Natural language processing (NLP), a subdiscipline of artificial intelligence, can be used to analyse organisational discourse. It provides a quantitative method for testing and refining theories of discriminatory language and its impacts in the workplace. Practically, it offers HR professionals and organisational leaders a tool for auditing their communications for unintended biases.This study demonstrates this methodology by investigating age-related biases in interview data from workshop discussions with employers and trade union representatives on workplace active ageing. We employed two NLP techniques: (1) Word Embedding Association Tests (WEAT) to measure relative bias between young and old terms, and (2) direct association analyses to examine how strongly each age group associates with negative descriptors. WEAT analysis demonstrated relative bias favouring younger workers regarding competence in both groups (union: s=0.581, p=0.033; employer: s=0.785, p<0.001), though no significant differences were found between unions and employers in their WEAT scores (t = -0.4031, p = 0.6879). Direct association analyses further showed the employer model associated old terms more strongly with negative physical descriptors than young terms (p = 0.049), but no significant differences were found between unions and employers in their direct negative associations with old terms (all p > 0.05). These findings suggest that despite some internal biases within each model, there are no significant differences between employer and union discourse patterns, challenging assumptions about fundamentally different stances between stakeholder groups. This methodology can be applied to measure bias in any written qualitative data.
