Synthetic Clinical Data for Intersectional Bias Assessment in Mental Health

A synthetic dataset containing demographic and clinical data used to test a classifier's performance in predicting schizophrenia or depression diagnoses. It was created by Marta Maslej for a tutorial presented at ACM FAccT 2022 to assess models for intersectional bias. The dataset is licensed under CC-4Y and hosted on OpenML.

Use Cases

• Testing classifier performance for mental health diagnosis prediction based on demographic and clinical features.
• Assessing intersectional bias in model predictions across demographic subgroups.
• Benchmarking fairness evaluation methods using synthetic clinical data.
• Reproducing or extending the methodology from the ACM FAccT 2022 tutorial.

Strengths

• Dataset is explicitly designed for intersectional bias assessment, a critical task in fair ML.
• Created for a peer-reviewed tutorial presented at a major conference (ACM FAccT 2022).
• License (CC-4Y) is clearly specified, facilitating reuse.

Limitations

• Row count, column definitions, and sample data are unknown, which limits suitability assessment.
• Description metadata is limited; actual data quality and structure require manual inspection after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

OpenML

Collection Method

Synthetically generated for model testing and bias assessment.

Time Range

Created for a tutorial presented in 2022.

Freshness

Last updated date is unknown.

Geography

Spatial coverage is unknown.