Reflective writing is known to support the development of students' metacognitive skills, yet learners often struggle to engage in deep reflection, limiting learning gains. Although large language models (LLMs) have been shown to improve writing skills, their use as conversational agents for reflective writing has produced mixed results and has largely focused on providing feedback on reflective texts, rather than support during planning and organizing. In this paper, inspired by the Cognitive Process Theory of writing (CPT), we propose the first application of LLMs to the planning and translation steps of reflective writing. We introduce Pensée, a tool to explore the effects of explicit AI support during these stages by scaffolding structured reflection planning using a conversational agent, and supporting translation by automatically extracting key concepts. We evaluate Pensée in a controlled between-subjects experiment (N=93), manipulating AI support across writing phases. Results show significantly greater reflection depth and structural quality when learners receive support during planning and translation stages of CPT, though these effects reduce in a delayed post-test. Analyses of learner behavior and perceptions further illustrate how CPT-aligned conversational support shapes reflection processes and learner experience, contributing empirical evidence for theory-driven uses of LLMs in AI-supported reflective writing.
Figures & tables
Figure 1: Interface of Pensée consisting of two screens. Left: A CA guides learners in the planning phase, with key concepts automatically being extracted to support translation . Right: Writing area where learners compose their text using key concepts and receive automated feedback on the text structure during the review stage.
Figure 2: Left: the reflection cycle from Glogger et al. [ 10 ] , starting from top-left, used in designing and steering the model behind our CA. Right: the architecture for the different LLM agents used in the planning and translation stages in our tool.
Figure 3: Comparing the depth score of reflections across different stages and groups.
Figure 4: Self-reported perception metrics by group (Mean ± SEM).
Reflection, the ability to revisit and revise prior reasoning, is central to how humans improve their answers. Large language models (LLMs) are increasingly prompted to "reflect," yet whether this resembles human revision remains unclear. We introduce the Human-LLM Reflection Framework (HRF), a controlled two-pass protocol comparing human and LLM revision under identical conditions across self-, peer-, and cross-agent settings. Using an information-theoretic analysis based on per-iteration cross-entropy reduction, we find two failure modes of LLM reflection. On objective tasks with finite answer spaces, reflection yields near-zero information gain (Delta I approx 0), behaving as neutral re-generation indistinguishable from re-sampling. On subjective tasks, it yields significant negative gain (Delta I < 0), moving predictions away from the target. Human revision, by contrast, yields positive gain in both settings. Cross-agent experiments localize the failure to the revision step, not input quality: LLMs degrade even high-quality human responses. Diagnostic analyses (revision conditioned on first-pass correctness, and oracle-guided revision against a random-reshuffle baseline) show that which sub-step dominates varies by task and by model rather than reducing to a single mechanism: self-error detection is present on objective multiple-choice tasks but weak on subjective ones, and recovery under an oracle error signal exceeds the baseline for some models and falls below it for others. The unifying account is structural: without external information, self-conditioned revision cannot reduce uncertainty about the target, so LLM reflection is better understood as conditioned re-generation than as genuine error-driven revision.
Large Language Models (LLMs) often produce explicit reflective traces during complex reasoning, accompanied by anthropomorphic markers such as wait, hmm, and alternatively. Although these markers are commonly used as visible indicators of reflection, their mechanisms remain unclear, which leaves the risk of overthinking associated with redundant and repetitive reflection markers. In this work, we revisit anthropomorphic reflection markers, examining their necessity for reasoning and role in the reflection. We suppress these markers through prompt-level and token-level interventions, and analyze their effects on task performance across four benchmarks and two model scales. Our results show that anthropomorphic markers are not uniformly necessary for reasoning performance: suppressing them can preserve or improve performance in several settings, especially under larger sampling budgets. Meanwhile, marker suppression does not necessarily remove reflection behavior, as models can still perform marker-free verification. These suggest that anthropomorphic markers tend to be surface cues rather than reliable proxies for reflection itself, and motivate future research on reasoning mechanisms beyond explicit marker patterns.
Dialogue systems have long supported learner reflections, with theoretically grounded, rule-based designs offering structured scaffolding but often struggling to respond to shifts in engagement. Large Language Models (LLMs), in contrast, can generate context-sensitive responses but are not informed by decades of research on how learning interactions should be structured, raising questions about their alignment with pedagogical theories. This paper presents a hybrid dialogue system that embeds LLM responsiveness within a theory-aligned, rule-based framework to support learner reflections in a culturally responsive robotics summer camp. The rule-based structure grounds dialogue in self-regulated learning theory, while the LLM decides when and how to prompt deeper reflections, responding to evolving conversation context. We analyze themes across dialogues to explore how our hybrid system shaped learner reflections. Our findings indicate that LLM-embedded dialogues supported richer learner reflections on goals and activities, but also introduced challenges due to repetitiveness and misalignment in prompts, reducing engagement.
Paras Sharma, YuePing Sha, Janet Shufor Bih Epse Fofang +6
University of Pittsburgh Pittsburgh, Pennsylvania, USA · Carnegie Mellon University Pittsburgh, Pennsylvania, USA · Bowie State University Bowie, Maryland, USA