Abstract
Cohen's kappa remains one of the most widely used metrics for assessing inter-rater reliability (IRR) in qualitative research. Yet, applying this statistic to text-based data coded at the micro-level of pragmatic function presents methodological tensions, particularly when conducted within computer-assisted qualitative data analysis software (QDAS) such as NVivo. This paper examines the epistemic and practical challenges encountered in establishing IRR during the coding of peer interaction data and extends the discussion beyond procedural concerns to consider how digital tools shape methodological decision-making and the appearance of objectivity in qualitative inquiry. By situating NVivo within broader debates on transparency and digital mediation in social research, we highlight the interpretive processes that underpin statistical representations of reliability. The paper concludes with a series of recommendations and reflexive strategies designed to enhance the transparency, rigour, and methodological accountability of IRR practices in QDAS-supported qualitative research.