AI Assistance May Narrow Human Choice Diversity
A commentary argues that AI assistants and recommendation systems may reduce both the range of individual choices and diversity across people. It cites Sandra Matz’s Baskin Robbins experiment, in which ChatGPT offered only two of the shop’s 31 flavors in 96 of 100 prompts. The commentary also says Matz’s team analyzed more than 110,000 social-media choices and an experiment involving 1,000 US participants, finding that choices became less distinctive with both general and personalized LLM assistance. It further discusses research on university essays, writing variation before and after ChatGPT’s widespread adoption, and a natural experiment during Italy’s temporary ChatGPT ban.
The commentary identifies a concrete concern: systems optimized to provide predictable, satisfying recommendations may also concentrate people’s choices and reduce their uniqueness. The cited experiments and studies offer evidence relevant to whether AI assistance broadens or narrows human decision-making.

我们对 AI 的信任,正在达到史无前例的高度。无论是在屏幕前码字,还是在生活中做决策,我们都习惯向 AI 寻求那个“完美建议”。用得多了,你甚至会发现它总是精准地踩中我们预期中的最优解。这种顺滑,很可能是一种被精心设计的心理按摩。哥伦比亚大学教授桑德拉·马茨用冰淇淋店 B......
本文来自微信公众号: APPSO ,作者: APPSO,题图来自:AI生成
我们对 AI 的信任,正在达到史无前例的高度。
无论是在屏幕前码字,还是在生活中做决策,我们都习惯向 AI 寻求那个“完美建议”。用得多了,你甚至会发现它总是精准地踩中我们预期中的最优解。
这种顺滑,很可能是一种被精心设计的心理按摩。