AI and Democracy

IN EUROPE, AS in many parts of the world, elected politicians are emerging as some of the most despised and least trusted of public figures. It’s ironic given the inherent value that is often placed on democracy. From the recent election of far-right leader Geert Wilders in the Netherlands to leftist Spanish Prime Minister Pedro Sánchez’s power-grabbing machinations—which in their latest twist, have him in bed with rightwing Catalan separatists—for many, politics is increasingly a source of disillusionment rather than hope. And the modern politician is often seen as venal and narcissistic.

But if people in democratic systems get the leaders they deserve, then why are they voting for the very candidates they claim to dislike? One common explanation is the TINA factor. That “there is no alternative”. In Europe, indeed across the globe, even countries with bountiful populations do not seem to be throwing up the honest, decent, public-minded politicians that citizens ostensibly desire. It is not entirely clear why this is the case. Perhaps decent men and women don’t want to be involved in politics given its tarnished reputation.

A Pew Research Center survey conducted in 2020 found that roughly two-thirds of adults in France and the US, as well as about half in the UK, believed their political system needed major changes or reforms. Similar surveys abound.

Among the more startling ‘solutions’ to this malaise is the idea, both tantalising and terrifying, that politicians should simply be removed from the picture and replaced by super-intelligent, disinterested, and incorruptible artificial intelligence (AI) actors who don’t need to sleep.

A survey by Spain’s IE University’s Center for the Governance of Change found strong acceptance of proposals for automating some political decision-making. Fifty-one per cent of Europeans polled were found to be in favour of replacing a number of national parliamentarians with an AI that would have access to their data.

The appeal of rule by AI, or algorithmic governance, is understandable. AI has proven to be superior to human decision-making in some areas, particularly when there is a need for advanced strategic reasoning based on the analysis of large data sets. Politicians must regularly face complex issues where long-term and short-term interests have to be balanced and decisions must be made with speed. Just think of the Covid-19 pandemic. But this requirement does not play to human strengths.

There is research that suggests that an AI economist, for example, could make better tax policies than any human. Policies that would improve productivity while reducing income inequality.

Given our instinctive, anthropomorphic worldview, it can feel shocking to imagine a world of AI technocrats, shaping and executing policy. This is especially the case since we already know there are a variety of problems with algorithmic intelligence. These include AI’s recreation of the biases and assumptions of its programmers. After all, artificial intelligence is a privately designed system.

THERE IS ALSO a lack of transparency and accountability in many of its decisions. AI decisions are not explicable to mortal minds because of the vast amounts of data they compute to arrive at any conclusion. Were we to decide to use AI to, for example, select which criminal qualifies as a terrorist, or which citizen is more worthy of a subsidy than another, that selection procedure would remain opaque to the human mind. It is extremely difficult to reverse engineer AI computations given the complexity and quantity of data involved.

And these problems, while grave, stop short of what is in fact the ultimate fear of putting AI in charge: what if there is a glitch in the system and it decides that humans should be decimated?

Against this apocalyptic train of thought, there are arguments that suggest such fears to be overblown. The idea some say is for a system where humans have control over the direction and fundamental goals of a society, while AI should figure out the best, most efficient way to achieve those goals.

In this sense, an algocracy, or rule by algorithm, would be similar to a technocracy, wherein political legitimacy would derive from outcome rather than process. In a democracy, rulers gain their legitimacy from the process of having been voted in, rather than the outcome of their policies. Rule by AI, on the other hand, would be considered legitimate if it delivered on certain social/ economic goals: on outcomes.

It is unsurprising, when explained this way, that according to the IE University survey, the nation where the maximum number of citizens expressed a desire to replace some politicians with AI was China—about 75 per cent of respondents.

The fact is that, for good or bad, AI is already an intrinsic part of the political landscape. The New York Times called Argentina’s recent election, “The First A.I. Election,” referring to the extent to which the political campaigns of both contestants used AI-doctored images and AI-devised messaging.

AI can deceive and confuse voters, casting doubt over what is real and adding to the disinformation that can be spread by social networks. But it can also assist political campaigns, if used responsibly, with content creation. It could help pollsters with predictions. It can sift through vast amounts of data to help inform policy decisions. It might be pressed into service to curb election manipulation and improve public services. It could give civil society inspiration for ideas generation and information gathering.

In the future, the possibilities are only restricted by the imagination. For example, AI could assist with raising political campaign contributions. There has been some discussion of AI being programmed to make a profit out of seed investment from a human by investing it well. That revenue stream could then be redirected to a political campaign.

It’s certainly a brave new world out there. And there have even been some experiments with AI establishing a political platform, based on which human candidates contest elections. For example, in 2022, an artist collective in Denmark launched the Danish Synthetic Party. The public face of the party was the AI chatbot, Leader Lars, which was programmed on the policies of Danish fringe parties and meant to represent the values of the 20 per cent of Danes who do not usually vote. While Leader Lars was never intended to be on the ballot, the human members of the Synthetic Party were committed to carrying out their AI-derived platform, if elected. In the event, the party failed to receive enough signatures to earn a spot on the ballot.

The merits of democracy have been debated since the days of ancient Greece, but the idea that humans may be expendable to politics is almost inconceivable. However, the general crisis of democratic politics around the globe, coupled with the emergence of powerful generative AI, means that it has become urgent to begin thinking deeply about the unthinkable. The future could very well be the algorithmic.

algocracyartificial intelligencedemocracyelectiongovernancegovernmnet policiestechnology