Human ai collaboration improves reasoning when users trust models selectively
Available but Unclaimed: An Empirical Study of Human-AI Synergy
Human-Computer InteractionArtificial Intelligence
Summary
People work together with AI language models to solve reasoning problems, but just having a good AI doesn't guarantee better results. The authors studied how people used several AI systems while solving puzzles and found that people relied more on AI help when the AI was better at specific tasks. However, people’s confidence in their answers after consulting AI was less accurate than when they answered alone. The research shows that some of the AI's accuracy helps people more than others, highlighting the need to support smart ways for humans to decide when to trust AI advice.
What this means in practice
- •For human-computer interaction designers: Design interfaces that help users decide when to trust AI advice for better collaborative problem solving.
- •For cognitive training software developers: Create training tools that use AI assistance while encouraging independent reasoning to improve user learning outcomes.
Authors
Robin Welsch, Michelle Rausch, Pascal Knierim, Thomas Kosch, Jochen Kuhn, Albrecht Schmidt, Daniela Fernandes
Abstract
People increasingly reason with large language models (LLMs), yet complementary capabilities do not guarantee outperforming both components. In a between-subjects study, participants (N=535) solved a 40-item battery of matrix reasoning, mental rotation, syllogisms, and letter-string analogies, unaided or with GPT-5.6-Luna, Claude Opus 4.8, Gemini 3.6 Flash, or Kimi K3. Each assisted trial required consultation with the model. Each model answered every item alone 100 times under matched elicitation. The assisted-unaided accuracy difference increased with item-level LLM competence. Deference varied across tasks and increased with competence within tasks. Post-advice confidence distinguished correct from incorrect answers less strongly than unaided confidence. In a reference comparison, about half the increase in LLM accuracy carried through to assisted accuracy. How much of that accuracy gain reached participants differed across the models. These findings motivate evaluating LLMs in interaction with humans and designing support for selective deference that preserves independent reasoning.