content that may be subsequently reviewed, labeled, downranked or removed\([56](https://arxiv.org/html/2608.20649#bib.bib63);[69](https://arxiv.org/html/2608.20649#bib.bib62)\)\.
- •Fact\-checking\- Fact\-checking assesses the accuracy of a claim by comparing it against available evidence\. It can be carried out by experts, journalists, and platforms, and includes both text and video\-based formats\([20](https://arxiv.org/html/2608.20649#bib.bib65);[109](https://arxiv.org/html/2608.20649#bib.bib61)\)\.
- •Debunking\- Debunking goes beyond determining whether a claim is accurate by providing additional evidence, explanations or contextual information to correct misleading claims\. In this sense, debunking can be understood as a form of "narrative intervention" that identifies falsehoods and provides a more coherent and accurate account of the issue\([66](https://arxiv.org/html/2608.20649#bib.bib66);[23](https://arxiv.org/html/2608.20649#bib.bib67)\)\.
- •Algorithmic Moderation\- Algorithmic content moderation refers to the use of automated systems to identify, evaluate and manage content at scale\. Such systems would automatically fact\-check, label, remove, downrank or otherwise reduce the visibility of content that violate policies, like misinformation\([15](https://arxiv.org/html/2608.20649#bib.bib56);[19](https://arxiv.org/html/2608.20649#bib.bib68);[36](https://arxiv.org/html/2608.20649#bib.bib53);[39](https://arxiv.org/html/2608.20649#bib.bib69)\)\. Algorithmic moderation can also function as a virality circuit breaker, temporarily limiting the amplification of rapidly spreading, unverified content, until its accuracy can be assessed\([1](https://arxiv.org/html/2608.20649#bib.bib19)\)\.
- •User Control\- User control shifts some moderation responsibilities from the platforms to the users, allowing users to take a more active role in deciding what appears in their feeds, rather than than leaving that decision entirely to the platform’s algorithms\([49](https://arxiv.org/html/2608.20649#bib.bib70)\)\.

### A\.3\.Content Distribution

Content distribution refers to a broad category of interventions that shape how content circulates, reaches audiences and is encountered on social media platforms\. Rather than directly removing or modifying content, these interventions influence the pathways through which information is spread on social media\. Specific interventions include:

