Toward Automatic Interpretation of Narrative Feedback in Competency-Based Portfolios

Abstract

Self-directed learning is generally considered a key competence in higher education. To enable self-directed learning, assessment practices increasingly embrace assessment for learning rather than assessment of learning, shifting the focus from grades and scores to provision of rich, narrative and personalized feedback. Students are expected to collect, interpret and give meaning to this feedback, in order to self-assess their progress and to formulate new, appropriate learning goals and strategies. However, interpretation of aggregated, longitudinal narrative feedback has been proven to be very challenging, cognitively demanding and time consuming. In this study, we therefore explored the applicability of existing, proven text mining techniques to support feedback interpretation. More specifically, we investigated whether it is possible to automatically generate meaningful information about prevailing topics and the emotional load of feedback provided in medical students' competence-based portfolios (N = 1500), taking into account the competence framework and the students' various performance levels. Our findings indicate that the text mining techniques topic modeling and sentiment analysis make it feasible to automatically unveil the two principal aspects of narrative feedback, namely the most relevant topics in the feedback and their sentiment. This study therefore takes a valuable first step towards the automatic, online support of students, who are tasked with meaningful interpretation of complex narrative data in their portfolio as they develop into self-directed life-long learners.

Leveraging Narrative Feedback in Programmatic Assessment: The Potential of Automated Text Analysis to Support Coaching and Decision-Making in Programmatic Assessment

“… 8–11 Automated text analysis holds promise as a valuable tool to support feedback processes. An initial exploratory technical study 12 demonstrated the technical feasibility of harnessing “topic modelling” and “sentiment analysis” to generate comprehensive overviews of the primary topics embedded within narrative feedback data. Study findings furthermore showed that this approach enabled the evaluation of the sentiment polarity (positive, neutral, or negative), associated with these identified topics.…”