Closing the Loop: Learning to Generate Writing Feedback via Language Model Simulated Student Revisions

Abstract

Providing feedback is widely recognized as crucial for refining students' writing skills. Recent advances in language models (LMs) have made it possible to automatically generate feedback that is actionable and well-aligned with human-specified attributes. However, it remains unclear whether the feedback generated by these models is truly effective in enhancing the quality of student revisions. Moreover, prompting LMs with a precise set of instructions to generate feedback is nontrivial due to the lack of consensus regarding the specific attributes that can lead to improved revising performance. To address these challenges, we propose PROF that PROduces Feedback via learning from LM simulated student revisions. PROF aims to iteratively optimize the feedback generator by directly maximizing the effectiveness of students' overall revising performance as simulated by LMs. Focusing on an economic essay assignment, we empirically test the efficacy of PROF and observe that our approach not only surpasses a variety of baseline methods in effectiveness of improving students' writing but also demonstrates enhanced pedagogical values, even though it was not explicitly trained for this aspect.

Figures (7)

• Figure 1: PROF Pipeline: The depicted figure illustrates the iterative optimization algorithm used in our approach. At each iteration $t$, the feedback generator $M_{t}$ generates multiple feedback samples, which are then evaluated for their effectiveness using the student simulator and then gpt-4 as a judge. These evaluations are used to establish preference relations over feedback using the quality of the corresponding revised essays , which are subsequently used to update the parameters of $M_{t}$ via DPO NEURIPS2023_a85b405e, resulting in the updated version $M_{t + 1}$.
• Figure 2: Temperature and sentence-level modifications
• Figure 3: The quality of the revised essay by student simulators vs. actual students.
• Figure 4: Evolution of segments
• Figure 5: Progression of problem/solution elements with refinement iterations.