Imagine, biases like "good formatting => better; no typos => better; social profiles => better;" etc being put within an AI/ML based resume parser, with an intent to mimic human behaviour of sourcing a resume out of a bundle of resumes, and scoring against a job description. Now, what might go wrong, when

- more such biases creep in to the models, thanks to groupthink, multiple developers/groups and their biases

- such systems end up gating the applicants before a human view is cast on the resume

This is happening now, at scale. And only a very few in the industry seem to be really tuned about such dangerous bias perpetuation engines.