Algorithmic Intermediation and Bias Mitigation
RCTReview
Traditional public employment matchings suffer from high manual search costs and inefficient screening, while automated algorithmic screening tools risk amplifying historical human biases against underrepresented demographic groups if trained on biased historical hiring data [6, 9].
Picture this
Imagine a digital matchmaker that instantly sifts through thousands of applicant profiles to recommend top fits, like a streaming service recommending movies based on actual viewing preferences rather than genre covers. However, if the recommendation engine is trained on biased past ratings, the engine will stop recommending minority content unless explicitly recalibrated to prioritize diverse options while maintaining fit quality.
What the evidence says
Algorithmic recommendation of workers to employers improved recruit pool quality and increased overall hiring [6]. Algorithmic writing assistance on job seekers' resumes boosted hiring probability by 8 percent [6]. Initial hiring algorithms selected far fewer Black and Hispanic candidates, but re-engineered algorithms successfully increased demographic diversity while maintaining high hiring rates [6].
- Who was studied
- Job seekers and employers participating in large online labor markets and field evaluation settings [6, 9].
- How
- Randomized evaluations testing machine learning recommendation engines and resume writing assistance tools against standard search interfaces, alongside algorithm re-design extensions [6, 9].
What to do
Deploy audited, diversity-recalibrated machine learning matching algorithms within public job placement platforms to automate candidate screening and resume preparation support [6, 9].
From the source
"In an extension, the same authors showed that using a different algorithm increased demographic diversity while keeping hiring rates high [6]."
ENG_Job-search-assistance.pdf