Skip to content
Florence School of Transnational Governance

New STG Chair analyses the relation between AI and Democracy

In the latest contribution for ‘Firenze, idee d’Europa’, a series by the EUI and La Repubblica Firenze, STG Professor Daniel Innerarity writes about the challenges of Artificial Intelligence for Democracy.

30 June 2022 | Opinion - Policy dialogue - Research

STG_AI_Chair_Launch

Prof. Innerarity is the first Chair in Artificial Intelligence (AI) and Democracy newly established at the EUI School of Transnational Governance, with support of the Spanish Secretary of State for Digitalisation and Artificial Intelligence, and the Institute of Democratic Governance.

Prof. Innerarity and his team study the effects of AI on democratic systems and the possibilities that the technological infrastructure linked to digitalisation offers for improving democratic systems.

The opinion below appears as academics, policymakers, officials and members of civil society meet in Florence to celebrate the launch of the new Chair, exchanging views on its priorities in a transnational context. The Policy Dialogue The Democratic Challenges of Artificial Intelligence takes place on Thursday 30 June and Friday 1 July.

 



Let Siri or Alexa vote for us?

Prof. Daniel Innerarity


At the beginning of 2013, the MIT Technology Review headlined an issue as follows: "Big Data Will Save Politics". Just five years later, in autumn 2018, under the impression of the Cambridge Analytica scandal, fake news and hate speech on the internet, the same magazine stated on its cover: "technology is threatening our democracy. How do we save it? A year later, The Economist was already talking about an "AIthoritarianism" that could destroy democratic institutions.

This seesawing between expectations and disappointments is perhaps revealing that we do not know what happens to politics and democracy when the technological environment changes radically, what political transformations we associate with robotisation, digitalisation and automation. This ignorance explains the fact that two types of diagnoses have been formulated that imply, albeit for opposing reasons, a certain farewell to politics: the prophets of enthusiasm announce the absolute power of technology over politics, which they see as fundamentally a good thing. And some prophesy that it could even serve the function of repairing or replacing weakened or absent political structures.

The new technology would solve the problems that the old politics has failed to address. The other diagnosis of the end of politics is pessimistic insofar as it holds the new technological environment responsible for the loss of governance over social processes and the de-democratisation of political decisions. Technophilia and technophobia share much more than what sets them against each other: a similar ignorance and the assumption that the logic of technology can replace that of politics; they differ only in considering this as good or bad news.

The challenge we face is conceptual rather than normative. Automation requires thinking about many sociocultural categories, such as subject, action, responsibility, knowledge and work. The three elements that will change politics in this century are increasingly intelligent systems, more integrated technology and a more quantified society. What we are asking ourselves is what democratic self-government means and what free political decision-making means in this new constellation. We need to develop a theory of democratic decision in an AI-mediated environment, to elaborate a critical theory of automatic reason.

The fundamental question is the place of political decision in an algorithmically governed society. Democracy is free decision, popular will, self-government: to what extent is this possible and does it make sense in the hyper-automated, algorithmic environments heralded by artificial intelligence?

Representative democracy is a way of articulating political power that attributes it to a particular body and according to a chain of accountability and legitimacy in which the principle that all power comes from the people is verified. From this perspective, the introduction of autonomised intelligent systems appears problematic. This problem is exacerbated in learning systems because the function that processes the data changes in the learning phase. The system works adaptively and not according to pre-programmed rules, making the chain of legitimacy and accountability - without which there is no democracy - more difficult to identify.

This would be the core of the problem. Democracy is about popular will, and when there is a change to the conditions under which the popular will takes decisions, our democratic practices are modified as well. The current increase in Automated Decision Systems (ADS) means that we are now in "an automated public sphere" (Pasquale). 

In both our societies and governments, there is a growing outsourcing of decisions toward decision-making systems based on algorithms. These automated landscapes are challenging democratic processes that stem from the popular will. Are democratic sovereignty and generalised automation compatible? As long as these processes are automatic, we cannot fully control them, but we should design the framework and values in which they are developed so we can continue to regard them as a true result of our will.

According to Lincoln's famous formula, democracy is a system of government in which the people are the owners, subjects and recipients of political action. To be in a position to answer the question of whether liberal democracy is inextricably linked to the analogue world, we need to elucidate what kind of political subjectivity corresponds to the people in the world of artificial intelligence, what kind of popular will is expressed in big data, how we decide when we sophisticate our automated processes. We need a Gettysburg Address for democracy in the age of artificial intelligence.

 

This article was first published in La Repubblica Firenze.

Last update: 30 June 2022

Go back to top of the page