Do reflection questions help people make better decisions with AI?
This explores whether conversational AI that prompts users to think through problems outperforms AI that simply provides answers. Understanding this matters for designing AI tools that genuinely improve human judgment rather than replace it.
Through a lab study (N=80), LLM-based "Thinking Assistants" that combine asking reflection questions with providing advice outperform conversational agents that only ask questions, only provide advice, or neither. The key insight: "Rather than adhering to the prevailing authoritative approach of generating definitive answers, LLM agents aimed at assisting with cognitive enhancement should prioritize fostering reflection. They should initially provide responses designed to prompt thoughtful consideration through inquiring, followed by offering advice only after gaining a deeper understanding of the user's context and needs."
This directly challenges the default LLM interaction paradigm. Since Why can't conversational AI agents take the initiative?, the Thinking Assistant approach provides an alternative to passivity that doesn't require proactivity in the traditional sense — instead of the AI taking initiative to redirect, it takes initiative to question. This is proactivity in the Socratic mode rather than the directive mode.
The approach leverages LLMs' encoded "world-knowledge" while avoiding the authority-without-accountability problem that since Does polished AI output trick audiences into trusting it?, direct answers carry unearned authority. Questions carry no such risk — they prompt the human to exercise their own judgment.
Inquiring lines that read this note 31
This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.
What enables conversational agents to guide rather than just respond?- What are the five specific conversation triggers where AI intervention adds value?
- Why do conversational systems benefit from post-thinking between user turns?
- Does conversational AI reduce learner control over information selection?
- How does AI assistance differ from search engines in cognitive impact?
- How does AI assistance affect human cognitive development over time?
- How does AI assistance change learning outcomes across different cognitive engagement levels?
- Does metacognitive feedback about AI assistance work beyond a single online session?
- Why does AI assistance trigger harsher judgment than other workplace tools?
- Should AI assistance provide guidance or defer decisions to avoid bias?
- Does metacognitive feedback reduce reliance on AI-generated answers?
- Can humans using superhuman AI actually increase their own novel problem-solving?
- Does receiving AI advice undermine people's own moral reasoning and decision-making skills?
- Does the timing of AI feedback relative to user reasoning change its effectiveness?
- How can AI systems help users clarify what they actually want?
- How do contrasting examples improve AI feedback quality over generic suggestions?
- Why do users prefer AI responses that actually harm their decision-making?
- How might automated evals eventually capture the human judgment designers exercise now?
- Can explicit reflection during AI-assisted work improve transfer of learning?
- Why does knowing what questions to ask improve AI output?
- What design features make tutoring AI preserve learning better than answer-giving AI?
- Can perspective-taking theory predict who benefits most from human-AI partnership?
- How does perspective-taking predict who benefits from human-AI collaboration?
Related concepts in this collection 1
This note in its neighbourhood — explore the map, then jump to a related concept in the list below.
Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph
-
Can models learn to ask clarifying questions instead of guessing?
Exploring whether large language models can be trained to detect incomplete queries and actively request missing information rather than hallucinating answers or refusing to respond. This matters because conversational agents today remain passive, responding only when prompted.
Thinking Assistants are the decision-support analog of proactive critical thinking
Related papers in this collection 8
Papers most semantically related to this note, ranked by cosine similarity in the embedding space.
- Thinking Assistants: LLM-Based Conversational Assistants that Help Users Think By Asking rather than Answering
- When Hindsight is Not 20/20: Testing Limits on Reflective Thinking in Large Language Models
- Toward Measuring AI's Effects on Skill Formation: The Stock-Formation Gap
- GenAI as a Power Persuader: How Professionals Get Persuasion Bombed When They Attempt to Validate LLMs
- Self-Reflection in LLM Agents: Effects on Problem-Solving Performance
- “Understanding AI”: Semantic Grounding in Large Language Models
- AInsight: Augmenting Expert Decision-Making with On-the-Fly Insights Grounded in Historical Data
- How AI Impacts Skill Formation
Original note title
thinking assistants that ask reflection questions outperform those that only provide answers — fostering reflection over authority improves human decision-making with AI