- •Account Limits\- Another type of content distribution is to impose account limits on users to restrict their forwarding capabilities, which caps the number of recipients the message can be forwarded to\([87](https://arxiv.org/html/2608.20649#bib.bib59)\), resharing limits that constrain repeated dissemination, or disable sharing functionality after content has had several rounds of redistribution\([1](https://arxiv.org/html/2608.20649#bib.bib19)\)\.
- •Friction\- Friction introduces additional steps or delays, such as an extra click or a confirmation window, into the process of sharing content so that users can pause and possibly reconsider before sharing\([12](https://arxiv.org/html/2608.20649#bib.bib50);[55](https://arxiv.org/html/2608.20649#bib.bib51)\)\. One example of such friction would be to prevent users from re\-sharing content that they had not opened and prompting them to consider reading the article first instead\([88](https://arxiv.org/html/2608.20649#bib.bib52)\)\.
- •Accuracy Prompts\- Accuracy prompts, sometimes referred to as "nudges," are brief questions or reminders about the importance of accurate information sent to users and intended to bring accuracy to the front of their minds when a user is deciding whether or not to post a piece of content\([85](https://arxiv.org/html/2608.20649#bib.bib48)\)\. One implementation involves asking users to evaluate the accuracy of one or more headlines to reflect the importance of sharing accurate news before they can continue using the platform\([32](https://arxiv.org/html/2608.20649#bib.bib49)\)\.
- •Redirection\- Redirection intercepts a search query on a potentially harmful or dangerous topic and directs the user toward vetted and official information\. In some cases, it may suppress the search results entirely\([18](https://arxiv.org/html/2608.20649#bib.bib45);[101](https://arxiv.org/html/2608.20649#bib.bib46)\)\. Sometimes, platforms may limit access to the requested content altogether\. A common example is surfacing guidelines from a public health organization, such as the CDC or the WHO, above or next to general search results when users search for a topic like COVID\-19 vaccines\. Similarly, in the immediate lead\-up to an election in the United States, many platforms highlight how to find one’s polling place using official government sources\.\([27](https://arxiv.org/html/2608.20649#bib.bib2);[114](https://arxiv.org/html/2608.20649#bib.bib47)\)\.
- •Advertising Policy\- Advertising policies govern how advertisements can be created, distributed, targeted or displayed on social media platforms\([15](https://arxiv.org/html/2608.20649#bib.bib56);[27](https://arxiv.org/html/2608.20649#bib.bib2);[46](https://arxiv.org/html/2608.20649#bib.bib58)\)\. These policies govern the rules that platforms or regulators can impose on advertising content, and they may include restrictions on the type of content that can run and how it must be labeled or displayed\([15](https://arxiv.org/html/2608.20649#bib.bib56);[27](https://arxiv.org/html/2608.20649#bib.bib2);[46](https://arxiv.org/html/2608.20649#bib.bib58)\)\. Examples include banning political ads, prohibiting the use of AI or manipulated content in political or targeted ads, requiring ads to undergo a fact\-checking process before posting, and labeling ads as "paid for"\([4](https://arxiv.org/html/2608.20649#bib.bib114);[24](https://arxiv.org/html/2608.20649#bib.bib64);[27](https://arxiv.org/html/2608.20649#bib.bib2)\)\.
- •Platform Alterations\- Platform alterations refer to interface\- or architecture\-level changes on social media platforms that shape how content is surfaced, sized, or displayed to users, which in turn affects how users interact with it\([63](https://arxiv.org/html/2608.20649#bib.bib54)\)\. Such interventions can reduce the size or visibility of a post\([36](https://arxiv.org/html/2608.20649#bib.bib53);[63](https://arxiv.org/html/2608.20649#bib.bib54);[100](https://arxiv.org/html/2608.20649#bib.bib55)\)or alter the platform interface to restructure how users can share content, such as directing posts toward more specific rather than general groups\([57](https://arxiv.org/html/2608.20649#bib.bib57)\)\.

### A\.4\.Generative AI\-specific

This category is for interventions that explicitly use generative AI tools like chatbots or LLM\-written content to counter misinformation\. Specific interventions include:

- •Generative AI Content\- Using AI\-generated content to create rebuttals against misinformation or to develop educational initiatives\.
- •Generative AI Chatbots\- Deploying conversational AI systems to dialogue with users as a way to reduce their belief in conspiracy theories or false claims\([26](https://arxiv.org/html/2608.20649#bib.bib106)\)\.

There are other generative AI\-adjacent interventions, such as prohibiting the use of AI or manipulated content to produce deepfakes \(Content Moderation\), banning the use of AI in political or targeted advertising \(Content Distribution\), or requiring clear disclosure on any ads that incorporate AI\-generated images, videos, or audio \(Content Labeling\)\([45](https://arxiv.org/html/2608.20649#bib.bib107)\)\. However, these types of interventions already fit into pre\-existing categories\.

### A\.5\.Content Labeling

Content labeling refers to interventions that attach disclosures, warnings or supplementary information to content in order to communicate its accuracy, credibility, source, or broader context\. These interventions provide users with information through labels that help them evaluate the content they encounter on social media\. These labels include fact\-checks \(see Content Moderation section\) or source information or credibility\([75](https://arxiv.org/html/2608.20649#bib.bib76);[114](https://arxiv.org/html/2608.20649#bib.bib47)\)\. Specific interventions include:

- •Crowdsourcing\- Crowdsourcing involves drawing on evaluations from ordinary users to access information and provide labels or contextual information\. It shifts the labeling task away from professionals and onto the user base, then aggregates user judgments into a label or note\([5](https://arxiv.org/html/2608.20649#bib.bib77)\)\. One prominent implementation of this approach is X’s Community Notes program, in which users collaboratively contribute and evaluate contextual notes that attached to the tweets\([6](https://arxiv.org/html/2608.20649#bib.bib78)\)\.
- •Context Labels\- Context labels attach additional contextual information to posts so that users can better interpret or evaluate the content\. These labels provide relevant background, clarification or evidence to the post\. X’s Community Notes are context labels, and the program uses crowdsourcing to create and maintain them\([6](https://arxiv.org/html/2608.20649#bib.bib78)\)\.
- •Source Labels\- Source credibility labels provide users with information about the reliability, identity, or institutional affiliation of the source behind a post\. These interventions may, for example, communicate assessments of a news source’s reliability\([34](https://arxiv.org/html/2608.20649#bib.bib80)\)or identify accounts associated with government officials or state\-run media sources\([78](https://arxiv.org/html/2608.20649#bib.bib81)\)\.
- •Warning Labels\- Warning labels flag posts based on the post’s content or sources that may be false, misleading, or otherwise warrant additional scrutiny\([84](https://arxiv.org/html/2608.20649#bib.bib79)\)\. Some implementations require an active click to bypass an interstitial screen before the user can view or share the underlying content\([100](https://arxiv.org/html/2608.20649#bib.bib55)\)\.
- •Notification\- Notifications directly inform users when content they have been posted, shared or otherwise interacted with has subsequently been flagged to contain harmful or false information, or originating from state\-run media sources\([27](https://arxiv.org/html/2608.20649#bib.bib2)\)\.

### A\.6\.User\-based Measures

User\-based measures refer to interventions that empower individuals and communities to respond directly to misinformation and shape their own information environments\([102](https://arxiv.org/html/2608.20649#bib.bib83)\)\. Specific interventions include:

- •Social Norms\- Social norm interventions leverage community expectations and social signals to encourage more responsible information\-sharing behaviors\. Platforms can support these norms by designing features or metrics that reward accuracy rather than engagement\([31](https://arxiv.org/html/2608.20649#bib.bib86);[37](https://arxiv.org/html/2608.20649#bib.bib87);[10](https://arxiv.org/html/2608.20649#bib.bib88)\)\.
- •Social Corrections\- Social corrections are peer\-to\-peer corrections in which one user directly fact\-checks another’s post, sometimes occurring in the open \(via a reply\) and other times occurring privately \(via a direct message\)\([11](https://arxiv.org/html/2608.20649#bib.bib84);[19](https://arxiv.org/html/2608.20649#bib.bib68)\)\.
- •Retractions\- Retractions occur when users or organizations delete or correct a false post that was previously published\. It may also involve efforts to notify users who previously viewed the post\.\([8](https://arxiv.org/html/2608.20649#bib.bib89);[30](https://arxiv.org/html/2608.20649#bib.bib90)\)\.
- •User Reporting\- Reporting enables users to flag potentially harmful or misleading posts or accounts for platform review\([82](https://arxiv.org/html/2608.20649#bib.bib82);[116](https://arxiv.org/html/2608.20649#bib.bib18)\)\.
- •Blocking\- Blocking gives users a way to reduce their exposure to unwanted content, such as content from specific users or on specific topics\([102](https://arxiv.org/html/2608.20649#bib.bib83)\)\.
- •Other/User Empowerment\- Individualized measures provide users with tools or interventions that encourage self\-moderation and behavioral change, such as nudges which will encourage users to reduce their social media use\([107](https://arxiv.org/html/2608.20649#bib.bib85)\)\.

