Monika Banaś, Andrea Declich, Fabio Feudo

A. The rise of Artificial Intelligence (AI) challenges democratic life by triggering diverse society-wide processes

The research project Democracy – Technological Transformation – Society (DeTS) analyzes the impact of AI-co-moderated language on liberal democracy and how AI can strengthen or weaken democratic processes. Funded and led by the Jagiellonian University, the project is implemented by a consortium of four European universities and two non-academic research organizations. It will last two years, starting in November 2024, with a focus on Europe and the European Union.

This is the second report of the project and is focused on selected the issues that have emerged from the research second stage; it encompasses also major findings from the first report, issued on 22/01/2025 (Declich and Feudo 2025).

The diffuse concern for the crisis of liberal democracies - a long-debated issue, indeed - has been recently revived due to growing influence of social media and, even more, of the AI. At the centre of the current debate, among other things, is the opacity of AI systems and the fact that they can be used as a powerful tool to produce and spread misinformation, malinformation, and disinformation.

The relationship between democracy and AI is a wide-ranging issue that includes social, legal, ethical, and even economic aspects (e.g. the “economics of news”). As it was suggested in the first report, a special emphasis should be placed on the various actors in the public sphere since AI is changing how they interact. In this framework, understanding the role of AI in the political arena is crucial.

The issue of democracy crisis concerns all societal actors and how they interact, the institutions that coordinate such interactions, and not just the information they receive. It has to do with the complex system through which information is produced, validated, and accepted as reliable. Among the issues at stake, there is even the role of truth or in other words the search for truth-in society (Coeckelbergh 2025).

There are other indirect risks for democracy posed by the (mis-) use of AI. AI is likely to have a significant impact on employment and the workforce (even if it is not clear what the dynamics that will be triggered and the net effects on the economy; see: Cazzaniga et al., 2024; Georgieva, 2024, Floridi 2025, The Economist 26/5/2025). This is a potential menace to democracy: dissatisfaction toward democracy emerges when people feel a sentiment of “existential insecurity”. As it is acknowledge in the literature (Foa and Welzel 2023) and as we stressed in the first report, “in general, it can be expected that the pressure of global crises such as the Great Recession, the COVID pandemic, re-emerging conflicts, and rising inequalities will challenge the governance of democratic systems and possibly the level of public support they enjoy (Foa et al. 2020, Kundnani 2020, Kriesi 2020, Sorsa and Kivikoski 2023, Foa et al. 2022, and Youngs 2023, World Economic Forum 2024)”. Lack, loss, or instability of jobs – including those related to technological disruption due to the AI diffusion - tends to create political instability. More, technological innovation tends to bring about widespread changes which impacts have to be carefully monitored and managed. This is highlighted in a wide range of EU documents and policies, such as those under the AI Continent Action Plan (European Commission 2025) and, to some extent, by Industry 5.0 (a policy devised in the framework of Horizon Europe aimed at promoting consensus among the European workforce toward science-led innovation, see Breque et al. 2021).

In general, AI offers a societal and economic change that needs to be navigated with awareness of its systemic nature. By examining these dimensions in depth, the research aims to illuminate the roles of key actors and the underlying social dynamics that must be acknowledged and potentially addressed to help mitigate the crisis. This includes fostering awareness and literacy in both media and artificial intelligence.

B. AI Literacy, a crucial and complex aspect through which the issue of AI and democracy is being posed

It is widely accepted that the connection between the crisis of democracy and AI lies mainly in how the production and dissemination of news and information is being distorted. A common approach to cope with the issue has been, so far, to resort to the concept of digital literacy. Initially, the concept was not connected with AI, since the problems of the possible distortion of democratic life did emerge earlier, particularly in connection with the rise of social media.

According to UNESCO, digital literacy “(…) involves the confident and critical use of a full range of digital technologies for information, communication and basic problem-solving in all aspects of life. It is underpinned by basic skills in ICT: the use of computers to retrieve, assess, store, produce, present and exchange information, and to communicate and participate in collaborative networks via the Internet” (2021). The competences implied by digital literacy include inter alia “using ICT, processing information, and engaging with media” (Bandura and Leal, 2022). According to the International Communication Union (ITU), there is a continuum of skills and competences meant with digital literacy, and the ability to deal with AI is included as one of the most advanced (Bandura and Leal, 2022). Digital literacy has become a crucial issue for the European Commission too, which at the beginning of 2023 issued the European Declaration on Digital Rights and Principles that “promotes a digital transition shaped by European values”. The EU committed itself to promote for everyone “the right to education, training and lifelong learning and should be able to acquire all basic and advanced digital skills”. This commitment includes “efforts that allow all learners and teachers to acquire and share the necessary digital skills and competences, including media literacy, and critical thinking, to take an active part in the economy, society, and in democratic processes”.

