Design of an Informed Consent Support System for Dental Implant Treatment Using Generative AI

Abstract

Implant treatment involves decision-making processes spanning multiple domains, including imaging diagnosis, surgical planning, prosthetic design, risk assessment, and estimation of cost and treatment duration. Patients, meanwhile, seek clarification regarding treatment options, intraoperative and postoperative complication risks, cost validity, and considerations aligned with their personal circumstances and values. However, meaningful dialogue is often constrained by limited consultation time and privacy concerns. This paper proposes a design concept for a generative AI-based informed consent support system for implant treatment, explicitly assuming that the system does not replace medical diagnosis or treatment decisions. The architecture employs Retrieval-Augmented Generation (RAG), which generates responses by referring to predefined prior knowledge, and is governed by an AI agent with clearly defined roles and constraints. The design aims to integrate evidence grounding, explicit indication of uncertainty, and response suppression in situations requiring professional medical judgment, thereby prioritizing safety. Furthermore, the knowledge base is structured into three layers consisting of public evidence, abstracted clinical knowledge, and institutional or regional information, under an operational framework compliant with relevant legal requirements. This paper does not present empirical validation；rather, it organizes key design considerations and requirements for applying generative AI to informed consent support in the medical domain.