The Synthetic Persona Fallacy in HCI-Q Methodology

Engineering notes

Key topics: autonomous driving, large language model. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Designing platforms for multi-cultural and multilingual political deliberation requires a deep understanding of citizens' subjective viewpoints and potential points of friction. While Q-methodology is a robust statistical tool for uncovering these qualitative archetypes (recurring patterns of users' behaviours, motivations, or attitudes), its manual execution is highly resource-intensive. This paper investigates whether Large Language Models (LLMs) can automate Q-methodology by acting as demographically grounded ``Synthetic Citizens''. Grounded in the context of MultiPoD - a European initiative building cross-cultural citizen assemblies - we designed a policy scenario, and used it to collect baseline concourse and Q-sort data from human crowdworkers. Subsequently we tasked LLM-driven ``Synthetic Users'', prompted with matching demographic profiles, to perform the same tasks. Through a comparative Factor Analysis, we evaluated the statistical correlation between human and AI-generated viewpoints. Evaluating the algorithmic fidelity between human crowdworkers (N=62) and simulated AI agents, we found a profound lack of structural correspondence. Rather than replicating human viewpoint pluralism, synthetic users exhibited severe variance collapse, defaulting to a rigid, homogeneous consensus. Furthermore, LLMs displayed distinct "normative amplification'' and "coherence biases'', failing to simulate the human's relational and emotional motivations. Consequently, we argue that LLMs cannot safely replace human subjectivity in qualitative design research. Instead, we propose re-conceptualizing synthetic users as a ``sterile baseline'' within a subjectivity wind tunnel: a tool where human deviations from the AI's predictable consensus serve to actively pinpoint the real cultural, emotional, and sociotechnical friction in deliberative systems.

Links and sources

Comments