What does Woven do to deter cheating?
What does Woven do in the event of cheating or plagiarism?
If you work in hiring long enough, you'll hear horror stories about cheating. So how does Woven identify and deter cheating?
We currently have several ways of identifying non-original code:
- We use an algorithm that compares each response to others from previous candidates and assigns it a “plagiarism score”. It also gives us some flashing warnings when the score is high (usually above 90%)
We are also alerted when big chunks of code was pasted into the solution box. We have the ability to see the entire playback of their solution (i.e. we can see exactly how they behaved while they were writing up their response, and whether they had long idle periods). - Sometimes our evaluation team is also able to identify unoriginal code just by looking at it, since the quality may vary between the code that they copied from somewhere else, and the code that they actually wrote
- We consistently scrape the internet for leaked Woven solutions and prompts and endeavor to remove them
- We regularly make changes to our scenarios that make them more difficult to find and copy exactly without alerting us to likely plagiarism
Whenever we identify non-original responses, we take several actions:
- We flag the code and explain the situation in their final recommendation
- We score the scenario which had non-original code as 0
- If available, we also add the source from where we believe they may have copied the code from
It's definitely possible for dedicated cheaters to get around all of these layers. Like most security, our goal is to make it so tough/expensive to cheat well with our scenarios that they're more likely to cheat on something easier or another job.
One thing you can do to avoid this kind of fraud is ask a couple of questions about the candidate's solution in your next interview. It usually becomes obvious quite quickly whether the candidate was responsible for their own work.
