Strategic Value Design in Medtech AI : Mediating the Conflict Between Institutional and Epistemic Logics

Sammanfattning

Artificial Intelligence is reshaping diagnostic medicine, yet most clinically validated tools fail to reach deployment. This thesis investigates why, focusing on a structural conflict that has been treated separately rather than together in the existing literature: the trust requirements of clinicians, who evaluate diagnostic AI through an epistemic logic of professional judgment, and the trust requirements of clinic managers, who evaluate the same tool through an institutional logic of operational and financial accountability. No published framework treats a single device interface as a sociotechnical boundary object that addresses both logics simultaneously. The thesis explores whether a dual-faceted interface can fill this gap, using a salivary diagnostic platform for early periodontitis detection as the concrete context. The interface allocates institutional content to a Management View and epistemic content to a Clinical View, both grounded in a shared data structure. The investigation follows a Research through Design posture. Value Sensitive Design guides the design work, Boundary Object Theory provides the analytical lens, and Business Intelligence and Explainable Artificial Intelligence supply the technical vocabulary. A high-fidelity prototype and a physical device model were evaluated through nine semi-structured interviews with dental practitioners and clinic managers in Sweden and Poland, analysed using Reflexive Thematic Analysis. The findings refine both Design Propositions. Trust signal sets diverge across gatekeeper categories as predicted, with substantial within-category variation conditioned on clinic type and clinical sub-specialty. The dual-faceted strategy functions as a boundary object under three conditions: referent separation, configurable role-conditional access, and integration with practice workflow infrastructure. Three framework extensions emerged: the patient as a third audience, workflow integration as a constitutive condition, and platform scope as a design variable. The contribution consolidates into thirteen design guidelines for Strategic Value Visualization in MedTech AI.