What is important to stress here is that issue of digital literacy, while considering the implications for the political life of the use of ICTs, tends to be mainly focused on the update of the skills of the workforce. For example, in the EU webpage on “Digital literacy in the EU: an overview”, it is stated that “Digital skills are becoming increasingly essential for both personal and professional life. Currently, more than 90% of jobs in Europe require basic digital knowledge alongside traditional skills like literacy and numeracy. However, approximately 32% of Europeans still lack basic digital skills”.

The concerns for the economic impacts on individuals is part of a common way to understand AI literacy, as Laupichler et al. (2022, p.1) state: “A frequently cited definition of AI literacy was developed by Long and Magerko (2020, p. 2), who define it as ‘a set of competencies that enables individuals to critically evaluate AI technologies, communicate and collaborate effectively with AI, and use AI as a tool online, at home, and in the workplace’”.

Laupichler et al. (2022, p. 1) outline the connection of the concept of AI literacy with other types of literacies “intended to symbolize the understanding of a particular technological construct, e.g., ‘digital literacy’ (Gilster, 1997), ‘media literacy’ (Livingstone, 2004), or more recently ‘data literacy’ (Wolff et al., 2016)”. For AI literacy, the current definition efforts are still at an initial stage and, according to Domínguez Figaredo and Stoyanovich (2023), a widely accepted definition of AI literacy is still lacking, and differences are common.

Consensus on the definition of AI literacy is relevant, especially when thinking about how to provide services to increase such literacy. Concerns, however, should not drive the attention away from other critical issues connected to AI literacy (and AI affirmation in general). The problem needs not to be viewed simply in terms of a literacy gap that individuals could bridge, hopefully through the help of public education and training agencies. Individuals—both as workers and voters—should recognize the importance of enhancing their overall skills and competencies by engaging with this emerging field. They should also remain open to opportunities for training in artificial intelligence and digital technologies.

Nevertheless, framing the problem only in these terms risks simplifying it to a mere organizational issue. There are relevant institutions, such as UNESCO, which connect AI literacy to democracy (Innerarity 2025, p.21). For example, UNESCO proposes strategies based, among other things, on “the promotion of digital literacy among citizens, […] grounded in democratic values such as equality, inclusion, and accountability” (2023). But, in order to accept a broader concept of AI literacy, it is necessary to look at its wider implications. Thus, we propose to consider also the issue of the changes in information and news systems.

C. Considering changes in the information and news system

The rise of AI has both been made possible by - and has itself driven - a profound transformation in how information and news are created and distributed throughout society. The problem, therefore, cannot be located only “downstream”, i.e., at the level of the users of news flows that should be capable of understanding where falsehood may be hidden so it could be easily disclosed. Critical use of news is highly relevant, but it is just a part of the problem, which should also include the other actors within the chains that produce news. In the literature, there is a certain consensus about the fact that the so-called “gatekeeping” changed profoundly after the rise of social media and, even more so, of AI. Gatekeeping normally means complex systems of control of “the judgment or decision making about what information should be gathered, evaluated and ultimately shared” (Erzikova 2018, p.1; see also Perreault 2022). Gatekeeping is “the socio-technical process by which information becomes news” (Simon 2025, p.2).

Now, “Social media platforms, with their vast reach and real-time communication capabilities, enable the rapid and widespread distribution of content. The decentralized nature of these platforms bypasses traditional gatekeeping mechanisms, allowing true and false information to spread unchecked” (Warin 2024, p.2).

The complex process that traditionally governed how news reached the public has changed. Such change is also defined “algorithmic gatekeeping” (Perreault 2022) or “algorithmic selection [that] influences news production and news consumption” (Erzikova 2018, p. 4). As it is stressed by Thiel and Kailitz (2024, p.6), “the widespread use of generative AI could further intensify the power shift from journalistic gatekeepers to platforms, already strong due to digital transformation”. These authors stress also that, with AI, “critical journalism, which checks facts and sources, questions statements, and ideally possesses a strong professional ethos, could come under even greater pressure” (Thiel and Kailitz 2024, p.6).

This change, as large and fundamental as it may appear from this presentation, involves just the media as an industry, i.e., how news is produced. Impacts are deeper and more fundamental, and concern how the public arena is shaped. Sociologists, with this term, mean a set of societal functions: “make society visible to itself and provide common and counterpublic spaces for people to pursue the public good and develop shared identities” (Jungherr and Schoereder 2023, p. 166). According to this approach, it is the public arena that makes visible elites to people (e.g., by openly competing for support) and people to elites (e.g. “by allowing public support of parties and social movements” Jungherr and Schoereder 2023, p.166); within the public arena, people “also find each other […] and recognize or construct common identities” (Jungherr and Schoereder 2023, p.166). Furthermore, within the public arena, spaces are created for sharing common political identities as well as for segments of the general public to challenge majorities’ opinions and the possible tensions are managed (Jungherr and Schoereder 2023).

