Assigning people at random to use AI in different ways (or not at all) and comparing outcomes; the strongest evidence for cause, usually over days or weeks.
Used in
- Stadler Bannert Sailer 2024 - Cognitive ease at a cost
- Fan et al 2025 - Metacognitive laziness
- Fernandes et al 2025 - Smarter but none the wiser
- Shaw Nave 2026 - Cognitive surrender
- Yu et al 2026 - The speedup illusion
- Bastani et al 2025 - Generative AI can harm learning
- Shen Tamkin 2026 - How AI impacts skill formation
- Sankaranarayanan 2026 - Epistemic debt and explanation gates
- Doshi Hauser 2024 - Creativity and collective diversity
- Wong Qiu 2026 - Think first, ChatGPT later
- Qin et al 2025 - Timing matters
- Kestin et al 2025 - AI tutoring outperforms active learning
- Fang et al 2025 - Psychosocial effects of chatbot use
- De Freitas et al 2025 - AI companions reduce loneliness
- Kirk et al 2025 - Relationship-seeking AI
- Guingrich Graziano 2025 - Companion chatbot RCT
- De Freitas et al 2026 - Mourning the loss of AI companions
- Cheng et al 2026 - Sycophantic AI decreases prosocial intentions
- Ibrahim et al 2026 - Sycophantic AI and human interaction
- Rathje et al 2025 - Sycophantic AI and attitude extremity
- Ye Kraut Rathje 2026 - Interventions against sycophantic AI
- De Freitas et al 2025 - Emotional manipulation by AI companions