Visual mapping in AI interfaces: to establish trust in explainable AI in mammography

Sammanfattning: Artificial intelligence (AI) is a broad topic that many researchers and stakeholders want to implement into their daily tasks. There are expectations to implement AI in hospitals, starting with mammography centers. But there is a need to guarantee that it will be efficient and safe to use, and that requires a lot of research. However, the challenge is about how radiologists should understand how AI will communicate when it shows unclear results during mammography examinations. When it comes to cultural communication and expectations, radiologists are working with visual communications, since that is where they see the x-ray images of breasts. However, there seems not to be any clear directions on how the AI is supposed to communicate back, in a visual and explainable way. The current studies do not focus on the practical usage of AI towards its target group and how they take in information and how they take in information that is consistent with their work habits and the use of visual communication, as well as human interaction. The goal of this thesis is to view this field from an information design perspective and contribute with a different view. This master thesis will explore the possibility of how visual mapping can contribute for radiologists, in mammography, that will eventually work alongside AI. To highlight important factors that are important when it comes to working with AI. As well as inviting AI researchers and experts to explore the expectations of AI in mammography. The theoretical framework contains explainable AI (XAI), scientific visualization, visual perception, human computer interaction (HCI) and visual mapping etc. This research will use design thinking processes, to empathize with the radiologist and conduct interviews, artifact analysis and finalize a prototype. The results of the thesis will highlight important keywords from the participants that are emphasized on the prototype.