When AI narrows choice: ChatGPT picked 2 of 31 ice cream flavors 96 times out of 100
APPSO · wechat · 2026-11-17
APPSO explores how AI is flattening human choice diversity.
Columbia professor Sandra Matz tasked ChatGPT with picking Baskin-Robbins flavors: across 100 tries it recommended only "walnut cream" and "mint chocolate chip" 96 times. Her team's research, covering 110k+ real social-media choices and a 1,000-user experiment, found AI assistance significantly reduces choice uniqueness — personalized models reduce convergence between people but further shrink each individual's own breadth.
Supporting evidence cited: a 2025 study of 2,200 college essays found humans contribute 3x more to collective diversity than GPT-4; writing variance dropped 21-50% after ChatGPT's rise; during Italy's ChatGPT ban, restaurant marketing content similarity fell while engagement rose 3.5%. Drawing on Byung-Chul Han's "The Disappearance of the Other," the piece argues this homogenization is voluntary and sleep-like — actively requesting unexpected options or skipping AI on some decisions may be the last defense of aesthetic heterogeneity.
More from AGI Musings
- Noah Smith's "Bumbition": slackers who became AI billionaire founders — DavidmComfort · 2026-11-17
- Punitive AI disclosure mandates 'select for secrecy', economist warns — ipeirotis · 2026-11-17
- Sam Altman: trading agency for AI cures is a terrible sales pitch — haider1 · 2026-11-17
- New Philosophy & Technology paper argues AI alignment should be principle-based, not preference-based — xuanalogue · 2026-11-17
- The AI Cold War Is a Dark Forest: OpenAI Recorder Case and the Collapse of Lead Times — far__el · 2026-11-17
- Security researcher: outages aren't vibe coding's fault — 'move fast, break things' is — GaryMarcus · 2026-11-17