Pretest 1: Human-AI Interaction Dialogues – Multi-LLM Shopping Assistant Evaluation

This dataset contains survey responses and interaction metadata from a pilot study (Pretest 1) evaluating human-AI dialogues in a simulated online shopping context. 156 participants recruited via Prolific interacted with one of four large language models (Claude, GPT, Gemini, DeepSeek) configured as shopping assistant chatbots across four product categories (Books, Cell Phones & Accessories, Sports & Outdoors, Office Products). The dataset includes self-reported measures of System Quality, User Satisfaction, Perceived Naturalness, and Ease of Use, as well as task parameters (task type: simple vs. complex; constraint level: low/medium/high), participant demographics, and chatbot experience as well as the full chat transcripts. The study was conducted as part of SNF project 100018-227553 ("Customer Interactions with Generative Artificial Intelligence") to validate the experimental infrastructure and inform LLM model selection for subsequent studies.