HackerRank open sourced its ATS. My resume scored 90/100. Oh wait 74. No – 88 (danunparsed.com)

Article summary

HackerRank's open-sourced ATS (Applicant Tracking System) has been tested and found to have inconsistent scoring due to its reliance on Large Language Models (LLMs). The system's grading rubric has been criticized for prioritizing open source contributions and personal projects over work experience. The author's resume scored differently each time it was run through the system, highlighting the non-determinism of the LLMs. This inconsistency raises concerns about the effectiveness of AI-powered hiring tools.

Main themes

• AI-powered hiring tools
• Resume screening
• Large Language Models
• Hiring biases
• Job application filtering
• Recruitment strategies

What commenters say

• The use of AI-powered hiring tools can lead to inconsistent and unfair candidate screening due to the non-determinism of Large Language Models.
• The current hiring pipeline is flawed and prioritizes the wrong qualities, such as open source contributions over work experience.
• The alternative to AI filtering is not to manually review all resumes, but to limit the number of applications and use more effective screening methods.
• Some argue that AI screening tools can still provide a more relevant pool of candidates than random distribution, despite their limitations.
• Others propose that companies should stop accepting applications once they have enough, rather than leaving the posting open for an extended period.
• There is a need for more transparent and deterministic hiring processes to ensure fairness and effectiveness in candidate selection.
• The use of AI in hiring can perpetuate biases and discriminate against certain groups of candidates, such as those without open source contributions or personal projects.
• Some commenters suggest that work sample hiring and blind hiring could be more effective and fair methods for screening candidates.