tl;dr: Artificial Intelligence research needs to focus on enriching people’s relation to each other, empower them to organise their life beyond the market, and assist in the transformation towards sustainability
Technology is never neutral. It transforms the world, our perceptions of it, and our relations towards one another in a fundamental way. Artificial Intelligence (AI) as technology, as a technological paradigm, as specific applications of machine (self-)learning as well as automation of highly cognitive decision routines, raises questions if we should use it, how we should use it, and what happens then.
There are several phenomena we can subsume under the headline of AI that revolve around the capabilities of machines to carry out tasks requiring some form of human intelligence and cognition. The lowest common denominator for AI might be the ability of a machine to mimic human abilities and/or exceed them in their execution.
Questions about technological applications are often addressed in an economic reference frame, if and how they enhance economic value added. Likewise the possibilities of ecological value added, especially when dealing with intelligent energy systems, organising sustainable mobility and making both our homes and our cities “smart”. To give a more holistic view on AI and the ethical conditions of its use, I will focus on a different approach: the concept of conviviality by Ivan Illich and if AI can be, at its core, a convivial technology.
Conviviality according to Illich is an ethical principle for technological societies: the conditions of individual autonomy realised in multiple interdependencies with each other, with human institutions, and our non-human natural environment. When Illich was developing the concept of conviviality in the 1970s, he was looking at the highly industrialised western societies and the heteronomous relations individuals had with anonymous market forces. As consumers in such a relation, you can only consume what the market delivers, what companies decide. Technology in this regard should become a means for liberation from heteronomy, especially from large-scale technological infrastructures and with a minimum of natural resource use. Industrial production for Illich had exactly one purpose: to provide tools for self-production, to enable passive consumers to become active producers for self-use (or in Toffler’s terms “prosumers“). Institutionalised forms of these acts of self-production for self-use can be found in the proliferation of repair cafés, urban gardening, community supported agriculture, and local exchange trading systems.
The hallmark of a convivial technology would then be its contribution to minimise heteronomous relations, to reduce over-reliance on both market as well as community relations, and thus maximise autonomy in interdependencies.
German degrowth researcher Andrea Vetter has developed a conviviality matrix based on Illich’s ideas for evaluating technologies. Vetter distinguishes dimensions and levels of conviviality. Dimensions are necessary materials, the physical conditions without which no technological application is possible; production of goods and services through technology; use of goods and services and their socio-economic use contexts; the necessary infrastructure to actually use these goods and services like e.g. the energy infrastructure. Levels of conviviality focus on the influence of technology on human relations (relatedness); on the access to technology, if it is open and socially inclusive or proprietary and expert-based; on the adaptability of technology towards local and individual contexts (cf. frugal engineering); on the bio-interaction of technology with its natural environment; as well as on the appropriateness of social-ecological benefits of technology in relation to its social-ecological damage.
Applied to AI it becomes clear that the current discourse some dimensions like materiality and physically necessary infrastructure do not play a large role. Likewise the level of appropriateness. When focusing on these dimensions and that level, the conviviality matrix gives us clear suggestions about future requirements for AI development and AI applications:
- Appropriateness of materiality: are the backbone resources of AI applications like precious metals and rare-earth metals locally available? Most likely not, so are these materials easy re-usable or recyclable, in short: is a local or regional circular economy possible? Currently this question has to be answered with no. Great efforts are necessary to change that, but at the same time these are interesting business opportunities as well.
- Appropriateness of infrastructures: can AI applications be used with a minimum of material and energetic supporting infrastructure? Ideally, can these applications continue to work if that infrastructure has a major outage? As software, AI has the potential to work indefinitely as long as the hardware works – and enough energy is available. Hardware elements of AI applications need to be durable and easy to maintain, probably something that is not the case today. Ideally, an AI application should be able to run on just one device, e.g. a smartphone with long life span, without the need to connect to the internet very often and with minimum energy requirements. This opens up new perspectives on future development of AI software and hardware.
The conviviality matrix enables AI research to focus on the question of increasing and enriching the relations between people. How strongly do AI applications support autonomy in interdependencies? Can local communities self-organise economic activities beyond the market and create a civil sphere of economic exchange? Just one example: how can AI enable a local community, a neighbourhood, a city quarter to organise their needs for mobility beyond commercial car- and ridesharing services as well as public transportation? All the dimensions and levels of the conviviality matrix can be critically examined with AI applications in order to determine novel AI research pathways beyond narrow-minded economic value added increases. What is also clear from the discussion here: currently, AI is not a convivial technology.
To make AI a truly convivial technology, to enable and empower people to fully realise their autonomy in interdependencies with other, with human institutions, and our non-human environment, some guiding principles can be derived from the conviviality matrix:
- Joint focus and development of AI software (applications), AI hardware (devices) and AI infrastructures (raw materials, energy).
- Durability, repairability, recyclability and low material and energy intensity.
- Social inclusion, adaptability to local contexts and individual needs.
- AI applications with focus on enabling relationships between people and empowering them to organise themselves beyond the market in a social and ecological friendly way.
If these guiding principles find their way into the mainstream discourse about AI it is safe to assume that this new fundamental technology can not just be another growth stimulus for an ailing global economy but deliver a significant contribution to a social and ecological transformation towards a more convivial and sustainable society.