Towards Personalized and Contextualized Code Explanations

Code understanding is a common and important use case for generative AI code assistance tools. Yet, a user's background, context, and goals may impact the kinds of code explanations that best fit their needs. Our aim was to understand the kinds of configurations users might want for their code explanations and how those relate to their context. We ran an exploratory study with a medium-fidelity prototype and 10 programmers. Participants valued having configurations and desired automated personalization of code explanations. They found particular merit in being able to configure the structure and detail level in code explanations and felt that their needs might change depending on their prior experience and goals.