### A\.7\.Media Literacy and Education

Media literacy and education encompasses interventions designed to strengthen individuals’ ability to critically evaluate and engage with information\. These approaches seek to improve the public’s ability to navigate digital sources and to think critically about the media they encounter\([43](https://arxiv.org/html/2608.20649#bib.bib92);[48](https://arxiv.org/html/2608.20649#bib.bib93)\)\. Specific interventions include:

- •Media Literacy\- Media literacy and related educational interventions include a range of initiatives intended to develop or assess media literacy skills\. These efforts may provide practical guidance for identifying misinformation\([43](https://arxiv.org/html/2608.20649#bib.bib92)\), evaluate individuals’ information, digital, or news literacy\([52](https://arxiv.org/html/2608.20649#bib.bib96)\), or develop and test communication strategies for improving the effectiveness of media literacy education\([105](https://arxiv.org/html/2608.20649#bib.bib97)\)\.
- •Inoculation\- Inoculation, sometimes called "pre\-bunking" in the literature, aims to build resistance to false information before exposure occurs\. These interventions typically warn individuals about common misleading arguments or manipulation techniques, and explain how they operate, making users better prepared to recognize and resist subsequent misinformation\([67](https://arxiv.org/html/2608.20649#bib.bib91)\)\.
- •Games\- Games like fake news games are a specific type of educational tool that use interactive, gamified experiences to teach users how to recognize misinformation and strengthen their critical evaluation skills\. These games often expose players to common misinformation strategies in a controlled environment, helping them identify similar techniques in future encounters\([70](https://arxiv.org/html/2608.20649#bib.bib16);[74](https://arxiv.org/html/2608.20649#bib.bib95);[96](https://arxiv.org/html/2608.20649#bib.bib94)\)\.

### A\.8\.Institutional Measures

Institutional measures refer to interventions undertaken by governments, civil society organizations, news media, and other institutions\([21](https://arxiv.org/html/2608.20649#bib.bib98)\)\. Specific interventions include:

- •Government Regulation\- Government regulation encompasses laws and regulations designed to govern platforms and increase their accountability\. These measures can be implemented at the local, state, or federal level\([82](https://arxiv.org/html/2608.20649#bib.bib82);[95](https://arxiv.org/html/2608.20649#bib.bib9);[113](https://arxiv.org/html/2608.20649#bib.bib105)\)\. Examples include modifying Section 230, developing comprehensive privacy laws similar to Europe’s GDPR, taking anti\-trust action to break up monopolistic technology companies\([95](https://arxiv.org/html/2608.20649#bib.bib9)\), or restriction of micro\-targeted advertising\([89](https://arxiv.org/html/2608.20649#bib.bib102)\)\.
- •Media Support\- Media support involves efforts to strengthen the broader news ecosystem, including funding local media outlets, giving credible local reporting greater visibility on social media platforms\([104](https://arxiv.org/html/2608.20649#bib.bib99)\), and providing training and resources to support high\-quality reporting\([14](https://arxiv.org/html/2608.20649#bib.bib100);[21](https://arxiv.org/html/2608.20649#bib.bib98)\)\.
- •Data Sharing\- Data sharing involves social media platforms sharing raw data and internal findings with researchers, thereby supporting greater transparency and independent investigation\([9](https://arxiv.org/html/2608.20649#bib.bib103);[16](https://arxiv.org/html/2608.20649#bib.bib101);[15](https://arxiv.org/html/2608.20649#bib.bib56)\)\.
- •Other Institutional Measures\- Other institutional measures include developing resources and tools to support civil society and strengthening coordination and collaboration among governments, researchers, media organizations, platforms and other relevant institutions\([21](https://arxiv.org/html/2608.20649#bib.bib98);[110](https://arxiv.org/html/2608.20649#bib.bib104)\)\.

## Appendix BOperationalized Interventions

This section describes the 40 operationalized interventions participants were asked about in the expert survey\. Each intervention is a specific implementation of one or more interventions described in the literature and in the previous appendix\. On the survey, participants were given the definitions of the specific intervention implementations, as shown in the first column of the following table\.

