Democracy is facing serious challenges. Some are contemplating whether artificial intelligence (AI) could help revitalise it. AI is, after all, heralded as the solution to a broad range of social and political challenges. Henrik Skaug Sætra argues that AI does indeed offer some hope, but also serious potential pitfalls
The relationship between AI and democracy is complex, and the search for technological solutions to democratic challenges might be futile. While some promote the potential of AI to support democracy, a recurring criticism has been that AI systems exacerbate challenges related to technocratic tendencies. These include opaqueness in decision-making processes, bias and discrimination, increased polarisation, the disruption of political processes by foreign actors, and a shift of political power to private corporations.
We also risk diluting and undermining the core foundations of democracy as we use AI to mediate, and optimise, political ideation and communication.
Might AI strengthen and promote democracy, despite these pitfalls? Of course, we want to govern AI democratically, but we also want to secure the future of the democracy that we desire. How should we go about it?
Proponents of AI sometimes claim it is a democratising technology – or a technology that can be democratised. However, those people often base such claims on a superficial understanding of democracy, or at least use the term in a way not related to democratic theory, as a recent article argues.
The information and people to whom we are exposed influence how we perceive our societies and the other people living in them
First, we must distinguish between digital technology in general, and AI. Not all means of communicating digitally, for example, include AI. But some do. One example is social media, in which AI plays a fundamental part in determining how content is disseminated, and who sees what. All these aspects relate to democracy, because the information and people to whom we are exposed influence how we perceive our societies and the other people living in them.
One way in which AI could play a role is through promoting dialogue, deliberation, and exposure to others and their ideas. These are all crucial requirements for a well-functioning democracy, as various proponents of different types of democracy emphasise.
Beyond connecting people and ideas, we have recently seen rapid progress in the development of large language models (LLMs), including the immensely popular ChatGPT and alternatives. Researchers in this field are exploring AI's potential to penetrate far more deeply and directly into democratic processes. For example, we could use AI to find the 'most popular' policies. We might also tune and develop AI models with the aim of fostering consensus.
Finally, AI systems provide superhuman analytical capabilities and performance in an increasing range of highly complex contexts. These systems have proved themselves in many different games and scientific settings – Google DeepMind’s AlphaGo, AlphaZero, AlphaStar, AlphaFold, etc. are prime examples – and could clearly apply to some of the key challenges of today’s political context. One example is how we could use AI to identify and elucidate environmental policies aimed at mitigating climate change. AI might, for example, help us better understand the dynamics of natural systems, and identify the most effective actions.
The promise of AI-supported democracy, however, also risks potential pitfalls. These relate in part to how technology-mediated democracy would be different, and potentially worse, than democratic activity through conventional means. Many also worry that AI – and other technical systems – are imbued with a certain politics of their own. It could be that AI is more likely to give rise to a form of technocracy instead of a revitalised democracy.
The first set of pitfalls are tightly linked to the benefits of AI, and in particular how social media and the use of AI in such contexts have effects essentially detrimental to democracy. Filter bubbles and echo chambers are well-known concerns. Yet evidence does not support fears that these mechanisms might result in negative effects related to, for example, increased polarisation.
Another set of potential pitfalls relates to technocracy. Some believe that technology is political. It thus follows that AI provides the foundations for a revival of, for example, the 'political cybernetics' of David Easton and Karl Deutsch.
The more data we have – and the better means of analysing them and designing political systems that measure and respond to people’s needs and preferences – the more redundant political activity becomes, right? If democracy is only instrumentally valuable because it provides the most effective way to distribute goods and make collective decisions, it will certainly face challenges as AI develops.
AI, some argue, is better than humans at making decisions. This, of course, could include political decisions
AI, some argue, is better than humans at making decisions. This, of course, could include political decisions. Wouldn’t it make sense, then, to use AI's potential to improve political decision-making? One way to do this is through the outright outsourcing of certain decisions to AI systems. Alternatively, we might use AI to augment human capabilities in hybrid intelligence systems.
A central objection to AI-supported democracy of the kind that helps us mediate and improve deliberation rests on a key insight from deliberative democratic theory. Deliberation is not simply about identifying the optimal consensus or outcomes. It is about the process that occurs when citizens meet and get puzzled by each other’s quirks and preferences. When this happens, citizens become curious about others, and try to understand them. If we want AI to support democracy, it should facilitate deliberation, not replace it.
If we want AI to support democracy, it should facilitate deliberation, not replace it
What we need now is for our governments to acknowledge that AI and digital technologies are shaping political systems. Governments must ensure their citizens have access to the most comprehensive information on AI, and join them in finding an answer to two key questions: What are our foundational values and goals? And how is AI contributing to or undermining our efforts to promote such values and reach these goals?