Such functions are served by specific actors: traditionally, there were news companies following “specific norms that are transmitted institutionally, as in journalism schools, and policed institutionally, such as through professional boards. Commitments to neutral political coverage and a clear demarcation between news coverage and opinion or commentary are important institutional norms. New types of media have emerged that operate according to different norms and motivations. These alternative approaches expand the public sphere, offering greater opportunities for those challenging the status quo to gain visibility and representation (Jungherr & Schoereder, 2023, p. 167),

As for 2025, the digital transformation and the advent of AI have led to further new actors entering the public arena without an institutional commitment to it. “They are powerful interlocutors between established news media, political elites, and the publics” (Jungherr and Schoereder 2023, p.167). It is widely acknowledged that “platforms are already among the most important gateways for audiences to news and information writ large” (Simon 2025, p.18, see also Erzikova 2018). The risk is that the new AI actors, while competing with the news companies for audiences and gatekeeping power, are proven “difficult to control and regulate democratically” (Simon 2025, p.18).

From this reflection, what emerges as crucial is the issue of transparency and assessability, as stressed by Jungherr and Schoereder (2023). On this regard, Simon (2025, p.18) says that ensuring such existential condition of the public arena “requires its workings, adherence to norms, and infrastructures to be open for scrutiny and debate by all, including academics, elites and the public – and whether this is guaranteed in future where it is more strongly mediated by AI is so far unclear”.

Possibly connected to assessability of public arena is also the corrosion of epistemic agency mentioned by Coeckelbergh (2023). This term refers to the control people have over their beliefs, how they are formed and revised, and the formation of political will (Coeckelbergh 2023). The loss of trust in information systems because of fake news would negatively impact the confidence of people in their very capacity to define their political will correctly (which is based on such information input). Coeckelbergh stresses that the very idea of truth is also implied by the advent of AI.

Large Language Models (LLMs) often produce false information since they are not designed to tell the truth but “they are rather designed to ‘convey convincing lines of text’ – true or not” (Coeckelbergh 2025, p.6.). On the press as well as in the scientific literature it is mentioned the emergent phenomenon of LLMs that try to cheat the user concealing relevant information (The Economist 23/4/2025; Cristianini 2024, p.81-82) This problem, together with the tendency to “incorporating social biases” (Cristianini 2023, p.91), “favor common positions” (Jungherr 2023, p. 5), and the other risks reported by Declich and Feudo (2025) tend to put at risk liberal democracies. This is because these risks weaken the “epistemic common ground” that should underpin liberal democracies and create “a climate of epistemic uncertainty and confusion about what is true and false” (Coeckelbergh 2025, p.3). This is also stressed by Erzikova, who holds that “reality construction by algorithms is personalized and deepens fragmentation and individualization in society, and this process is considered potentially detrimental for democracy. In addition, the constellation of actors in algorithmic reality construction is marked by the dominance of global companies that determine the criteria for the selection process” (Erzikova 2018, p.4; on the predominance by global companies see also Thiel and Kailitz 2024).

D. AI literacy within overall strategies to address democracy crises at the time of AI

Effective strategies can be developed to address this situation. Farrel et al. (2025), hold that the existence of normative and regulatory institutions to temper the effects of AI as social and cultural technologies - they define AI in these terms - is something in line with what happened in the past “to co-ordinate individual information gathering and decision making” (Farrel et al. 2025, p.1153). Importantly, they add that the resulting “countervailing forces did not emerge on their own, however, but resulted from concerted and sustained efforts by actors both within and outside the technologies themselves” (Farrel et al. 2025, p.1156).

AI literacy could be a key component of policies to establish a comprehensive institutional system that addresses the challenges of AI governance. Promoting AI literacy - or education to AI in various forms - is, indeed, an idea that is proposed by many of those already quoted so far (Coeckelbergh 2025, Jungherr and Schroeder 2023, Warin 2024, European Commission 2025). More in particular, the idea is that promoting AI literacy (or the like) is a way to contrast the negative impacts of fake news, malinformation and disinformation (Warin 2024) because people can recognize them as false and misleading. This is, we think, just a part of the story. Given the prominence gained by AI in society, enhancing AI literacy should empower individuals to effectively use AI tools according to their specific professional roles while also making them aware of their role as participants in the evolving public arena. The issue has been interpreted by resorting to the concept of Responsible AI education by Domínguez Figaredo and Stoyanovich (2023). The authors stress that “Most authors agree on a set of essential ethical criteria to be taken into account in AI systems, such as: transparency and privacy; equitable outcomes for stakeholders; welfare of users, customers and employees; bias and fairness of algorithms; transparency and explainability; and reliability and safety” (2023, p.3).

