Working Papers
This series showcases the diverse and impactful research in our Training Centre
Author: Katie J. McDonnell, Steven J. Bickley, Ho Fai Chan, Benno Torgler, Stephen Whyte
Title: Human values and the uneven cultural alignment of artificial intelligence
Abstract: Human values shape which interests count and how societies organize collective life. As people increasingly use AI systems for information, advice, and judgment, model-generated responses may influence how values and cultural practices are expressed across diverse contexts. A demanding case arises when AI serves as synthetic respondents or proxies for people in research and decision-making: can they faithfully represent the values of the populations they stand in for? We examine this question by comparing GPT-4o responses with those of 13,736 people across 91 countries and 110 survey components from the joint European Values Study and World Values Survey (EVS/WVS) 2017–2022 dataset.
