I watched the excellent Stanford Conference on Human Centred Artificial Intelligence (AI) this week (25 March 2021). It made me more committed to the idea of inclusion than ever. The conference addresses three areas - AI in healthcare, art, and education. It is over six hours to watch but well worth dipping into the agenda for anyone who is caring, creating, or teaching.
http://hai.stanford.edu/streaming-2021-spring-conference
The main theme of the conference is that we are living in an age of artificial intelligence (AI) implementation, where algorithms, targeted machine learning, and deep learning will reap the benefits of:
- faster data processing power
- more accurate probabilities
- better predictive analytics
- more depth to understanding of what can be observed and measured
In this golden age there are many benefits of AI for human health, wealth, and social justice. These benefits can have a global impact.
There are also some down sides to AI, and many open questions. Such as:
- costs/resources needed to develop AI can detract resources from immediate social problems
- issues about data quality and data protection
- harms associated with use such as surveillance, less direct human interaction
- adverse impact on minority groups
- the problem of the ‘anomalous individual’, who experiences a discrepancy between AI’s intelligence and their personal experience or circumstances, because they are an exception to the model, data or probabilities on which AI is based
In this new and exciting era, it is easy to get distracted by aspirations for the AI rather than the value of AI to the individual human and to humanity. Unfortunately, AI can be a more efficient path to something we do not want – such as scaling up ineffective pedagogies or superficial categorisations of identity.
This, failing to hit the mark, is the reason why I believe design justice – paying attention to human needs, preferences, and circumstances - must to be up front in the creation of AI.
Design justice means the involvement and inclusion of people right from the beginning of the design process. It can ensure AI makes sense in a world where it is difficult to know:
(1) what is real
(2) what we can trust
(3) how to respond or behave
(4) and most importantly how to alleviate suffering through human unity
Machines that look like or behave like humans raise issues of comfort levels with augmentation, replication, and replacement, transparency of production and attribution of contribution.
The verge of implementation of useful AI technologies is driving new forms of governance, regulation, new types of job roles/upskilling, learning policies, interaction design, data flows/feedback loops/data trails, integrated security and privacy techniques, and surveillance and monitoring agreements. AI is requiring rapid societal change. Governments (G20) are quickly working on international treaties to ensure humans do not give AI the decision-making power to take human life.
On a more profound level, AI also raises questions about the reconstruction of ‘being’ - for example, what being right, being sensitive, being compassionate mean - from the perspective of the changing machine/human relationship.
A reflection is becoming visible - between how technology development positions people and AI. For example, the possibilities for exclusion, instrumental use, control, collaboration, and ‘agent’-led, can be applied to both the public and to how society approaches AI.
It is naïve to think that the design of AI-driven technologies for COVID, or any other societal challenge, or frontier of technological development, is anyway democratic. But it could be more diverse and inclusive.
Few people outside of research and technology arenas understand what AI is or could mean.
Few people have a direct say in choices about the design of technology.
Few people have a say in funding of AI research or evaluation of AI.
Few people in technology design understand the relevance of human sciences or see the moral, ethical, and technical benefits of inclusion.
Design is driven by the aspirations of technologists and the knowledge of practitioners, rather than centred on the human and what is important to them. Enacting design justice is of course challenging in the context of intellectual property, data ownership, patents, and commercial interests. And, even if technologists aim for a more collaborative co-design process, there are pervasive cultural assumptions and unconscious scripts for design that are difficult to unmask or stand outside of.
Bringing diversity of world views into human centred designs could embed human values into the technologies we live by.
Principally my argument is that anyone working in AI can seek to be more inclusive in deciding the possibilities of what we create. If people chose to, they could follow four principles of design justice in AI. (These principles are derived from an unpublished analysis of AI development for mental health and wellbeing).
Four principles for design justice in AI
1. Meaningful and authentic public engagement in all areas of AI technologies can be supported and guided by the core principle that such technologies should sustain, heal, connect, and empower people and communities
2. Substantial ethical concerns, such as inequalities, cultural and population biases, safety, acceptability, and broader socio-political issues, could be better understood and moderated if design focused on the concerns of the community over the intentions of the designer
3. AI technologies need to emerge from an accountable, accessible, and collaborative process that describes how patients and the public have been involved
4. AI-developers and their designs can be informed by, and contribute to, shared knowledge in design and tools to support design justice
If you are interested in these issues, please watch the Stanford Human Centred AI Spring Conference at the weblink address above - or get in touch with me for a chat about how we make design justice real.