Specified Intervention ImplementationsIntervention Type\(s\)Account ModerationAccount Suspensions\- Permanently or temporarily ban certain users\.Account SuspensionDeplatforming\- Coordinated efforts to ban prominent misinformation spreaders\.Account SuspensionShadow Banning\- Limit the visibility and the spread of posts from certain policy\-violating accounts\.Shadow BanningDemonetization\- Remove or restrict monetization features for a user account\.DemonetizationContent ModerationContent Removal\- Remove posts verified to contain misinformation\.Misinformation DetectionBan Deepfakes\- Prohibit the use of AI to misrepresent the speech or actions of public figures\.Misinformation DetectionDownranking\- De\-emphasize posts in news feeds that contain misinformation\.Algorithmic ModerationUpranking\- Uprank high\-quality news sources and downrank low\-quality ones\.Algorithmic ModerationVirality Circuit Breakers\- Briefly halt algorithmic amplification of unverified content\.Algorithmic ModerationUser Control\- Give users more control over the algorithms powering their news feeds\.User ControlContent DistributionLimit Forwarding\- Cap the number of users one can forward a given message to\.Account LimitsLimit Resharing\- Remove share buttons on posts after several levels of sharing\.Account LimitsFriction\- Temporarily delay users from posting content they did not open via a pop\-up\.FrictionAccuracy Prompts\- Nudge or remind people to consider accuracy before posting or sharing content\.Accuracy PromptsRedirection\- Redirect users to other content, such as official content or no content\.RedirectionBan Political Ads\- Ban political ads on social media platforms\.Advertising PolicyFact\-Check Ads\- Fact\-check ads on social media platforms\.Advertising PolicyAI in Ads\- Prohibit the use of AI or manipulated content in political or targeted ads\.Advertising PolicyPlatform Alterations\- Reduce the size or visibility of a post containing misinformation\.Platform AlterationsGenerative AIGen AI Content\- Generate rebuttals to misinformation or create educational initiatives\.Generative AIGen AI Chatbots\- Use chatbots to reduce belief in conspiracy theories or misinformation\.Generative AIContent LabelingCrowdsourcing\- Labels generated by the public rather than professionals \(e\.g\., Community Notes\)\.Crowdsourcing, Context LabelsFact Check Labels\- Link information from third\-party fact\-checkers on misinformation posts\.Fact\-Checking, Context LabelsSource Labels\- Labels indicating the reliability of news sources\.Source LabelsGovernment Labels\- Label the accounts of government officials or state\-run media\.Source LabelsAI Disclosure\- Require disclosures on advertisements that use AI\-generated content\.Source Labels, Context LabelsClick\-through Warning Labels\- Place misinformation posts behind click\-through labels containing facts and context\.Warning LabelsUser\-Based MeasuresSocial Norms\- Use social or community norms to encourage social corrections and self\-corrections\.Social Corrections, RetractionsAlter Platform Metrics\- Reward accuracy rather than engagement to discourage misinformation sharing\.Social NormsReporting\- Improved reporting functionality and transparency\.User ReportingMedia Literacy / EducationDigital Media Literacy\- Invest in and promote content on how to critically evaluate online information\.Media LiteracyInoculation\- Inoculate people against misinformation with games or videos\.Inoculation, GamesInstitutional MeasuresGovernment Regulation\- Hold companies accountable for content on their platforms\.Government RegulationPrivacy Legislation\- Develop comprehensive privacy legislation, similar to Europe’s GDPR\.Government RegulationAnti\-Trust Action\- Break up monopolistic big tech companies\.Government RegulationTaxes / Fines\- Tax or fine social media companies for their use of personal user data\.Government RegulationTargeted Advertising\- Limit or ban micro\-targeted advertising\.Government RegulationMedia Support\- Promote and invest in local media\.Media SupportJournalism Support\- Support and train the next generation of journalists\.Media SupportData Sharing\- Have companies regularly release data and internal research reports to researchers\.Data SharingTable 7\.Operationalized misinformation interventions across the eight general categories\.