This repository contains the rating data supporting the study “Integrating AI avatars into intelligent cockpits: The effect of communicative content and visual appearance on usage intention.” The study examined how communicative content (negative vs. positive valence; humorous vs. non-humorous delivery) and visual appearance (avatar gender and appearance style) influence drivers’ intention to use in-vehicle AI assistants and avatars in a driving scenario involving impeded progress. The data are organized in the data/ directory, which contains four Excel files: - 1_experiment1_18experts_manipulationcheck.xlsx: Ratings from 18 human factors and ergonomics experts who evaluated the valence and humor of 16 communicative sentences. The file contains 288 observations and includes content conditions, sentence identifiers, valence scores, and humor scores. - 1_experiment1_241participants.xlsx: Data from the text-based survey in Experiment 1. A total of 241 participants evaluated 16 descriptions of AI assistant communication across four content conditions. The file contains 3,856 observations, including emotional valence ratings and intention-to-use ratings. - 2_experiment2_102participants_real human appearance.xlsx: Data from Experiment 2, in which 102 participants evaluated prerecorded videos simulating male and female AI avatars with real human appearances. Each participant completed 32 evaluations across content and avatar-gender conditions, yielding 3,264 observations of intention to use. - 3_experiment3_102participants_animated human appearance.xlsx: Data from Experiment 3, which followed the same design as Experiment 2 using animated human appearances. The file contains 3,264 intention-to-use observations from 102 participants. Each row represents one evaluator’s response to one sentence or video, rather than an average across items. All rating variables use 7-point scales. Participant identifiers are numeric and local to each dataset; matching identifiers across experiments do not identify the same person. Experiments 2 and 3 involved separate participant samples, so appearance style is a between-participants factor in their combined analysis. Please see the accompanying README for variable definitions, scale anchors, appearance-label mappings, and guidance on interpreting the identifiers.