In AI-based educational systems, transparency and understandability are particularly important to ensure reliable human-AI interaction. This paper contributes to the ongoing research on developing analytics for AI-based educational systems by delivering feedback throughout indicators that learners can easily interpret and act upon during their studies. Specifically, this paper introduces a mixed methods study that examines the types of indicators that ought to be incorporated into the feedback offered by an AI-based system designed to help students develop competencies in programming. Building upon prior work in Human-Centered Design, the card sorting technique was used to collect both qualitative and quantitative data from 31 Computer Science students. We created 16 cards that presented students with different indicators to explain the reasoning behind the system’s decisions and feedback. These indicators were displayed in different formats (visual and textual representations; temporal vs. non-temporal and social vs. non-social reference frames). Our goal was to discover the most interpretable and actionable method for delivering feedback to learners. Our study found low consensus among students. Overall, students found indicators based on social comparison to be less actionable and interpretable compared to those without; and textual indicators were perceived as less actionable and interpretable than visual ones.