---
title: "Research | Stanford HAI: How Culture Shapes What People Want from AI"
author: "Unique Research"
sourcePublication: "Unique Research (FFCAP English library)"
originalPublishedAt: "2024-08-26T16:00:00+08:00"
canonical: "https://ffcap.cn/en/research/20240826a05vvp00"
source: "https://ffcap.cn/en/research/20240826a05vvp00"
language: "en"
cnSourceUrl: "https://view.inews.qq.com/a/20240826A05VVP00"
---

# Research | Stanford HAI: How Culture Shapes What People Want from AI

_Unique Research · Insight / Digest · August 26, 2024 · Source: Nikki Goth Itoi | Stanford HAI; researchers Xiao Ge, Chunchen Xu, Jeanne Tsai, Hazel Rose Markus_

If AI agents play active roles in our social lives — not only running quietly in the background — is that desirable? The answer may depend on culture. Applying cultural psychology to expectations of AI, a Stanford team found clear links between cultural background and the kind of AI people idealize.

Mainstream AI discourse often assumes people want **control** over technology — an impersonal, hierarchical tool that serves individual goals. Psychologists **Xiao Ge** and postdoc **Chunchen Xu** argue that view is not universal; it reflects an agency model common in many European-American middle-class contexts.

Broader observation shows other ideals: some people want AI with greater capacity to influence its surroundings — emotion, affect, autonomy — even machines that act spontaneously in social settings. “Future AI urgently needs the views, imaginaries, concerns, and creativity of diverse groups,” Ge says, so stakeholders can widen which worldviews shape design and use.

Drawing on independent vs. interdependent cultural models and two online surveys, the team proposes a framework in the paper _How Culture Shapes Expectations of AI_, supported by a Stanford HAI seed grant.

## Flipping the HCI lens

Much HCI work studies how tech affects people across countries; few reverse the question — how culture shapes AI design, or how products encode cultural defaults. Co-author **Jeanne Tsai** (Stanford Culture and Emotion Lab): culture is often considered late (usability, UI), but findings suggest it can matter at _initial_ creation — including designers’ imaginations of benefit and outcome.

They use a mature cultural-psychology frame: **independent** selves as unique and separate from context vs. **interdependent** selves as fundamentally connected to people and environments — plus “influence capacity,” whether people want the environment to act on them. Co-author **Hazel Rose Markus** (Stanford SPARQ): those two dimensions surface human–AI ideals far from what individualism middle-class settings default to.

## Two (plus one) cultural models

-   **European-American contexts:** independent agency — people more influential than environment; change environment to match preferences.
-   **Chinese contexts:** interdependent agency — permeable person–environment boundary; environment can be more alive and influential.
-   **African American contexts:** elements of both, shaped by switching between mainly independent and more interdependent settings.

Hypotheses: vs. Chinese respondents, European-Americans want more control of AI and less connection; Chinese respondents want more connection and more influence-capable AI (autonomy, spontaneity, affect). African American preferences expected in between.

## Testing

First survey (n=373): seven diagrams of self–environment balance from “environment strongly influences person” to reverse; cultural differences matched expectations. Second study (n=348): short AI description, random assignment to one of six scenarios (home, health, teamwork, education, wildfire protection, manufacturing), then preference items aligned with independence/interdependence and HCI traits (autonomy, perceived affect).

## Findings

Culture shapes not only what “being human” means but what people want from AI. Chinese participants rated **control** of AI as less important and **connection** as more important than European-Americans; European-Americans preferred lower AI influence capacity. African American preferences sat between on connection and optimal influence. Notably, Chinese means still showed _some_ desire for control — lowest of the three, not zero.

## Why it matters for HCI

Markus and Tsai: the work shows rigorous empirical methods can map culture-shaped preferences for AI purpose, form, and function — and how Western HCI defaults make shared agency outside the individual hard to imagine, though elsewhere it is obvious. Xu: relying only on existing cultural models may constrain creativity and AI’s global upside; rethinking human agency with a wider cultural palette could open a new innovation era.

_Original: Nikki Goth Itoi | Stanford HAI. Partner signup channel noted in CN; not expanded here._

— END —

---

Editorial English edition published on UniqueCapital: https://ffcap.cn/en/research/20240826a05vvp00
On-site reading page: https://ffcap.cn/en/research/20240826a05vvp00