authorship

An author tests Pangram's AI-detection accuracy after a French literary controversy, finding that short samples flagged paragraphs as AI while the full chapter scored 100% human, raising questions about false positives with short inputs.

Dartmouth provost Santiago Schnell's Washington Post op-ed on AI cheating was flagged as largely LLM-generated by the Pangram detector, and even his email defending his use of AI was found to be '100% AI-written,' raising broader questions about authorship and accountability in AI-assisted writing.

The paper proposes three approaches for few-shot author-style adaptation of LLMs—contrastive activation steering, a steering-vector prediction network, and a hypernetwork that predicts LoRA adapters—and finds the hypernetwork offers the best trade-off between style imitation and output quality for both seen and unseen authors.

The article discusses concerns that AI-generated books could overshadow human-written fiction, particularly in self-publishing, and advocates for transparent disclosure of AI involvement in writing.

The article critiques AI detection tools for relying on surface-level patterns that can misidentify human writing, especially when polished, and highlights their failure to capture the unique voice of individual writers.

OpenAI has updated ChatGPT to block direct requests to mimic specific authors' writing styles, responding to ethical and legal concerns. The change affects users who relied on style replication, while other AI models vary in their compliance with such requests.

An article analyzing the legal battle over copyright for AI-generated art, centered on Jason Allen's case against the Copyright Office and the broader implications for human authorship in the age of generative AI.

A pilot study with 24 college students examines how varying levels of LLM access (none, limited, unlimited) affect essay writing quality, behavior, and perceived authorship, finding that constrained access preserves authorship confidence while unlimited access reduces creative expression and ownership.

An article discussing the controversy over a prize-winning short story that was accused of being generated by AI, and the broader implications for authorship and detection in the age of large language models.

Tom Dietterich reminds arXiv authors that signing as an author means taking full responsibility for all contents, regardless of how they were generated, highlighting implications for AI-generated content.

An analysis arguing that the optimal balance for AI-assisted writing is around 50% AI and 50% human input, where AI handles structure and organization while humans provide voice, judgment, and editorial control. The author contends that 100% AI reads as slop while 0% AI leaves capability on the table, and that meaningful AI assistance requires genuine expertise, strong structure, and distinctive human voice.