Introducing Asynchronous Proctoring: A Human-Centered Approach to ChatGPT Cheating Detection
Candidates cheating via external tools is now an epidemic. After months of research across thousands of candidates, we’ve found that 1 in 10 developers cheat on their tech assessment or take-home exercise with a tool like ChatGPT.
So Woven is excited to introduce our latest advancement: Human-powered Asynchronous Proctoring is now included for free on all plans.
Without Proctoring, Engineering Spends 30% of Time with Cheaters
Even worse, Woven’s data shows that candidates using those tools without proctoring are 3 times more likely to advance to the next round. For hiring managers not using some type of proctoring, that means 1 in 3 of tech interviews are with someone who cheated on their assessment or take-home.
Algorithmic Detection Doesn’t Work
And to pile on, automated detection methods just don’t work. They’re riddled with both false negatives and false positives. Even OpenAI, with $1b in funding, couldn’t figure out algorithmic detection of ChatGPT usage.
Only with recording-based proctoring of the candidate completing the work can you reliably catch anything but the lowest-effort usage of ChatGPT. It then requires trained, expert humans to use this recording to differentiate between harmless cases like using Stack Overflow from cheating situations where a candidate’s actual tech skills aren’t being tests. Believe us, because we tried very, very hard. Adding async proctoring is expensive for us.
It’s the right move to further our mission to eliminate the gap between talent and opportunity. If our customers can’t trust that a candidate actually performed the work, then they’ll be less likely to save engineering time and less likely to take a bet on a candidate whose performance is much better than their pedigree.
Introducing Asynchronous Proctoring
Our technique involves skilled human experts reviewing every single assessment for signs of external assistance. They watch a live playback of the candidate’s submission, evaluating their behavior and final responses against guidelines our team has defined and tested.
Woven is uniquely positioned to catch this kind of plagiarism. Since we already score candidates with a team of certified human engineers and detailed rubric, ChatGPT detection is only one more rubric item for our team to evaluate.
If you’re using an async assessment that relies on algorithmic scoring, you may be particularly susceptible to candidates using external assistance. Human-powered Asynchronous Proctoring is the only way to retain the advantages of async assessments and still get the signal you need to make great hiring decisions.
Interested in learning more about how we do it?