SCHB Annual Conference June 2024: The Ethics of AI-Person Interactions

14.10 – 15.00 hrs: People At Its Heart: Trustworthy, Ethical & Inclusive AI: Mr. Calum McDonald (Scottish AI Alliance)

15:00 – 15.50 hrs: Patient and staff experiences with the Transplant Benefit Score: Lessons for the use of AI in high-stakes medical decision making : Mr. Jamie Webb (University of Edinburgh’s Centre for Technomoral Futures)

We were very happy to welcome Calum McDonald and Mr James Webb to the Golden Jubilee, not really knowing what to expect.

Calum focussed on his work with the AI Alliance and their outreach programme, which seeks to demystify the hype and fear around AI. They seek to deliver on Scotland's vision to be a leader in trustworthy, ethical and inclusive AI. They seek to empower Scotland's People, Supporting Scotland's businesses and organisations and Influencing policies that impact Scotland. He brought out the good uses and limitations of this new science and put people at the heart of AI, by frequent educational events and the use of the ladder of engagement. He discussed some of the dangers such as deep fake videos and also the environmental impact of building and powering huge computer systems (including the use of rare metals). He also discussed the need for international regulation and agreed that this is very hard to implement. On job displacement, though a concern, he emphasised how AI can free up workers to do other important tasks.

Jamie spoke of the use and limitations of AI in medicine, with human input of protocols and data. The concept of the use of historical data input, leading to predictive output, without human modelling, introduced the idea of machine learning. The large amounts of data in the black box could be analysed with greater than human capacity, using a random forest of branching decision making or a deep learning matrix. He then went on to describe his PhD study of the Transplant Benefit Score (TBS), which is one of the more commonly used algorithms for the Liver transplant waiting list allocation for Donors by Brain Death (DBD), which collated 29 characteristics to score needs and match donors and recipients. However, the young, due to their better prognosis were often left longer on the list. The teams who dealt with the patients directly, experienced the human dimension of suffering, and were left with a sense of unfairness. The systems can be tweaked, but often not explained to patients as the interface was lacking this aspect in its output. The question of trust in the system arose.

After an interesting discussion, we concluded that the human factor is needed. We can rely on computers, but we maybe cannot yet trust them, due to the human input defects and lack of a good human interface output. These could be improved.