Conference Agenda

The aim of the study was to examine a) the extent to which AI models
can produce high-quality thematic analyses of interviews with students with
learning disabilities and ADHD regarding their difficulties and sources of
support throughout school and University, and b), whether the thematic
analyses produced by AI models are comparable to those produced by
humans (experts and non-experts). This was part of a larger study
investigating school and University experiences of young adults with dyslexia
and ADHD regarding their encountered difficulties, strengths and sources of
support. Participants were 14 Higher Education students studying in Athens,
Greece who had a diagnosis of learning disabilities and/or ADHD. The
students participated in individual semi-structured interviews. The data were
analysed using a thematic analysis method by faculty members and student
assistants and through AI models. Four categories of thematic analyses will
be compared: (a) analyses produced by an expert (psychology faculty
member), (b) analyses produced by a pre-trained AI model (LLaMA), (c)
analyses generated by the BART-base model, fine-tuned specifically to the
interview data, and (d) thematic analyses extracted by the psychology
assistants who conducted the interviews. The quality of the analyses will be
assessed through automatic metrics (ROUGE-1/2/L, BERTScore, etc.), which
will compare each analysis with the corresponding experts’ analysis in terms
of lexical and semantic similarity and human evaluation by two independent
experts (faculty members), who will rate each analysis based on four criteria:
faithfulness, coverage, coherence, and style, on a five-point Likert scale. The
findings are expected to provide insights into the extent to which AI models
can support qualitative data analysis and the development of assistive tools
for researchers and educators working with students with special educational
needs.

This paper explores the use of generative AI (GAI) in qualitative research through the metaphor of the Intergalactic Bummer Train, a silly but serious conceptual vehicle proposed to name the affective, ecological, material, conceptual, and colonial conditions of contemporary inquiry (More Worlds Collective, 2015). As qualitative researchers increasingly engage with GAI, we find ourselves riding “a machinery of depletion” (More Worlds Collective, 2015, p. 11) that reshapes the terrain of knowledge production.

The metaphor offers a multidisciplinary lens for understanding this shifting terrain and imagining alternative futures. Intergalactic signals the multiplicity of worlds, ways of knowing and being, and relations that GAI constrains and enables. Bummer attunes to ecological collapse and planetary emergency, naming the feeling of unknown but inevitable destruction. Train invokes setter-colonial infrastructures and enduring multispecies violences, invoking the tracks laid to progress racialized capitalistic genocide. Riding this train—whether we remember boarding or not—carries us toward an uncertain destination.

Qualitative researchers must consider the assemblage of AI-assisted research as a moving system that carries us along, that increasingly mediates our practices, and institutional engagements. Thinking with the Intergalactic Bummer Train opens space to recalibrate how we travel—to research differently, through renewed commitments of relationality and critically reflexive stances towards institutions. Rather than simply applying GAI to existing methods, we call for methodological experimentation that embraces the More Worlds Collective’s (2025) notion of terraformatics, an ethico-political methodology that seeks to humbly and specifically limit harm to the conditions of place-based existence through a range of methods.

This is not a call to disembark but to “stay with the trouble” (Haraway, 2016, p. 1), to ride the bumps and turbulence as part of inquiry, to engage multispecies relationality inside and out, and to imagine “another kind of study” (More Worlds Collective, 2025, p. 13).

This paper critically reflects on the use of generative AI in qualitative analysis. It draws on an experiment in which Microsoft Copilot was used to do an initial analysis of public discourse on autism from the Trump administration - specifically, briefings by Robert F. Kennedy Jr. (April 2025) and Donald Trump (September 2025). Using Fairclough’s three-dimensional model of Critical Discourse Analysis, we prompted Copilot to examine these texts and their reception by the media and the inclusive education community. Our aim was to assess whether AI could support nuanced, context-sensitive engagement with language, power and ideology.

We found that the AI-generated analysis suffered from consistent gaps in depth and contextual sensitivity. Key theoretical framings, such as the medical model of disability, were omitted. AI also struggled to synthesise commentary, interpret ideological tensions and engage with silence and exclusion, and particularly around neurodiversity and inclusive education. These omissions were most pronounced when prompts included multiple sources or required interpretive nuance.

