Adverse impact reduction and job performance optimization via pareto-optimal weighting: A shrinkage formula and...

Area of Science:

• Industrial-Organizational Psychology
• Quantitative Psychology
• Machine Learning Applications

Background:

• The Pareto-optimal weighting approach is valuable for reducing adverse impact and enhancing diversity in personnel selection.
• This method generates hiring solutions reflecting a diversity-job performance trade-off, potentially improving selection outcomes and minority applicant offers without compromising job performance.

Purpose of the Study:

• To address the issue of shrinkage in Pareto solutions upon cross-validation.
• To develop methods for approximating diversity-job performance trade-off curves that are robust to shrinkage.

Main Methods:

• Introduction of a shrinkage approximation formula for the Pareto trade-off curve.
• Derivation of a novel regularization technique for Pareto-optimal predictor weights, adapted from machine learning (elastic net).
• Evaluation of both methods using Monte Carlo simulation.

Main Results:

• The proposed shrinkage approximation formula provides a way to estimate shrinkage across the Pareto trade-off curve.
• The regularization technique yields predictor weights less susceptible to shrinkage.
• Both methods were evaluated for their effectiveness in accounting for shrinkage.

Conclusions:

• The developed shrinkage approximation and regularization techniques offer improved methods for approximating diversity-job performance trade-off curves in personnel selection.
• Recommendations are provided for practitioners to account for shrinkage when utilizing Pareto-optimal weighting.