Intersectional Bias Assessment Training Data for Mental Health Diagnosis

A synthetic dataset containing demographic and clinical data used to train a classifier to predict diagnoses of schizophrenia or depression. It was created by Marta Maslej for a tutorial presented at the ACM FAccT 2022 conference. The dataset is licensed under CC-4Y and is hosted on the OpenML platform.

Use Cases

• Training classifiers for mental health diagnosis based on synthetic demographic and clinical features.
• Assessing algorithmic bias and fairness in predictive models based on the described demographic data.
• Benchmarking bias mitigation techniques in healthcare AI based on the synthetic training data.
• Conducting intersectional analysis of model performance across demographic subgroups as suggested by the dataset's purpose.

Strengths

• Dataset is explicitly designed for fairness and bias assessment in AI, a well-defined purpose.
• Created for a tutorial at a major conference (ACM FAccT 2022), suggesting pedagogical utility.
• Has a clear, permissive license (CC-4Y) for reuse.

Limitations

• Row count, file formats, and column-level documentation are unknown, which limits suitability assessment.
• Data is synthetic, which may limit its realism and direct applicability to real-world clinical settings.
• Last update date is unknown; freshness unverified.

Provenance

Source

OpenML

Collection Method

Synthetically generated for a tutorial.

Time Range

null

Freshness

null

Geography

null