Discrepancies from Pre-Registration
Our studies were pre-registered on AsPredicted (Study 1, Study 2). To maintain a focused presentation of analysis results and prevent scope creep, several analyses and presentation formats deviate from the initial pre-registration documents. The table below details these deviations and their justifications.
Primary Simple Linear Regression Models (without Covariates)
The following tables present the simple linear regression models (without demographic covariates) for the primary relationships reported in the main manuscript for Questions 1 and 2: Self-Reported Religiosity predicting Frequency of Seeking Moral Advice from AI Chatbots, and Religious Behavior Score predicting Frequency of Seeking Moral Advice from AI Chatbots.
Pilot Studies
Two pilot studies provided initial evidence for the relationship between religiosity and seeking moral advice from AI chatbots. Pilot 1 (N = 30, November 22, 2024) and Pilot 2 (N = 106, January 09, 2025) recruited participants through Prolific. In both pilot studies, seeking moral guidance from AI chatbots was assessed with the item “How much do you usually seek moral guidance from Chatbot/AI assistant?” (1 = None at all, 5 = A great deal), and self-reported religiosity was measured on a scale of 1 (Not at all) to 4 (Very). Demographic characteristics for the pilot studies are presented below.
Pilot Study Correlations: Self-Reported Religiosity → Seeking Moral Guidance from AI
Sources of Moral Advice
Source Ranking Plots
Frequency of Seeking Moral Advice
Interest in Seeking Moral Advice
Accessibility of Moral Advice Sources (Study 2 Only)
Source Correlation Heatmaps
Frequency of Seeking Moral Advice
Interest in Seeking Moral Advice
Study Variable Correlation Analysis
Correlation Analyses: Predictor → AI Moral Interest
The following analyses examine the two supplementary predictor–outcome relationships not covered in the main body: Religious Behavior Score → Interest and Self-Reported Religiosity → Interest. Correlations are presented with predictor first, then study.
Regression Analyses: Predictor → AI Moral Interest
We report regression analyses for both supplementary outcome models: Religious Behavior Score → Interest and Self-Reported Religiosity → Interest.
Religious Behavior Score → AI Moral Interest
Self-Reported Religiosity → AI Moral Interest
Moderation Analyses
The main body (Question 4) tested whether access to moral advice sources moderates the Religious Behavior Score → frequency relationship. Here we report supplementary moderation analyses for three additional predictor–outcome combinations: Self-Reported Religiosity → Frequency, Religious Behavior Score → Interest, and Self-Reported Religiosity → Interest. For each, we test both AI chatbot access and overall access to non-AI sources as moderators (Study 2 only).
Self-Reported Religiosity × Access → AI Moral Frequency
AI chatbot access did not significantly moderate the Self-Reported Religiosity → frequency relationship (b = 0.01, t(343) = 0.4, p =0.686). Overall access to non-AI sources did not significantly moderate this relationship (b = 0.08, t(343) = 1.45, p =0.149).
Religious Behavior Score × Access → AI Moral Interest
AI chatbot access did not significantly moderate the Religious Behavior Score → interest relationship (b = 0.04, t(343) = 1.19, p =0.234). Overall access to non-AI sources did not significantly moderate this relationship (b = 0.12, t(343) = 1.92, p =0.056).
Self-Reported Religiosity × Access → AI Moral Interest
AI chatbot access did not significantly moderate the Self-Reported Religiosity → interest relationship (b = 0.02, t(343) = 0.67, p =0.501). Overall access to non-AI sources did not significantly moderate this relationship (b = 0.09, t(343) = 1.6, p =0.111).