Methodology for a proactive and collaborative development and implementation of wholesome and ethical Artificial Intelligence in Healthcare in Australia

Artificial Intelligence (AI) is pervading all sectors of industry including healthcare. The COVID-19 crisis accelerated the penetration of AI into healthcare with the development of applications ranging from case identification to population monitoring. While the situation called for the rapid deployment of these COVID-19 AI apps, ethical considerations should not be foregone or become an after-the-fact remedial exercise. The high-tech industry developing AI-based Healthcare Applications (AIHAs) does not share the same culture of ethics as the medical field, and regulations around AI are still in their infancy. Hence, it is important to understand how ethics implementation in AIHAs can be done. The aim of this thesis was to explore how to implement ethics in an AIHA. My scoping review on the topic found that implementing ethics in AIHAs is a complex issue requiring stakeholders’ involvement. Therefore, a systems approach was adopted for the exploration.

The research used an exploratory, two-stage qualitative design involving focus groups, and semi-structured and in-depth interviews. Critical Systems Thinking principles guided the design, and facilitation of the study. In the first stage, a transparent and inclusive participatory process engaging a diverse group of clinicians, patients, and AI developers was set up to capture the different worldviews about a fictitious COVID-19 app scenario. The chosen app scenario was set in the Australian context and based on an aggregation of real life COVID-19 apps that were developed in 2020. One finding was that ethical issues could be illuminated through mapping the flow of knowledge in the patient-clinician-AIHA system. Consequently, in the second stage, the flow of knowledge between the different agents in the system was mapped. Data analysis followed principles of a reflexive methodology where data are examined from different perspectives and includes a reflective piece from the researcher.

A methodology was developed to conduct the participatory process for implementing ethics in medical AI, mapping the flow of knowledge between the agents of an AIHA. The methodology facilitates inclusive, respectful dialogues allowing candour in exchanges that are foundational to the implementation of ethics in an AIHA, while maintaining transparency of the process. Mapping the flow of knowledge between the elements of an AIHA allows for surfacing assumptions and ethical issues in an effective way. Additionally, key considerations to examine at the inception of an AIHA were identified. These are (1) the readiness of the different agents in the system to exchange flow of knowledge, (2) the consequences on human-to-human relationships when introducing the AIHA, (3) the transparency of the values embedded in the app, (4) the alignment of purpose of the different agents of the system with the system’s purpose, and (5) the custodianship of the data, algorithm, and stakeholders’ engagement. This program of research has significant implications for policy and practice including provision of a methodology for implementation of ethics in AIHAs and identification of key considerations for reporting on ethics implementation.