Rationality of AI Custobots and applicability of the EU “average consumer” test

Determine whether AI-driven purchasing agents ("Custobots") are more rational than human consumers in terms of exhibiting fewer behavioral biases in transactional decision-making, and ascertain whether the European Union consumer law "average consumer" benchmark (a consumer who is reasonably well-informed, reasonably observant, and circumspect) can be validly applied to AI agents or whether a distinct "average Custobot" benchmark must be formulated.

Background

The article examines how the rise of autonomous AI agents will challenge foundational premises of EU consumer law, particularly the human-centric "average consumer" standard used to assess unfair commercial practices. It contrasts human behavioral biases with emerging findings in machine psychology and AI psychometrics, noting mixed evidence on whether LLM-based agents exhibit human-like decision biases.

Within this context, the authors explicitly state that it is currently unresolved whether AI-driven Custobots are more rational than humans. They further highlight uncertainty over whether the established average consumer test can be applied to AI agents or if a new "average Custobot" test is necessary, given factors such as the variability of AI agent behavior depending on prompts and indications of deceptive tendencies in some agents.