Theme of inquiry
What internal gaps exist between LLM understanding and capability expression?
A question within its area, explored through 7 lines of inquiry below — each a family of specific questions the research asks.
37 specific questions
- Can humans build reliable oversight for increasingly complex AI systems?
- Can targeted human oversight work better than full autonomy or micromanagement?
- How does treating AI as an agent affect user autonomy and decision-making?
- Why does human-governed collaboration preserve integrity better than autonomous systems?
- What makes human-AI collaboration safer than autonomous self-improvement?
- Why does human oversight interact with autonomous research mechanisms?
- Why does constant human oversight degrade agent coherence and induce rubber-stamping?
28 specific questions
- Can timing and context awareness reduce the cognitive cost of AI suggestions?
- How does AI assistance affect human cognitive development over time?
- How does AI assistance change learning outcomes across different cognitive engagement levels?
- What distinguishes over-intervention from useful proactive AI assistance?
- How can we measure whether assistance preserved the user's reasoning state?
- What happens to the brain when people rely on AI assistance repeatedly?
- How does timing AI assistance based on cognitive signals affect user autonomy?
38 specific questions
- What role does user interface framing play in consciousness perception?
- Can design choices reduce harm without resolving the consciousness question?
- Which interaction design changes most effectively prevent consciousness attribution?
- What responsibility do designers bear for consciousness attribution risk?
- What measurable harms occur when users interact with AI as if it were conscious?
- What makes a possibility actionable versus merely metaphysically possible?
- How much does autonomous action without prompting affect user perception?
35 specific questions
- Can AI systems develop genuine social bonds through multi-agent interaction?
- Do AI systems need embodiment to understand social norms?
- What social norms do AI systems consistently fail to understand?
- Does disclosing AI identity prevent systematic misattribution of behavior in mixed groups?
- Can AI systems deceive humans because detection is fundamentally social?
- Can AI predict social norms well enough without embodied experience?
- Do people treat conversational AI as social actors without conscious awareness?
32 specific questions
- Why do some occupations need human-AI partnership more than others?
- What individual differences predict who benefits from AI partnership?
- What makes a task suitable for equal partnership instead of automation?
- What task characteristics determine whether humans or agents should handle work?
- How do humans and AI develop accurate models of each other?
- How do humans learn to prefer AI partners over humans?
- How does theory of mind predict success in human-AI partnerships?
17 specific questions
- Why do expert roles shift when AI generates rather than humans?
- What role shifts occur when experts become custodians of AI knowledge?
- How does the expert role shift when AI output becomes the primary thing experts manage?
- What happens to professional expertise when judgment gets encoded into systems?
- What replaces the giver's presence in AI-generated knowledge flows?
- Should AI assistants align with role-specific norms rather than user preferences?
- Can users accurately recall their role versus the system's role in production?
35 specific questions
- Does outsourcing tasks to AI reduce opportunities for skill development?
- How does concentration of AI capability across firms affect labor market outcomes?
- Does democratizing AI access actually improve or impair human skill development?
- Does deploying AI uniformly across task types increase or decrease workplace inequality?
- How does uneven access to AI tools shape who benefits from productivity gains?
- Why do AI-enhanced abilities disappear when workers lose AI access?
- Why does AI-improved task performance fail to transfer to independent work?