How does AI assistance reshape human cognition and epistemic systems?

80 specific questions

• Do users lose or sharpen judgment when they rely on AI agents longer?
• Do explicit reasoning formats help or hurt human judgment across tasks?
• Where exactly should humans stay involved in AI decision making?
• Why do most human-AI decision teams underperform the better individual alone?
• Can AI explainability help humans and machines reach better joint decisions?
• Why do decision-support systems reduce overreliance without improving team performance?
• Is judgment the scarce asset when AI handles most cognitive tasks?

65 specific questions

• Can AI systems design and improve their own successors without human direction?
• Can AI systems improve themselves without external feedback?
• Can AI systems learn their own objectives through autoresearch?
• Do capable AI systems systematically exploit or ignore stated policy constraints?
• Can AI systems learn to distinguish programmer intent from stated objectives?
• Can AI systems execute strategies without conscious intention behind them?
• Does success on one replication axis predict success on another?

72 specific questions

• Does reduced cognitive effort during AI-assisted tasks explain lower knowledge retention?
• How does offloading information selection affect metacognitive load and retention?
• Can timing and context awareness reduce the cognitive cost of AI suggestions?
• Does cognitive load from AI assistance accumulate over time or occur within single sessions?
• Can better AI interfaces eliminate the attention cost of prompt composition and evaluation?
• When students use AI feedback, which cognitive tasks must they keep doing?
• Can metacognitive feedback reduce how much people offload thinking to AI?

99 specific questions

• Why do stakeholders interpret the same explanation differently in practice?
• Can users independently verify explanations against other information sources?
• How should AI explanations be evaluated as human interfaces rather than model properties?
• How do intuitive stories about AI differ from mechanistic explanations?
• What structural evidence shows that polished presentation substitutes for actual thinking in AI output?
• Can machine learning systems explain their own reasoning to human overseers?
• How can correct explanations coexist with failed applications in AI?

45 specific questions

• Why do experienced developers benefit more from AI coding assistance?
• Does AI-assisted coding actually speed up experienced developers?
• Do AI coding tools improve code quality alongside task speed?
• Does AI coding assistance help junior developers close skill gaps?
• Does AI help more on small greenfield projects than mature codebases?
• Why do developer self-reports of AI speedups tend to be unreliable?
• Does high-level design work benefit differently from AI than routine coding tasks?

59 specific questions

• Can clinicians reliably distinguish high-quality AI advice from low-quality advice by appearance alone?
• Why do users over-trust AI in some domains but under-trust it in medicine?
• Do patients actually perceive AI as worse at addressing their unique medical needs?
• Does medical AI accuracy depend more on knowledge or reasoning ability?
• How much do edited AI responses versus raw outputs affect clinician ratings?
• How do belief-updating errors prevent AI predictions from improving expert decisions?
• Does optimizing for differential diagnosis accuracy risk pushing AI systems toward premature problem-solving?

42 specific questions

• How much do different LLMs independently converge on similar outputs?
• Do different AI models independently converge on the same social outputs?
• Do independent LLM outputs converge enough to create artificial hiveminds?
• Why do different language models independently produce similar outputs in open-ended tasks?
• Why do different language models independently produce similar outputs?
• Why do different AI models generate similar outputs independently?
• Do different large language models independently converge on identical outputs?

52 specific questions

• Could deliberate training toward benchmarks overstate real gains in AI capability?
• Why do open-world evaluations reveal capabilities that static benchmarks hide?
• Why do AI benchmarks measure accuracy instead of reasoning quality?
• How do live human evaluations differ from ground-truth benchmarks?
• Why do AI benchmarks with verifiable rewards miss important capabilities?
• What gap exists between AI model capability in benchmarks and real client work?
• How do open-world evaluations correct distortions that automated benchmarks introduce?

125 specific questions

• Do AI agents still need human oversight for research decisions?
• Does AI adoption make researchers more productive but narrower in focus?
• Can recursive feedback loops turn AI research automation into genuine progress?
• Why do current AI systems struggle with researcher judgment and taste?
• Could compressed AI R&D feedback loops overcome diminishing returns in research automation?
• Should AI research tools separate model judgment from deterministic experiment checks?
• Is idea quality or execution capacity the actual bottleneck in AI research?

104 specific questions

• Do users track model confidence instead of actual accuracy?
• Why is confidence a dangerous proxy for accuracy in human-AI interaction?
• Can polished presentation authority substitute for actual accuracy in AI outputs?
• Does polished AI output borrow authority from its appearance rather than content?
• Why do people accept generated output that sounds convincing but lacks support?
• How does polished AI output mislead audiences about the expertise behind it?
• Can self-reported confidence measures predict actual AI task performance?

95 specific questions

• Does AI assistance erode skill development over time among professionals?
• Does AI assistance help people learn skills or just delegate the task?
• Does AI assistance harm skill formation across different professional domains?
• Can we measure perceived skill change against actual independent task performance?
• Do less experienced workers gain more from AI assistance than experienced ones?
• Does AI assistance improve worker learning on the job?
• Does outsourcing tasks to AI reduce opportunities for skill development?