DeepMind RCT with 992 participants: repeated LLM use, not personalization, drives perceived closeness
Dr_Atoosa · x · 2026-11-03
Google DeepMind researchers ran a five-day longitudinal RCT with 992 participants seeking daily relationship advice from LLMs, comparing a non-personalized baseline with memory-based and survey-based personalization.
Key findings:
- Repeated exposure itself drove increases in perceived closeness, usefulness, and competence — not personalization
- Different personalization approaches did change users' advice receptivity and information-sharing attitudes and behaviors
The team highlights that downstream effects on self-perception and interpersonal relationships remain largely unexamined, with implications for responsibly designing personalized AI systems.
More from AGI Musings
- Gary Marcus doubles down: Anthropic's AGI projections were 'insane,' $10K bet still on — GaryMarcus · 2026-11-03
- 'AI is just matrix multiplication' is the new 'emotions are just chemicals' — karmicoder · 2026-11-03
- LLMs forecast economics RCT effects with 0.054 SD median error, beating naive benchmarks — RobbWiller · 2026-11-03
- Addy Osmani on keeping critical thinking alive when AI agents do the work — addyosmani · 2026-11-03
- Indie dev: AI coding erased the technical moat — your rival is now the better seller — sven_ai · 2026-11-03
- AI practitioner argues frontier labs are overlooking chemistry as the richest RL sandbox — gabepgomes · 2026-11-03