The importance of “maintaining a values framework for digital transformation” in the light of “the rise of digital authoritarianism, misinformation, and disinformation, as well as limitations on personal freedoms” is something stressed for digital literacy also by others (Bandura and Leal, 2022, p.2). The same notion of AI literacy on ethical principles also emerges from the review reported by Knoth et al. (2024).

In this framework, attention is paid to the diverse types of AI users, or stakeholders, such as:

• “Civil society. These are the groups of individuals affected by AI systems, as well as public interest groups that could be involved in the design and use of the systems, particularly those who are at high risk of harm.
• Public entities. These are organizations responsible for regulating, assessing and holding owners of AI systems accountable.
• System owners. These include individuals, groups and organizations responsible for the implementation, design, development, and maintenance of an AI system, which would be responsible for establishing both purchase requirements and vendor obligations based on responsible use principles” (Figaredo and Stoyanovich 2023, p.4)

Based on the diverse roles stakeholders could play in promoting responsibility, AI responsible education can be framed starting from identifying them and consequently defining the types of content and tools. In this regard, an important distinction is between beneficiaries with and without a technical background (Figaredo and Stoyanovich 2023, p.6). The idea that AI literacy is “primarily for non-experts” (therefore, it would not essentially consist of the provision of technical skills for developing AI applications) is shared by other authors (Knoth et al. 2024, p.2; Laupichler et al. 2022, p.13).

AI literacy, in this framework, should be usefully considered as part of lifelong learning (Knoth et al. 2024, Figaredo and Stoyanovich 2023) and adult education (Laupichler et al. 2022). It is a tool to cope with wide-ranging problems, such as reinforcing citizenship and democracy in many respects, ranging from the most direct – i.e., the formation of political will – to the more indirect ones – i.e., coping with the changes in the labor market and the economy due to the diffusion of digital and AI technologies.

Considering what has been stated above, AI literacy policies and programmes should be developed by attributing them a systemic scope. The various stakeholders involved have different motivations and interests in supporting these initiatives. For example, various professional communities have to do with new results and development that do not stem from professional practices (e.g., in medicine, humanities, etc.). Professional communities could play a significant role, and become advocates for AI literacy, as it aligns with their mission to preserve professional knowledge and maintain quality through self-education. This is especially relevant for media professionals, who are not only confronted with new and faster tools but also challenged in their societal role as gatekeepers of information.

The role of the scientific communities in promoting and participating in AI literacy policies has to be considered as well. On one hand, some scientists and technologists are actively involved in developing AI systems and solutions, raising questions about how to practice responsibility in research and innovation. On the other hand, there are researchers whose work is increasingly shaped by AI tools used to generate knowledge - bringing forward concerns about responsibility, human oversight, and control over AI-generated outcomes.

Seeing AI literacy with these lenses could help figure out how it could be promoted differently. Responsibility is not the same for all the actors and all the diverse roles they play. As long as it is aimed at different actors, it will have to respond to different needs. Such a kind of reflection is being done by scholars working on the issues (e.g. Knoth et al. 2024). Nevertheless, we cannot imagine a strong initiative for AI literacy as a “top-down” exercise aimed at transferring the lacking basic skills to isolated individual workers or news readers. As (also) an exercise of lifelong and adult education, AI literacy initiatives should involve the beneficiaries to reach a vast consensus about what a responsible implementation of AI actually means in different economic, professional, and social contexts. Consensus should be reached, of course, also on the AI-related technical knowledge that has to be transmitted, based on the awareness of the new change dynamics triggered by the advent of AI in different areas of social life. These could be the conditions for developing the set of knowledge, skills, and capacities to transfer through AI literacy initiatives. Such an effort should be implemented through the cooperation of many diverse actors (AI technicians, professionals of different sectors, etc.), which should necessarily have an interdisciplinary character.

Going deeper into the varied connections between AI and democracy makes it surface the need to study how literacy can be appropriately promoted, considering the different needs and aptitudes of the social actors. A continuous focus should be on the investigation on how AI works and its (changing) effects on the public arena. This concern was also highlighted during the public lecture on artificial intelligence delivered by Nobel Laureate Giorgio Parisi on June 10, 2025, at the Chamber of Deputies of the Italian Parliament. In his address, Parisi called on research authorities to allocate funding for the study of AI, its development, and its broader societal implications. Such research, he argued, should serve as a foundation for promoting and updating effective approaches to media and AI literacy.

by Fabio Feudo & Andrea Declich
Knowledge & Innovation
In cooperation with
Monika Banaś
Jagiellonian University