Research
Instrumentalizing ChatGPT for Academic Identity Formation (AIF) in First-Year Composition: A Self-Regulated Learning Theory
Abstract: This article explores the theoretical potential of integrating ChatGPT as a learning coach (ChatGPT-LC) – first proposed for medical students as a life coach – within first-year composition (FYC) courses to support self-regulated learning (SRL) and bolster academic identity formation (AIF). I argue that ChatGPT-LC can play a pivotal role in helping students develop their writing self-efficacy and enhance their writing skills by providing an opportunity for metacognition even as it does not de-center students’ agency as they form their academic identities. The article articulates a SRL model that incorporates instructor guidance, combining insights from early adopters of generative AI in writing pedagogy with advances in educational psychology to position ChatGPT-LC as a classroom aid fostering metacognition, self-efficacy, and exploration of academic community among FYC students.
Status: Under review, Composition Forum
Instrumentalizing ChatGPT for AIF in First-Year Composition: Human Subjects Research Study
In this IRB-approved collaborative research study, students in Santa Clara University’s two-course first-year writing sequence – Critical Thinking & Writing (CTW) 1 and CTW 2 – were led through multiple surveys and interactions with OpenAI’s ChatGPT (GPT-4 series).
Purpose: To determine if ChatGPT-LC enhances students’ reflective practices, metacognitive awareness, self-efficacy in writing, and whether ChatGPT-LC encourages students to view themselves as active participants in the academic community.
Background
Recent research has highlighted the growing role of AI in education, particularly in writing instruction (Hart-Davidson 2018; Miller 2018; Knowles 2024; McKee and Porter 2020; Cummings et al 2024; Ranade and Eyman 2024). Studies have shown that tools like ChatGPT can assist with basic writing tasks but raise concerns about potential over-reliance and the impact on students’ critical thinking and creativity (Bedington et al. 2024; Jamieson 2022). Despite these concerns, AI’s potential for personalized learning and support remains significant. Previous research in the field of educational psychology has emphasized the importance of self-concept and self-efficacy in student success, particularly in writing-intensive courses like first-year composition, or FYC (Barclay et al. 2018; Dulay 2017; Peterson 2015; Çikrıkci 2017; Pajares and Schunk 2005). Additionally, the concept of Professional Identity Formation (PIF) in medical education, as explored by Huang and Lin (2024), suggests that AI tools can play a crucial role in helping students internalize professional norms and practices. This study builds on these findings by using ChatGPT-LC to support students’ development as academic writers and thinkers within the academic community. This qualitative human subjects research study aims to address gaps in the existing literature by exploring the specific impact of AI on academic identity formation in a liberal arts writing course context.
Procedures:
- Initial Survey: All subjects, including control group, complete intake survey assessing current self-concept, self-efficacy, and engagement with academic community. The survey questions derive from relevant surveys (Tennessee Self-Concept Scale; General Self-Efficacy Scale; and National Survey of Student Engagement, respectively)
- Interactions: Throughout each quarter, subjects engage with ChatGPT-LC as they complete writing assignments. They will use the tool to receive personalized feedback, set writing goals, and intermittently reflect on their writing progress. Subjects submit session logs within the ChatGPT interface.
- Ongoing Surveys and Reflections: Subjects provide reflective statements that describe their ongoing experiences with ChatGPT-LC, focusing on how it has influenced their understanding of themselves as writers and their engagement with the academic community.
- Final Survey: All subjects, including control group, complete final survey similar to Initial Survey.
- Data Aggregation & Anonymization: The data collected from the surveys and logs for fall, winter, and spring quarters (2024-2025) are aggregated and anonymized by assigning each subject a randomized alphanumeric ID.
- Data Analysis: Data will be analyzed using a phenomenological approach – coding common themes and patterns in reflective statements, focusing on how interactions with ChatGPT-LC have influenced self-concept, self-efficacy, and academic identity.
- Written Report: Complete article-length report of findings.
Recruitment: In-class recruitment script followed by ad hoc Q&A with research team member who is not their instructor; Informed consent form.
Queer Refractions: Queer Storytelling in the Age of AI
Given AI-generated texts are often marked by a sense of the uncanny, familiar yet eerily strange, this uncanniness queers conventional forms of communication by introducing ambiguity and instability into otherwise structured language forms. Speculative AI is at the nexus of understanding a new age of writing, storytelling, and constructions of (anti)normativity. My monograph project is tentatively titled Queer Refractions.
I use tools like Sudowrite and Squibbler, which both claim to generate full novels in users’ own style in mere seconds, to critically examine the potential and limitations of generative AI to produce queer/non-normative narratives. By generating and then close-reading corpora of creative writing focused on LGBTQ+ themes, I interrogate how these AI-generated texts ultimately fail to reflect the complexity of queer storytelling. Instead, I argue, these present what I call a refraction of queerness, rather than a reflection of it. In line with emerging subfields of queer AI and critical AI, the book exposes the heteronormative constraints embedded in these tools.
I end the monograph with potential fixes for this issue. One such solution I anticipate including based on my background in critical Indigenous studies will be the incorporation of Indigenous ways of knowing into the design of generative AI. By integrating Indigenous epistemologies – which is to say prioritizing relationality, multiplicity, and storytelling as acts of communal memory – I suggest that generative AI could move beyond linear, normative narratives toward more expansive, inclusive approaches to creativity. This reimagined framework not only challenges the dominant cultural assumptions embedded within current AI systems but also opens up possibilities for AI tools to co-create narratives that honor the fluidity and intersectionality of queer identities. The monograph thus advocates for a broader, more ethical engagement with AI in creative writing, urging developers, writers, and publishers to consider how marginalized perspectives can transform the future of emerging technologies.
Generative AI and the Speculative
I investigate the correlation between LLMs and the speculative mode of writing, a genre based in the creative exploration of hypothetical scenarios and future possibilities. Rather than merely serving as tools to inspire speculative thought, I argue that LLMs actively generate speculative narratives, thereby reshaping the very function of the speculative mode in contemporary writing and culture. Advancing recent scholarly threads in critical AI and cultural studies, I argue that LLMs, in their capacity to generate creative content, simultaneously bolster and expand traditional boundaries of writing.
Pulling from design theory, I argue this capacity to simultaneously strengthen and expand is the LLMs’ capacities to buttress and splay. Splayed forms in design introduce new dimensions or expansions while buttresses provide structural reinforcement, ensuring stability and continuity. As this pertains to generative AI and speculative writing, LLMs reinforce traditional narrative structure in that their production is entirely based on their training data. Yet, they also extend these boundaries by producing “new” texts
Preston Taylor Stone, Ph.D. 