“Hermeneutic burden” and clinical responsibility: a response to Sparrow et al. on explanation and machine learning

This paper critically reviews Sparrow et al.'s notion of the "hermeneutic burden" placed upon clinicians by the demand for explainable artificial intelligence (XAI) in the context of adaptive machine learning (ML) systems. While Sparrow et al. highlight important additional labour that may be required of clinicians, this response argues that framing explanation primarily in terms of such a burden obscures its overall ethical significance. This paper therefore offers a supplementary account of the interpretive work associated with XAI in medicine that places it within existing models of the patient-clinician relationship. In particular, Emanuel and Emanuel's influential typology consisting of four models of the patient-physician relationship is used to extract possible justifications for the responsibility to grasp and explain not only patients' values and conditions but also ML outputs. This allows us to distinguish between 'hermeneutic burden' and 'hermeneutic responsibility' and emphasise that explanation in medicine is not an incidental task but part of a clinician's professional role, particularly on 'interpretive' and 'deliberative' models. The paper thus argues that viewing explanation as a hermeneutic responsibility linked to patient autonomy clarifies the ethical significance of XAI in terms of both the grounds and scope of clinicians' responsibilities. At its core, the ethical challenge raised by XAI in clinical practice concerns not only the burdens it may impose on clinicians but also the evolution of clinicians' traditional interpretive duties in the novel context of ML-mediated care.