Human-Centered Design to Build Symbiotic AI Systems: A Case Study

The spread of Artificial Intelligence (AI) can bring many advantages improving the way in which humans carry out activities. In this scenario, humans and AI collaborate towards shared objec- tives enhancing both parties’ capabilities, reaching a symbiotic relation- ship. Human-AI symbiosis can be reached through the Human-Centered AQ1 Design (HCD) approach to ensure that users’ needs, characteristics, and preferences are considered when building such systems. This contribution presents SAI4HCD, an approach that integrates the indications to create Symbiotic AI (SAI) systems (i.e., guidelines to design and Success Crite- ria to evaluate) in HCD. In order to validate the approach, SAI4HCD was applied to a case study in the medical domain by creating BrainMed, an AI-based systems to support neurologists in the detection and diagnosis of Alzheimer’s Disease. The contribution reports the process of design- ing and evaluating an high-fidelity prototype of the system, focusing on its interaction paradigm and a Large-Language-Model-based component to provide explanations in different forms. These aspects reflect the pil- lars of SAI, guiding humans towards an aware decision-making process, especially in such high-risk scenarios, where wrong decisions could lead to irreversible consequences to other individuals.