Member-only story
Why Resume Ranking is a Bad Idea?
If you are applying for jobs and working on editing your resume, it might come to your mind the famous recruiting myth:
“Recruiters take 6-seconds on average to read your resume.”
Though it may not be exactly true, it does manifest the challenge of effective evaluation of candidates.
Amazon used a resume ranking system in 2014 to rank top candidates in order to automate the hiring process, until they found out that the engine did not like women.
But by 2015, the company realized its new system was not rating candidates for software developer jobs and other technical posts in a gender-neutral way.
To understand the problem, we’ll cover the baseline of how the model works and dive deeper into what can go wrong and why.
At a high level, if the objective of the algorithm is to evaluate the “fit” of a candidate to a role, the model would take the resume of the candidate as input, feed it into a model and predict how good of a “fit” the candidate is according to past successful matches between candidates and jobs. The raw inputs from the resume can be featurized using word embeddings.
The problem was:
Amazon’s computer models were trained to vet applicants by observing patterns in resumes submitted to the company over a 10-year period. Most came from men, a reflection of male dominance across the tech industry.
