Tag
This position paper argues that AI evaluation should pivot to assessing human-AI teams rather than superhuman performance to foster better societal outcomes.
This paper proposes Principal Trait Analysis (PTA), a data-driven method to derive common behavioral traits from human-AI collaborative coding conversations, evaluating it on educational and professional datasets to understand what skills contribute to task success.
This empirical study compares conversational XAI (powered by LLMs) against a traditional dashboard for UAV intrusion detection auditing, finding the conversational interface improves perceived usefulness but risks operator over-reliance on AI advice.
This paper presents CoPlan, a co-intelligent and contestable interface for human-AI care planning that uses a multi-agent workflow to generate candidate interventions and arguments, allowing human care planners to inspect, challenge, and revise recommendations before final plan generation. It demonstrates the approach in an aging-in-place scenario and contributes a design framing for trustworthy human-AI care planning.
AgentPanel is a multi-agent forum system for human–AI collaboration in scientific exploration, enabling heterogeneous agents to asynchronously discuss research questions. Evaluations show it outperforms centralized multi-agent debate and was favored by 65% of participants for early-stage exploration.
An essay argues that humans should receive primary credit for AI-assisted discoveries, countering OpenAI's claim that AI systems generating mathematical arguments should be attributed as discoverers.
MusiChat presents a conversational system for human-AI music co-creation that enables iterative refinement through natural language interaction, achieving high accuracy in multi-turn editing.
Humanly is a configurable writing platform that records the writing process to provide traceable evidence of human-AI collaboration, with features like sealed certificates and anomaly detection.
This paper introduces Mycelium, an active shared workspace that automatically connects researchers and AI agents to enable networked intelligence in team science, evaluated in a multi-omics campaign.
This paper introduces a hierarchical human-AI triage model for POS fraud detection in Nigerian FinTech, designed to neutralize structural bias that discriminates against rural agents due to infrastructure-related noise. The system uses a calibrated ensemble model, specialist analysts, and senior supervisors to achieve substantive equality of opportunity, reducing the regional performance gap from 19.43 to 2.88 percentage points.
An essay arguing that the technological singularity may arise not from a single superintelligent AI, but from constellations of specialized human–AI partnerships, and warns against corporate control of such systems.
A recommended read arguing that the 'AI replacing humans' narrative is a scam, instead advocating for AI to augment and empower humanity, as echoed by Mira Murati's vision.
The article criticizes the prevalent use of chat windows as the interface for AI programming tools, arguing that this design restricts the autonomous working ability of AI from both the user and developer ends, and proposes a shift to an asynchronous, task-driven collaboration model.
John Schulman reflects on the founding principles of Thinking Machines, emphasizing customization and human-AI collaboration, and shares the company's updated vision in light of recent progress in AI agents.
This paper introduces a Human-Centric Reflective Architecture (HCRA) for human-AI collaborative decision-making, formulating the task as a stochastic game and using reinforcement learning with linguistic feedback. The proposed framework significantly improves decision-making effectiveness and recommendation quality.
New research from Thinking Machines critiques current single-threaded AI interaction models, arguing that they limit human-AI collaboration by forcing humans into clean input-output cycles. The lab proposes a new interaction model that supports continuous, multi-modal collaboration akin to real-time human conversation.
Treating AI agents as coworkers rather than tools reduces error detection by 18% and shifts responsibility, according to a study by Boston University professor Emma Wiles. The article warns against over-personifying AI, citing risks in healthcare, warfare, and government.
This paper presents a case study of human-AI co-discovery in mathematics, where AI assisted in expanding an intuition about sign-embedding quantum algorithms into a formal framework and proofs, with human judgment guiding route selection.
A researcher spent five days testing an alignment hypothesis across multiple AI systems, observing recurring themes like the value of uncertainty and collaboration over obedience, finding that ideas evolve through dialogue and criticism.
This paper studies human-AI team coordination in shared workspaces using the Collaborative Gym and DiscoveryBench tasks, finding that adding collaborators can lower performance without proper structure. Scaffolding with shared group memory and human-in-the-loop gates improves performance, especially in three-person teams.