ChatGPT Social Support Ratings for Bulimia Nervosa Scenarios

32 bulimia-related patient scenarios were used to evaluate ChatGPT's responses. A psychiatrist and clinical psychologist rated each response across four social support categories using a 5-point Likert scale. The dataset contains the resulting ratings and inter-rater reliability metrics.

Use Cases

• Analyze the distribution of Likert scale ratings across emotional, informational, appraisal, and instrumental support categories.
• Compare inter-rater reliability scores (Cohen’s Kappa) between the psychiatrist and clinical psychologist for each support type.
• Examine the statistical test results (Friedman and Mann–Whitney U) comparing performance across the four social support categories.

Strengths

• Data is derived from 32 structured bulimia-related scenarios based on the Edinburgh Bulimia Test.
• Each response was independently rated by two qualified professionals: a psychiatrist and a clinical psychologist.
• Ratings use a standardized 5-point Likert scale across four defined social support categories.

Limitations

• The dataset is very small at 9.5 KB, containing ratings for only 32 scenarios.
• The data reflects a specific evaluation context and may not generalize to other conditions or AI models.
• The sample data and specific column structure are unavailable for review.

Provenance

Source

Emre Vuraloğlu

Collection Method

Responses generated by ChatGPT-4o to 32 scenarios were rated via structured content analysis by two professionals.

Time Range

null

Freshness

null

Geography

null