In this week's blog, we explore the research of the THEMIS 5.0 project on the role of Artificial Intelligence (AI) in the healthcare sector, looking particularly at the insights from the workshop's first session. As discussed last week, the first session divided findings into four key themes that we shall examine in detail, which were:
Responsibility
Transparency and Accuracy
Cost vs. Efficiency
Attitudes Towards AI
The THEMIS 5.0 workshops were in agreement that it is certain that AI will play a significant role in the future of the healthcare sector, becoming a regular feature of hospitals and work spaces. Additionally, this is a source of great optimism, as AI could very well enhance the work of healthcare professionals by making processes more efficient and providing more objective and dynamic support to decision making. On the other hand, caution was expressed at the potential shortcomings of AI: mainly that it may be reliant upon biased data, be misused by workers due to a lack of proper education or training, and finally being too expensive to take up. With this in mind, the main recommendation for the use of AI was that it should act as an assistant to healthcare staff so that they can make well informed choices, but that the staff themselves should ultimately be entrusted with final decisions.
The research of THEMIS 5.0 is critical to identifying these concerns and guidance so that both AI developers and users can take them into account, and that the implementation of AI can be done with greater confidence.
Responsibility
Responsibility was noted to be especially important in the healthcare sector where human lives and livelihoods are impacted by workers' decisions on a daily basis. Not only is there a professional responsibility, but a powerful ethical obligation to patients to ensure that the correct choices are made regarding care. There was a strong impetus from the workshops that AI should be designed and implemented in the healthcare sector with the consideration of patients at the forefront.
Both workshops stressed that professionals must be entrusted with ultimate responsibility on decision making, especially with regard to patient care. The workshops expressed caution with AI having complete autonomy when considering patient care, but this does not prevent the tool being used to assist with diagnosis, treatment, planning and administration, where its impetus can still be supervised by individuals. In those areas, there is a tremendous deal of optimism for AI’s ability to analyse a large amount of data quickly, provide the latest knowledge and research regarding treatment which would aid immensely with early diagnosis and objective, evidence-based recommendations.
One of the most effective ways of ensuring that responsibility is maintained both by AI and the professionals using it, is for there to be as much information as possible to accompany the headline recommendations of AI tools. Workers would only be able to act as a safeguard against poor recommendations if they are empowered to make informed decisions with the help of AI, not merely regurgitating its recommendations. This must be reinforced by providing comprehensive AI training in education and workplaces, so that professionals may be able to use it correctly and properly integrate it into the healthcare sector.
Informing healthcare workers about both the reasoning and shortcomings of AI guidance would equip them to make informed decisions and utilise AI to its fullest extent without being encumbered by its limitations.
Transparency and Accuracy
As we have alluded to, the workshops recognised that it is vital for users to fully understand the data and reasoning of AI when employing it in the healthcare sector. Any lack of transparency will inevitably undermine trust in it. As well as the information driving AI being fully transparent and available to relevant parties, it is also critical that this information is as accurate as possible so as to prevent misuse. A major concern in the workshops was that the data employed by AI might not be fully representative of the population, creating inequitable as well as sub-optimal outcomes. Workshop participants advocated using as diverse a range as possible of patient populations to improve the quality of data, with the additional suggestion that allowing a freer exchange of information between AI datasets would help significantly in improving information accuracy.
When it came to how to mitigate errors in AI datasets, there was a difference in emphasis between the workshops. Danish participants focused on evidence and data led solutions, supporting rigorous testing, validation and constant updating of information. On the other hand Bulgarian participants, while certainly not opposing the suggestions of the Danish workshop, instead highlighted enhancing the adaptability and transparency of AI tools, so that healthcare professionals could critically assess the counsel that they were given. It is reasonable to argue that both these approaches would go a long way towards improving the transparency and accuracy of AI use in the healthcare sector.
Cost vs Efficiency
Balancing costs and efficiency was a concern for both workshops. However financial constraints were a larger concern amongst participants in the Bulgarians workshop, showing that there is legitimate concern that AI is as likely to be underutilised as much as it is over-utilised. Nevertheless, it was recognised that upfront investment in AI provided huge potential for efficiency gains, especially in diagnostics, planning and administration, creating savings in the long run and facilitating higher quality care.
In order to fully appreciate these efficiency gains, participants stressed aligning AI implementation with long term healthcare goals instead of AI developing parallel to the sector that it is meant to be employed in. This issue was of acute note in the Danish workshop, where participants argued that politicians and developers were not tailoring AI tools towards what professionals need the most. While the focus of AI so far has been on simple tasks like administration, there is hope that this new technology can be utilised towards complex endeavours like tackling rare diseases which can be often overlooked outside the healthcare sector.
Attitudes towards AI
Overall, participants took a nuanced approach to the potential of AI in healthcare, showing both optimism and scepticism about the tool’s tangible opportunities and limitations. They acknowledged its major role in the future of healthcare, and concerns were aimed at the technology either being misapplied or used without critical assessment. While the possibilities of AI in the future seem boundless, the participants were clear that its current capabilities are still underdeveloped and expressed reasonable caution about its implementation too soon, especially when dealing with complex needs in healthcare. In the long run, there was also the consideration that an over-reliance on AI at the expense of empowering professionals could result in an erosion of skills, making future workers less well equipped to carry out their tasks successfully.
While there was large unity of opinion, there were some differences of emphasis in attitudes towards AI amongst the participants. Younger professionals were more receptive towards the uptake of AI, while older participants were more likely to stress the importance of education to ensure that all workers were adequately prepared for using the technology. The Danish workshop focused on overcoming scepticism towards AI through rigorous testing and evidence in order to build trust in it. On the other hand, the Bulgarian workshop emphasised stronger dialogue and collaboration between experienced doctors who have worked without AI and younger professionals who are more likely to be familiar with it.
It is evident that the first session of the healthcare workshops alone provided a plethora of insights into the future of AI in the sector, in which a healthy combination of hope and caution was articulated. We look forward to building upon these findings in next week's blog, exploring the second session of the healthcare workshops.
Comments