AI And Ageing: Towards Gerontechnology
Bill JohnstonIn our book ‘All of Our Futures’ (2021) Craig Dalzell and I identified four change forces impacting society and democracy in the 21st century:
• Climate change.• An ageing population.• Technological change.• Economic change.
We suggested that: “The challenge is to link demographic ageing to the other three macro level global dynamics which are currently treated as separate issues in policy formation.” (P4). That challenge remains and this article addresses the ageing population and technological change linkage. The specific technology is AI (Artificial Intelligence), broadly described as the simulation of human intelligence by machines, particularly through research on the technology of machine learning.In the current hyperactive political narrative AI has morphed from a specialist field in computer science to the holy grail of economic policy. For example the UK government’s “AI Opportunities Plan” of 13 January 2025. This was presented to Parliament with the rousing sub-title: ‘Ramping up AI adoption across the UK to boost economic growth, provide jobs for the future and improve people's everyday lives”. Even the slowest MP couldn’t miss the point!The aspect of demographic ageing in this article is ageism because it remains a powerful yet underappreciated threat to older people’s rights, distorting intergenerational relationships and undermining social cohesion. My sense of the current issue is that whilst AI is presented as an innovative and beneficial information technology it is not clear that its adoption will necessarily result in direct benefit for our ageing population and other groups in society. Because AIs are trained by mashing up and repeating existing works, they are prone to repeating and amplifying systemic biases within the material they are trained on. Racial bias is an obvious category in need of action and as we shall see, so is ageism.To that extent, the present headlong rush for corporations to profit from AI, and politicians to harness AI to their ambitions, is likely to maintain inequalities rather than removing them. This problem is pressing due to the anticipated upsurge in adoption of AI technology across all facets of society and requires a response based in concern for equalities.
AI and Ageism: The example of health care
In our book we described ageism as negative stereotyping, bias, and discrimination on grounds of age. The consequence being that older people’s rights and concerns are much less likely to feature in mainstream politics and media representation. Ageism is pervasive but overlooked and does not get the attention paid to categories of race, gender, sexuality etc despite age being a protected characteristic in the 2010 Equalities Act.AI ageism can be approached from the starting hypothesis that the biases in the data sets used to train AI systems will mirror and amplify existing ageist biases unless steps are taken to prevent this. Consequently, those biases and stereotypes will be preserved, tending to undermine the proposed benefits of AI (1.2.3.). How might this impact health care?The World Health Organisation (WHO) briefing on Ageism in Artificial Intelligence for Health (4) positions AI within the concept of ‘gerontechnology’ described as the “technological software and devices that meet the needs of older people”. From that holistic perspective the briefing illuminates “… the potential interplay between ageism and AI for health as it affects older people, including the conditions in which AI for health can exacerbate forms of ageism…”. Several key aspects of AI are identified: (i) remote monitoring of care services; (ii) drug development processes; (iii) risk of encoding ageism in data; (iv) reinforcing digital divides in access to AI technologies; (v) exclusion of older people from the design, development, and deployment of technology.The briefing concludes by listing ways to maximise the benefits of AI for older people by avoiding ageism:
• Involving older people in the design of AI technology, including the provision of training and support, and creating cross generational teams in the AI workforce.• Building age diverse data science teams to select and validate data. Including age as a criterion for selecting data to ensure that data sets are representative This would apply especially to government sponsored services.• Investment not only in infrastructure but in developing older people’s digital literacy. This to include carers and care providers.• Right to consent and contest AI supported decision making.• Age sensitive regulation and governance structures designed to prevent ageist practices.• Increased research in how to design and operate AI systems to avoid ageism in a fast-moving technology. This should include studies to identify interactions between ageism and other biases such as racism and sexism.• Robust ethics processes applying across all organisations involved in development and implementation of AI systems to ensure that all ethical challenges are at the forefront of design work and quality assurance.
If adopted these precepts would go some way to ensuring that ageism is taken seriously in the development of AI to the benefit of health and care systems aimed at older people.By combining technological change and population ageing using the interdisciplinary construct gerontechnology the WHO has challenged the neoliberal assumption that tech is the fiefdom of billionaires and tough guy politicians. This is a strategic position to build on in the coming years.The work of the WHO notwithstanding it is still an open question as to how seriously the UK and other governments take ageism in their desire to exploit AI for economic growth. The UK 50-point plan referenced above is not encouraging (the list of sources does not include WHO). The three main pillars of the UK plan are: investing in infrastructure; rapid roll-out in both public and private sectors; positioning the UK as a world leader in AI. Whilst there is reference to issues of regulation and governance the emphasis is on ‘turbocharging’ AI for national economic advantage.Once you get down into the list of detailed points covering research, data collection, training scientists and so forth there should be opportunities to counterbalance the neoliberal drive of the plan by insisting on greater attention to the public good. At those points the WHO proposals offer a valuable foundation to consider AI and population ageing in the interests of older people and the wider society.
References
1. Nielsen, A. and Woemmel, A. Invisible Inequities: Confronting Age-Based Discrimination in Machine Learning Research and Applications. Workshop on Generative AI and Law, co- located with the International Conference on Machine Learning, Vienna, Austria. 2024.2. Chu, C.H., Donato-Woodger, S., Khan, S.S. et al. Age-related bias and artificial intelligence: a scoping review. Humanit Soc Sci Commun 10, 510 (2023). https://doi.org/10.1057/s41599-023-01999-y.,3. Stypinska, J. AI ageism: a critical roadmap for studying age discrimination and exclusion in digitalized societies. AI & Soc 38, 665–677 (2023). https://doi.org/10.1007/s00146-022-01553-54. Ageism in artificial intelligence for health: WHO policy brief. © World Health Organization 2022.END.