Logo image
IAT LLM Associative Interference
Dataset   Open access

IAT LLM Associative Interference

Achraf Cohen and Andrew Kincaid
Zenodo, v1
2026

Metrics

2 Record Views

Abstract

Large Language Models Implicit Association Test Artificial Intelligence Statistics
This dataset contains trial-level responses from large language models evaluated using an adapted Implicit Association Test (IAT). Each row corresponds to a single forced-choice classification trial. Variables include the IAT domain (iat_type), experimental condition (block), stimulus item (item), response options (pairing_A, pairing_B), model identity, and raw model output (choice). Derived variables include valid_response (whether a valid A/B response was produced), choice_clean (parsed response), and task_consistent (whether the response matches the predefined mapping). Analyses are conducted in two stages: modeling response compliance and conditional task consistency to estimate associative interference.
url
IAT LLM Associative InterferenceView
Published (Version of record) link to dataset Open CC BY V4.0

Related links

Details

Logo image