ChatGPT Incorrectness Detection in Software Reviews
/ Authors
/ Abstract
We conducted a survey of 135 software engineering (SE) practitioners to understand how they use Generative Al-based chatbots like ChatGPT for SE tasks. We find that they want to use ChatGPT for SE tasks like software library selection but often worry about the truthfulness of ChatGPT responses. We developed a suite of techniques and a tool called CID (ChatGPT Incorrectness Detector) to automatically test and detect the incorrectness in ChatGPT re-sponses. CID is based on the iterative prompting to ChatGPT by asking it contextually similar but textually divergent questions (using an approach that utilizes metamorphic relationships in texts). The underlying principle in CID is that for a given question, a response that is different from other responses (across multiple incarnations of the question) is likely an incorrect response. In a benchmark study of library selection, we show that CID can detect incorrect responses from ChatGPT with an Fl-score of 0.74 - 0.75.
Journal: 2024 IEEE/ACM 46th International Conference on Software Engineering (ICSE)