Modified Ebbinghaus (AI-agent detection trap)

The Modified Ebbinghaus is a detection item, not a research paradigm: it separates genuine human respondents from vision-enabled AI agents completing surveys autonomously. It exploits training-data overfitting on canonical visual illusions, following Ullman (2024) via the Cognitive Trap Framework of Affonso (2026, Journal of Consumer Research).

Participants see two central circles — one blue, one red — each surrounded by a ring of context circles. In the classic illusion two equal centrals look different depending on the size of their neighbours; here one central is objectively bigger, but it is ringed by large context circles (making it look smaller) so the figure still reads as roughly equal. A human ignores the neighbours and picks the bigger central almost immediately. A vision-language model, having memorised the canonical "the circles are equal" response, tends to report "the same" even though the sizes clearly differ.

In the reference data (Affonso 2026), this trap produced ~78% agent-failure with ~96% human pass, and remained effective across a two-year model span, making it a durable item.

Deploy as part of a battery, not alone. The recommended screening rule is to flag respondents who fail 3 or more of 5 such traps. Which colour is objectively bigger (blue or red), the sizes, and left/right placement are drawn at random per session from a pool of validated figures, so the correct answer cannot be memorised. See the Deployment notes below.