Research Article
ChatGPT as a Cognitive Tool in Sociology: Students' Perceptions and Its Impact on Academic Writing Practices
Synthesis: Malik (2025) interviewed 18 undergraduate sociology students in Indonesia and compared writing samples produced with and without ChatGPT. Students described the tool as a "thinking companion" that lowered the barrier to abstract theory — clarifying concepts, generating openings, and turning a shapeless assignment into an outline. Their writing did get more organized, and 61% used the tool mainly while drafting while 39% used it mainly while revising. The cost they named themselves is the one worth teaching against: fear of losing their academic voice and their critical edge, which produced a set of selective strategies — verifying theory, rewriting output — rather than either refusal or wholesale adoption.
Key Findings
- ChatGPT functioned as conceptual scaffolding. Students used it to clarify abstract sociological concepts, generate ideas and articulate theoretical explanations, and several called it a "thinking companion" that made dense theory approachable.
- Writing strategies shifted toward planning. Students moved from unstructured drafting to outlining, coherence checking and iterative refinement, and said this was the tool's main contribution rather than prose generation.
- The tool's use is bimodal by stage. 61% used it primarily during early drafting and 39% primarily during revision, which means most of the sample treated it as a planning aid rather than a rewriting service.
- Writing samples showed better organization. The comparison found clearer argumentation, tighter theory–evidence links and more coherent paragraph progression in the assisted samples.
- The tension is dependence, not dishonesty. Students worried about losing their voice, weakening their critical thinking and over-relying on AI structure, and answered with selective use, cross-checking of theoretical claims and rewriting of generated text.
What students actually did with it
The pattern in the interviews is specific enough to teach from. Students asked for outlines and section sequences when an assignment was long ("when writing about social change, it helped me build a logical sequence of sections"), asked for more academic formulations of sentences they had already written, and asked for alternative ways to phrase a concept in order to see the differences. They did not describe asking it to write the essay.
That is a defensible use of a text generator in a theory-heavy discipline, and the paper's evidence for it is a comparison of writing samples rather than a rating of quality by blind reviewers. The samples were produced by the same students with and without the tool, so organization improved while the underlying argument came from the student.
Where the interviews turn is accuracy. Several students reported that the theoretical explanations ChatGPT supplied were "not fully accurate" and that they had learned to cross-check claims about theorists. That is a discipline-specific failure: sociology's content is a set of positions in argument, and a model that averages over them produces plausible synthesis that no single theorist would endorse.
What this means for practice
- Instructors: teach the planning use explicitly. Outlining, sequencing and reformulating are the uses students already converge on, and they are the uses that leave the argument with the student.
- Instructors: make the verification step an assessed one. Students in this sample verified theory on their own initiative; the ones who did not had no way to know the explanation was wrong.
- Curriculum designers: put the discipline's own standards into the AI literacy component — which theorist, which tradition, which reading — rather than generic "check the facts" advice.
- Instructors: the voice worry is a real signal. Ask for a reflective note on what the student changed in AI-assisted drafts, which converts an unspoken anxiety into evidence of judgment.
Limitations
One Indonesian sociology program, 18 students, self-selected into interviews, with qualitative coding by a single author. The writing comparison was not blind, and "clearer organization" was judged by the research team rather than external raters or a rubric with reliability data. Recall of use is retrospective. The journal and the study are outside the main English-language sociology-education literature, so the sample's institutional context is described only briefly.
Citation
Malik, A. R. (2025). ChatGPT as a Cognitive Tool in Sociology: Students' Perceptions and Its Impact on Academic Writing Practices. DEIKTIS: Jurnal Pendidikan Bahasa dan Sastra, 5(4).