You can follow my new X account for posts like this: https://x.com/utkusen_en
Since the pandemic, most technical interviews are conducted remotely. This shift has created opportunities for cheating, particularly with the development of advanced LLM models. Cheating in online technical interviews is happening more often than you might think. If you’re running a technical interview, it’s a good idea to be ready to keep AI help from slipping in. Here are some of my thoughts around this:
How to Spot AI Cheaters
Delayed Responses Followed By Perfect Answers: People who use AI to cheat in interviews often start answering questions with a delay, then suddenly come up with a perfect answer. However, those who genuinely know the answer don’t typically stutter at the beginning. Even if they’re gathering their thoughts, they usually talk through the process.
It’s not natural to hear something like, “Hmm, I’m compiling my thoughts now… [7 seconds pass] Okay, here’s the perfect answer.” If it feels unnatural, trust your gut. If every answer requires a 7-second pause, something fishy is likely going on.
Talking About Unrelated Things: When you ask ChatGPT a question, it often provides a brief overview of the subject. However, humans in interviews typically don’t do this — they answer the question directly. If a candidate gives a short introduction to the topic or starts talking about something else first, it might be suspicious. Trust your gut if it feels unnatural.
Eye Movements: People usually look around when they talk — at the screen, to the side, or down. Someone using AI to cheat might stare at one spot and move their eyes as if they’re reading.
Tips for Preventing AI Cheating
- Avoid asking questions with textbook answers. Instead, provide a context and let candidates discuss it. Use scenario-based questions to assess their understanding.
- If something seems suspicious, interrupt frequently and ask a follow-up question. If they’re shocked and can’t provide a meaningful answer, there might be a problem.
- Ask “why?” and “why do you think..” questions often. This helps you understand how much they really know and if they’re just repeating memorized information.
- Share your screen and ask questions from there. Cheaters often use speech-to-text tools, so displaying questions visually may catch them off guard (though some might be prepared).
- Present a diagram or visual aid and ask related questions. If they’re using AI, their responses are likely to be irrelevant.