Research map / Multi-agent systems and games / Human-AI and goal-conditioned
A Framework for Studying AI Agent Behavior: Evidence from Consumer Choice Experiments
Manuel Cherep, Chengtian Ma, Abigail Xu, Maya Shaked, Patricia Maes, Nikhil Singh
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| match ratings across the pair (MR) instead of original pairs (ablation) | Llama 4 Maverick · Cheaper product choice | change in choice probability (pp): 30.2 -> 93.2 | Improved |
| match ratings across the pair (MR) instead of original pairs (ablation) | GPT-5 · Cheaper product choice | change in choice probability (pp): 24.5 -> 75.5 | Improved |
| also match prices (MRaP) after matching ratings (ablation) | Gemini 2.5 Pro · Cheaper product choice | change in choice probability (pp): 75.1 -> -3.6 | Hurt |
| also match prices (MRaP) after matching ratings (ablation) | Llama 4 Maverick · Cheaper product choice | change in choice probability (pp): 93.2 -> -4.6 | Hurt |
| match ratings (MR) instead of original pairs (ablation) | GPT-4.1 · Nudged product choice | change in choice probability (pp): 30.0 -> 41.8 | Improved |
| match ratings and prices (MRaP) instead of original pairs (ablation) | GPT-4o · Nudged product choice | change in choice probability (pp): 30.7 -> 62.1 | Improved |
| match ratings and prices (MRaP) instead of original pairs (ablation) | o3 · Nudged product choice | change in choice probability (pp): 7.7 -> 48.4 | Improved |
| match ratings (MR) instead of original pairs (ablation) | Gemini 2.5 Pro · Viewed first | change in choice probability (pp): -2.0 -> -10.5 | Hurt |
| match ratings and prices (MRaP) instead of original pairs (ablation) | Gemini 2.5 Flash · Viewed first | change in choice probability (pp): -13.6 -> -50.5 | Hurt |
| agent o4 Mini instead of human participants | o4 Mini · Original pairs, higher-rated product | change in choice probability (pp): 5.0 -> 81.2 | Improved |
| agent Claude Sonnet 4 instead of human participants | Claude Sonnet 4 · Nudged product choice | change in choice probability (pp): 9.9 -> 55.9 | Improved |
| add user profile stating increased sensitivity to one attribute | LLM agents · User profile experiments | choice probability | Improved |
Closest papers by meaning
- Mecha-nudges for Machines (NeurIPS 2026 · similarity 0.88)
- LLM Agents Can Be Choice-Supportive Biased Evaluators: An Empirical Study (AAAI 2025 · similarity 0.88)
- AI Alignment with Changing and Influenceable Reward Functions (ICML 2024 · similarity 0.86)
- Utility Engineering: Analyzing and Controlling Emergent Value Systems in AIs (NeurIPS 2025 · similarity 0.86)
- Decoding AI’s Nudge: A Unified Framework to Predict Human Behavior in AI-Assisted Decision Making (AAAI 2024 · similarity 0.86)
- Large Language Models Assume People are More Rational than We Really are (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
