Hierarchical Bayesian Models to Mitigate Systematic Disparities in Prediction with Proxy Outcomes
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912736699482112 |
|---|---|
| author | Mikhaeil, Jonas Gelman, Andrew Greengard, Philip |
| author_facet | Mikhaeil, Jonas Gelman, Andrew Greengard, Philip |
| contents | Label bias occurs when the outcome of interest is not directly observable and instead, modeling is performed with proxy labels. When the difference between the true outcome and the proxy label is correlated with predictors, this can yield systematic disparities in predictions for different groups of interest. We propose Bayesian hierarchical measurement models to address these issues. When strong prior information about the measurement process is available, our approach improves accuracy and helps with algorithmic fairness. If prior knowledge is limited, our approach allows assessment of the sensitivity of predictions to the unknown specifications of the measurement process. This can help practitioners gauge if enough substantive information is available to guarantee the desired accuracy and avoid disparate predictions when using proxy outcomes. We demonstrate our approach through practical examples. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_00639 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Hierarchical Bayesian Models to Mitigate Systematic Disparities in Prediction with Proxy Outcomes Mikhaeil, Jonas Gelman, Andrew Greengard, Philip Methodology Applications Label bias occurs when the outcome of interest is not directly observable and instead, modeling is performed with proxy labels. When the difference between the true outcome and the proxy label is correlated with predictors, this can yield systematic disparities in predictions for different groups of interest. We propose Bayesian hierarchical measurement models to address these issues. When strong prior information about the measurement process is available, our approach improves accuracy and helps with algorithmic fairness. If prior knowledge is limited, our approach allows assessment of the sensitivity of predictions to the unknown specifications of the measurement process. This can help practitioners gauge if enough substantive information is available to guarantee the desired accuracy and avoid disparate predictions when using proxy outcomes. We demonstrate our approach through practical examples. |
| title | Hierarchical Bayesian Models to Mitigate Systematic Disparities in Prediction with Proxy Outcomes |
| topic | Methodology Applications |
| url | https://arxiv.org/abs/2403.00639 |