Empowering Citizens: Can AI Strengthen Deliberative Democracy?

• In Event: AI, Governance, and Power

Abstract

AI has been used in autocracies to control public opinion and to suppress political dissent. In democratic states, algorithmic innovations may have been contributing to polarization and to a decline in information trustworthiness. But could AI be used to enhance democracy? This paper aims to provide a theoretical and empirical framework for a digital infrastructure of a robust future democracy, by conciliating technological innovations with old democratic ideals.
A significant portion of citizens lacks the time or opportunity to thoroughly inform themselves about the activities of decision-makers. On top of that, an insurmountable amount of digital information makes it epistemically difficult to deliberate on the relative success of public policies or to hold decision-makers accountable for their electoral promises. The ongoing scenario contrasts deeply with the ideal that democracy's virtue relies on its capacity to use the reason and common sense of ordinary citizens. In order to strengthen democracy, we should therefore continue to promote evidence-based debate and civic education on the digital realm, without compromising freedom of expression and diversity of opinions/sensibilities.
Habermas' deliberative conception of democracy together with intergroup contact theory could help reduce political polarization on the digital sphere and reverse the trend of the growing distance between government and civil society (Katz and Mair). Dahl argued that the 21st century should bring new and creative technologies to enhance political engagement (Dahl). Now, AI has arrived, and it could be used to increase political knowledge, promote democratic accountability and strengthen democracy.
Natural Language Processing (NLP) and machine learning can accurately process vast amounts of linguistic and statistical data. These tools could summarize political information, facilitate deliberation and reduce the cost of acquiring political knowledge. Through the allocation of statistical information, public scrutiny of public choices could be enhanced, and the success of public policies more easily accessed. Citizens could be regularly informed about the decisions of elected officials. Public debate could be fostered and enriched through the attribution of different arguments and data from a diverse and reliable set of public, media and academic sources. Research institutes, journalists and higher education institutions would then play a substantial role in the 21st century infrastructure of civic education by enhancing digital democratic debate. The possibilities are endless. Citizens could become empowered by technology to become more fully informed and democratically demanding.
Building upon this theoretical understanding, we conduct experimental research on a digital platform designed to boost deliberative democracy. Quantitative and qualitative analyses showed: a rise in interactions among various political groups, an enhancement in citizens' political knowledge (as evidenced by pre and post-test surveys), a deepening of deliberation and argumentation quality (as shown by content analysis), and an improvement in the accessibility of public information. The overall effect of the independent variable (platform usage) on other variables is assessed through a Difference in Differences Model and through regression analysis (STATA). This digital framework for deliberative democracy could also be adapted to diverse political settings (local, national etc.) and could be applicable to different digital communication channels, such as social media networks.

Author

• Luís António Nunes de Andrade, Instituto Superior Técnico