Introducing Asynchronous Proctoring: A Human-Centered Approach to ChatGPT Cheating Detection

kodiva.ai
September 11, 2024
3 mins
Proctoring

Cheating on technical assessments with tools like ChatGPT has become a widespread issue. After extensive research across thousands of candidates, we’ve discovered that 1 in 10 developers resort to external assistance during their assessments. At Kodiva, we are committed to addressing this challenge. That’s why we are excited to introduce Human-powered Asynchronous Proctoring, now included in all our plans at no extra cost.

The Impact of Cheating on Engineering Time

Without effective proctoring, companies spend valuable time interviewing candidates who have cheated. In fact, Kodiva’s data shows that candidates who cheat are 3 times more likely to advance to the next round, meaning 1 in 3 technical interviews could be with someone who didn’t genuinely complete their assessment. For hiring managers, this results in a significant waste of resources and time.

Why Automated Detection Fails

Despite claims from many platforms, algorithmic detection simply doesn’t work. Automated systems struggle with both false negatives and false positives, making them unreliable for catching more sophisticated forms of cheating. Even OpenAI, with its billion-dollar budget, couldn’t develop a reliable ChatGPT detection method.

Realistically, only recording-based proctoring—where the candidate’s work process is captured and reviewed—can effectively catch the misuse of AI tools. And even then, it requires skilled human reviewers to distinguish between benign actions, like referencing Stack Overflow, and actual cheating. We’ve invested significant resources into this approach because it aligns with our mission: eliminating the gap between talent and opportunity. If employers can’t trust the results of a technical assessment, the entire hiring process becomes compromised.

The Solution: Asynchronous Proctoring

Kodiva’s Asynchronous Proctoring is powered by human experts who review every assessment for signs of external assistance. Our team watches a playback of each candidate’s submission, carefully analyzing their behavior and responses. By integrating this review into our AI + human-scored assessments, we ensure that ChatGPT detection is a seamless part of the process.

Our solution is uniquely positioned to catch plagiarism and cheating because it builds on Kodiva’s existing human evaluation system, where each candidate’s work is reviewed by a certified team of engineers using detailed rubrics. Adding ChatGPT detection is just another layer to our comprehensive scoring process.

Why This Matters for Hiring Teams

For teams using asynchronous assessments that rely solely on algorithmic scoring, there’s a higher risk that candidates could be using external assistance. Human-powered Asynchronous Proctoring is the only reliable way to maintain the advantages of asynchronous assessments while ensuring that you’re getting accurate, trustworthy results. This approach ensures that hiring managers have the signal they need to make informed decisions and reduce wasted engineering time.

Want to learn more about how Kodiva’s Human-powered Asynchronous Proctoring works? Reach out today! We’re committed to helping you make the best hiring decisions while maintaining integrity in your technical assessments.