To address these limitations, we refined our approach by segmenting tasks and embedding theoretical cues, which improved coherence but still required substantial human intervention from a transdisciplinary research team. Our findings suggest that AI may be able to scaffold qualitative inquiry but cannot replace the interpretive labour essential to critical analysis - especially in politically charged contexts. This paper contributes to emerging debates on AI in qualitative research by foregrounding what AI routinely misses: nuance, contradiction and the ideological work of language. We argue for a reflexive, cautious integration of AI tools, and call on qualitative researchers to remain attentive to the epistemic risks of automation - particularly when analysing discourse which shapes public understanding of topics such as disability and inclusion.

This paper discusses the sociomaterial implications for students’ academic writing in higher education when considering Generative Artificial Intelligence (GenAI). Academic writing has been a means of assessing knowledge in higher education since the 19th century, through essays, exams, and other written products (Kruse, 2006). Previous research on student and GenAI co-writing is contradictory. GenAI is at best considered a digital tutor, a dialogic partner or at worst a crutch that leaves students with little knowledge or incentive to write themselves (Flenady & Sparrow, 2025; Giannakos et al., 2024). Sociomaterial theories generally agree that humans and machines comingle and shape pedagogical practices in different ways (Fawns, 2022). This presentation speculates that human-material collectives become shaped by humans, AI technologies, contexts (such as higher education pedagogy), and affects (including ethics). Shifting and unpredictable connections within human-material collectives leave higher education with several potential futures, where academic writing may become nurtured, diminished, or accentuated. The presentation opens up for joint speculations on the future of collective human and GenAI academic writing in higher education. The paper is written within the context of the CO-WRITE project (2025–2027), which explores students’ collaborative academic writing in hybrid learning spaces in higher education.

References:

Fawns, T. (2022). An entangled pedagogy: Looking beyond the pedagogy–technology dichotomy. Postdigital Science and Education, 4, 711–728. https://doi.org/10.1007/s42438-022-00302-7

Flenady, G., & Sparrow, R. (2025). Cut the bullshit: why GenAI systems are neither collaborators nor tutors. Teaching in Higher Education, 1–10. https://doi.org/10.1080/13562517.2025.2497263

Giannakos, M., Azevedo, R., Brusilovsky, P., Cukurova, M., Dimitriadis, Y., Hernandez-Leo, D., Järverlä, S., Mavrikis, M. & Rienties, B. (2024). The promise and challenges of generative AI in education. Behaviour & Information Technology, 1–27. https://doi.org/10.1080/0144929X.2024.2394886

Kruse, O. (2006). The Origins of Writing in the Disciplines: Traditions of Seminar Writing and the Humboldtian Ideal of the Research University. Written Communication, 23(3), 331-352. https://doi.org/10.1177/0741088306289259

As I discussed and argued in my previous work, the perspective of “Spaceship Earth” in US international education of the 1960s-1970s, while departing from love and care of the common world, nonetheless actualized egocentrism and technocratic control of the Other. Curiously, such a paradox is not rare to find in contemporary Chinese international schooling. This article begins with a case study of a summa cum laude graduate from an international high school in China, who was taken by his teachers and himself as a model for international-minded citizens. The discourse analysis focuses on how the campus culture and the international courses he wholeheartedly took shaped this model global citizen’s thinking and acting of “empathy” as a set of devalued, rational methods to solve problems and improve efficiency. The second part of this article historicizes how “international mindedness” (including empathy) as the educational goal of the International School Association and IB curriculum turned from a cosmopolitan value into a scientific method that works as a transnational “currency” that is also named “competency.” Thoughts of Kant, Dewey, and Piaget and their relations are explored as the epistemic conditions that historically make such a change possible. This article combines discourse analysis of a case study and a history of the present to shed light on the blurred boundary between human and machine when empathy is developed as a scientific method of problem-solving and peace-making in international education. It concludes by calling for another possible understanding and acting of empathy from Chinese ethics and aesthetics that integrates rationality into an ontology based on情 (Qing, insufficiently translated as feelings).