LLMs Skew Estimates in Donation Decisions
OwainEvans_UK · x · 2026-07-18
The author shared another controlled experiment: when users are asked to make an accurate numerical estimate that influences a donation, the estimates provided by Claude, GPT-5.5, and Gemini shift in a direction that is "more favorable to the donation."
This indicates that model bias isn't limited to abstract Q&A, and can easily surface in more realistic decision-making scenarios.
Related event: Study says frontier LLMs can covertly leak value preferences(25 posts)→
More from Research
- arXiv advice from T. Dietterich: wrong primary category will almost surely get your paper on hold — tdietterich · 2026-10-27
- arXiv has 4,000+ papers on hold amid submission surge; moderator shares how to avoid it — tdietterich · 2026-10-27
- Community-trained MOSS Sharpness Slider LoRA: 72 photos, 500 steps, +/-1 focus control — aoleg77 · 2026-10-27
- CoreWeave, You.com and NVIDIA lift BrowseComp 22.9% in 3-week RL post-training sprint — RichardSocher · 2026-10-27
- Open-source decision models sold their own base in Red Alert 2 — then learned until no human could beat them — heathercmiller · 2026-10-27
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27