From bullshit to friction: a dialogic and distributed cognition view on LLMs in reflexive QDA

Abstract

The integration of Large Language Models (LLMs) into qualitative research has sparked significant debate, in particular, in terms of their capacity for reflexivity, which is generally understood to be a researcher’s active engagement in self-interrogation of their own subjectivity and context.

We examine the limitations of LLMs with respect to reflexivity, supporting the view that they are incapable of genuine reflexive analysis. We argue that LLMs, as producers of unfaithful self-explanations and “bullshit” (indifference to truth, as per Hicks et al.), cannot be shown to engage in the necessary internal work of meaning-making or establishing researcher-subject relationships.
Rather than entirely rejecting the use of LLMs in reflexive qualitative data analysis (QDA), we find potential to reframe their role through the lenses of distributed cognition and dialogic spaces. The use of LLMs as automated Socrates or devil’s advocates can enable researchers to harness the friction generated by the model’s otherness to challenge their own assumptions and enhance their reflexive practices.

More information

Identifiers

ORCID for Bethany Ahluwalia: orcid.org/0009-0000-2797-2983

ORCID for Adriana Wilde: orcid.org/0000-0002-1684-1539

ORCID for Richard Gomer: orcid.org/0000-0001-8866-3738

ORCID for Daniel Ashton: orcid.org/0000-0002-3120-1783

Catalogue record

Export record