Context-aware GenAI feedback for fostering student feedback literacy and informing revision decisions in social science writing

Abstract

Generative AI (GenAI) provides immediate feedback but often lacks the contextual grounding for discipline-specific writing such as social science writing. This decontextualization leads to low cognitive engagement and surface-level revisions. This study investigated how a context-aware GenAI, integrated with Bitzer’s rhetorical situation theory, influences student feedback literacy (SFL) and revision decisions. 84 undergraduate students were assigned to either a standard GenAI group (n = 42) using rubric-based feedback, or a context-aware GenAI group (n = 42) using feedback grounded in rhetorical situation theory. GenAI-student interaction logs were analysed using Epistemic Network Analysis to examine structural patterns of SFL. Revision decisions were analysed using thematic analysis and ANOVA. The two groups revealed different structural patterns of SFL. The context-aware GenAI group demonstrated stronger cognitive-behavioural connections, enabling students to evaluate feedback for rhetorical relevance. In contrast, the standard GenAI group showed stronger affective-behavioural connections, where emotional burdens led to behavioural compliance and surface-level revisions. The ANOVA results showed that the context-aware group made significantly more rhetorical-level revisions, while the control group focused on surface-level corrections. This study contributes to the reconceptualization of SFL by revealing the interfering role of affective burdens in the feedback process.