Making AI Intelligible via Design

British poet, William Blake famously condemned the Albion Flour Mills in London, labelling them ‘Dark Satanic Mills’ when their automated processes threatened the bread manufacturing processes that were dominant prior to The Industrial Revolution. The disruption of new technologies generates anxiety in the minds of those whose lives are affected by them; seeming conceptually difficult, counter-intuitive and alienating.

Today, AI has created an air of suspicion in popular culture via apocalyptic portrayals in science fiction. All too often, these issues revolve around AI not being properly understood, even by researchers. We haven’t developed adequate conceptual models with which to make sense of its workings, and partly because it’s simply not, and may never be, fully understandable.

Complexity Theory is a framework for providing radical new ways of understanding the physical, biological, ecological and social universe. It’s based on the idea that everything is made up of Complex Adaptive Systems — from intergalactic systems and terrestrial weather patterns, to human central nervous systems; from cosmic to nano. AI can be considered one such system as it exhibits the following properties:

Emergence
— AI does not follow grand narratives; instead the nodes in the system interact in seemingly random ways. From these interactions, patterns emerge, informing behaviour of the other nodes; and as a result, of the system itself. The unimaginable designs created by Generative AI showcase these processes of emergence.

Co-Evolution
— Machine Learning capabilities mean that AI systems are able to adapt based on variety between nodes, and prior actions can devise better predictions for future patterns; hence the term predictive analytics. Soon, however the evolution of AI will be able to not only describe and predict, but also to direct or prescribe the best paths for future operation.

Connectivity
— AI amasses data from an expanding array of sensors, monitoring aspects of the physical world like air quality, traffic flow, ocean wave heights; as well as our own electronic footprints such as ticket sales, online searches, blog posts, and credit card transactions, these systems can glean patterns and grasp insight inaccessible to the human mind.

Self Organisation
— AI typically operates in distributed decentralised networks, without top-down control or hierarchy, constantly self-organising through emergence and feedback. ‘High Frequency Trading on stock markets mean that statistical and machine learning techniques are currently the best tools available to pan for gold’ argues Jerry Kaplan, American computer scientist, author, futurist, and serial entrepreneur.

Edge of Chaos
— Elon Musk, among other things, is the creator of Open AI and more recently Neuralink — a venture to merge the human brain with AI. Yet despite his involvement in the AI revolution, Musk warns of AI as ‘Our greatest existential threat’. These opinions may seem paradoxical, conflicted between optimism and peril. Yet, this is the very nature of AI: balanced between order and anarchy, at the edge of chaos.

Nested Systems
— Much like the Internet or any organism, AI consists as a network of networks; a system of systems — where agents at one level are the building blocks for agents at the next level.

Even though complex systems like Machine Learning algorithms come across as unknowable black boxes, this shouldn’t stop us from working together with them. ‘Complexity is both necessary and manageable’ says Don Norman, director of The Design Lab at the University of California.

In a world riddled with AI complexities, humans are essential for interpreting and humanising these mixed signals; and no one is more fit for the job than designers. ‘Designers are masters of complexity’ argues Katerina Alexiou, Senior Design Lecturer at The Open University ; managing a multitude of variable external factors such as constraints, expectations, satisficing requirements, regulations, hypothetical possibilities and uncertainties, to create artefacts and systems that exhibit ‘extensive domain-specific expert knowledge’. In this light, confronting complexity is a key methodology of the design process.

Once dissociated from the dumbed-down, conservative conceptions of such socio-technical phenomena, the designer’s role is to accelerate towards the novel and valuable world afforded by computer creativity. The task is not to simply grasp complex concepts, but to effectively communicate them onwards in an intelligible and engaging way, giving them cultural meaning. The key to meaning-making is to decode or understand the complexities and recode them into conceptual models with which to transform the complex reality into ‘workable, understandable mental concepts’, says Norman. Conceptual models should not take away from the features or capabilities of the complex system, but should simplify the complexity in order to increase understanding, usability and functionality. He concludes, ‘Simplicity is not the opposite of complexity — complexity is a fact of the world, whereas simplicity is in the mind’.