Lexical Alignment in Argumentative Dialogues: Study Data

This dataset contains the anonymised data and the analysis code of our between-subjects study (N = 31) on lexical alignment in argumentative dialogues. The participants and a partially wizarded conversational agent planned a route in a fictional rescue travel planning task. The agent was either lexically aligning or misaligning, and it always countered the participant's proposal with an argument for another destination. The dataset contains the dialogues, the questionnaire responses on trust, cognitive load, and the propensity to trust AI-based systems, the coding of agreement, information uptake, and task success, and the lexical alignment measures. The dataset contains only the 31 analysed participants, and it contains no Prolific IDs, names, or IP addresses. The Ethics Committee Information and Computer Science of the University of Twente approved the study (reference 230463). README.md describes the files. We report the study in the following paper, and in Chapter 5 of the dissertation of Sumit Srivastava (University of Twente): Sumit Srivastava, Mariët Theune, Alejandro Catalá, and Chris Reed. 2024. Trust in a Human-Computer Collaborative Task With or Without Lexical Alignment. In Adjunct Proceedings of the 32nd ACM Conference on User Modeling, Adaptation and Personalization (UMAP Adjunct '24), 189–194. https://doi.org/10.1145/3631700.3664868