Material for "Compassion Over Warmth: AI Communication Style and Advice Utilization in High-Stakes Decision-Making"

This dataset contains experimental data from two studies examining how AI communication style (Control/Neutral, Warm, Compassionate) influences advice utilization in high-stakes decision-making contexts.
Study 1 (N = 566) employed a medical treatment decision paradigm where participants chose between cancer treatment options after receiving recommendations from "CareAI," an AI assistant. The study used a between-subjects design with three communication style conditions.
Study 2 (N = 563) employed a financial advisory paradigm where participants estimated retirement portfolio survival probabilities for clients after receiving analysis from "WealthAI." The study used a 3 (Communication Style: between-subjects) × 2 (Task Stakes: within-subjects) mixed design with 16 client scenarios (8 high-stakes, 8 low-stakes).
Variables include:
Demographics (age, gender, education)
Pre-task measures: AI experience, health/financial literacy, trust propensity, risk perception, algorithm aversion, need for cognition
Trial-level behavioral data: initial choices/estimates, final decisions, AI recommendations
Post-task measures: perceived warmth, competence, compassion, perceived risk, affective trust
Computed dependent variable: Weight of Advice (WOA)