As generative AI becomes increasingly integrated into sensory and consumer research, questions arise about the representational biases these tools may introduce. While synthetic respondents (i.e., “silicon samples”) offer potential benefits for research efficiency, we argue that their routine use may normalise particular models of consumer behaviour: coherent, articulate, and rationally structured responses that systematically filter out the inconsistency and ambiguity characteristic of real consumer decision-making. Adapting the WEIRD critique from behavioural science, we examine whether generative AI embeds specific cultural assumptions and cognitive models that may not reflect actual consumer diversity. Evidence from adjacent domains shows that AI tools can increase research productivity while narrowing the range of questions asked. In consumer research, this manifests as path dependence: upstream use of generative AI in pretesting and design may shape which attributes, hypotheses, and preference narratives advance to human testing. As human and synthetic responses increasingly blend in online panels, researchers must remain cautious about how these tools privilege certain forms of preference articulation over others.
Califano et al. (Mon,) studied this question.