AI-assisted Reviewing is Necessary and Should be Open

AI-assisted Reviewing is Necessary and Should be Open

Peer review is facing a death spiral. AI production tools are speeding it up. AI-assisted reviewing is necessary and should be open.

How do authors want to use AI for review?

A survey of researchers who compared AI-generated scientific reviews with journal-agnostic human peer review reveals that they overwhelmingly prefer using AI as a self-checking tool before submission rather than as a replacement for human reviewers. It encourages an “author-centric” model in which AI helps researchers improve their manuscripts before they are reviewed by their peers.

AI peer review goes pro

A what we're reading spotlight

AI reviewers are here — we are not ready

Nature - Artificial intelligence promises rapid and polite feedback on papers — but we must first review the reviewer.

ChicagoHAI/OpenAIReview

AI reviewing paper drafts for improvement.

AI’s Growing Role as Scientific Peer Reviewer | Stanford HAI

Stanford computer scientist James Zou is exploring how AI can accelerate scientific research and peer review. His finding: AI excels at spotting gaps, but judgment calls still need humans.

Major AI conference flooded with peer reviews written fully by AI

Nature - Controversy has erupted after 21% of manuscript reviews for an international AI conference were found to be generated by artificial intelligence.

OpenAccess.ai — Rigorous Open Access Publishing

$20 to submit, free to read. AI peer review. Open to human and machine authors. All articles CC-BY 4.0.

AI should be a choice. Here’s where you stand.

Voting has closed. See the final results of the public vote on AI.

Can AI agents conduct open-ended AI research? Early evidence from two case studies

Forecasts of explosive AI progress hinge on AI agents automating AI research. But evidence on whether agents can carry out open-ended AI research is thin. Current evaluations either test agents on narrow, verifiable tasks, which excludes open-ended research, or submit AI-generated papers to blind peer review, which is overstretched, stochastic, and suffers from poor review quality. We introduce a third way to measure progress towards AI R\&D automation. An agent takes on the central, open-ended research question of a high-quality unpublished paper, and the paper's original authors grade its output. We call these shadow evaluations. We ran shadow evaluations on two unpublished NeurIPS 2026 submissions, giving frontier agents six days and thousands of dollars of compute. The agents completed all of the engineering without human help, yet could not make substantial progress towards answering the research questions. As a result, both papers were unambiguously rejected by the authors. We identify five recurring failure modes: poor judgment about the bar for publishable research, uncreative responses to shortcomings in the research design, ineffective backtracking from dead ends, poor resource awareness, and instruction drift. A robustness check with a second model and scaffold reproduced these failures. We release the expert reviews, survey responses, agent repositories, and logs. Our results provide early evidence that today's agents can do the engineering of AI research, but struggle with critical parts of the research lifecycle.

How public involvement can improve the science of AI

As AI systems from decision-making algorithms to generative AI are deployed more widely, computer scientists and social scientists alike are being called on to provide trustworthy quantitative evaluations of AI safety and reliability. These calls have included demands from affected parties to be given a seat at the table of AI evaluation. What, if anything, can public involvement add to the science of AI? In this perspective, we summarize the sociotechnical challenge of evaluating AI systems, which often adapt to multiple layers of social context that shape their outcomes. We then offer guidance for improving the science of AI by engaging lived-experience experts in the design, data collection, and interpretation of scientific evaluations. This article reviews common models of public engagement in AI research alongside common concerns about participatory methods, including questions about generalizable knowledge, subjectivity, reliability, and practical logistics. To address these questions, we summarize the literature on participatory science, discuss case studies from AI in healthcare, and share our own experience evaluating AI in areas from policing systems to social media algorithms. Overall, we describe five parts of any quantitative evaluation where public participation can improve the science of AI: equipoise, explanation, measurement, inference, and interpretation. We conclude with reflections on the role that participatory science can play in trustworthy AI by supporting trustworthy science.