Matching Markets in Recruitment
Recently, Amazon has scrapped its AI-driven recruiting algorithm due to bias that the algorithm was unable to self-correct. The algorithm was trained on thousands of examples of resumes from successful and unsuccessful applicants, and because the algorithm looked for familiar patterns, it learned quickly that male applicants were more favorable to female because generally, there are more male applicants than female. While engineers attempted to patch up these bias, they were so ingrained in the algorithm that Amazon ended up shelving their recruitment AI.
Amazon is not the only company trying to use AI for recruiting. The article mentions a handful of other tech companies who are relying on AI to figure out the most optimal candidates from a pool of thousands of applicants. If we think of the recruitment process as a matching market, then it makes sense why companies are trying to outsource job matching to AI.
Each company has a limited number of open roles with different requirements and desired skill set; therefore, companies want to maximize their overall “welfare” (may be profit or employee retention) by matching the most fitting candidate to the correct role. Theoretically, if a company had a “perfect matching” of roles to employee, then they will never need to recruit again. But because, people grow old and lose efficiency or time, and there is an endless number of people interested in working for tech companies, there will always be a better assignment–the grass is always greener. As employees leave the company or move out of a position, there will always be a pool of people waiting for the next matching.
By principle, the matching will end because each round in the matching market will cause the system to lose “potential energy”. However, it also feels like there is an unlimited amount of energy in the recruitment system: even as the “prices” of the roles increase, the applicants’ skills also improve and there will also be a good enough match for the market at the current time.
This perpetual matching market is what makes recruitment season is one of the most stressful times, especially as companies are working towards optimizing and automating this whole process through AI to find the most “optimal” matching.