“…Importantly for this article, this excerpt from the interview highlights how the objective of AI development is to build models that are accurate enough and highlights how accuracy is negotiated (Mackenzie, 1990) which for Laurent and Thoreau (2019: 165) (Carusi, 2016): labels/codes become essential criteria and underpin judgements about the accuracy of validation tests (Scheek et al, 2021). Importantly, for this article, internal validation tests provide further opportunities to mediate practice-orientated trust between collaborators (Clarke et al, 2006a(Clarke et al, , 2006bOudshoorn, 2008;Kuutti and Bannon, 2014;Papangelis et al, 2019). The next section will illustrate how this trust building deepens, paying particular attention to how clinical experts generate meaning with respect to the labels/codes or variables in the model -a process which is particularly useful when it comes to the 'interpretability' of outputs and continues the bridging between internal and external validation.…”