Artificial intelligence is increasingly being presented as more than merely a tool. In media discourse, it increasingly “knows,” “decides,” “chooses,” “learns,” and sometimes even “organises” social reality. A recent study by researchers from SWPS University and the University of Wrocław shows that the anthropomorphisation of AI is not merely a stylistic device. It is a communication strategy that affects how we understand technology, agency, and responsibility.
Artificial intelligence has become one of the most important topics in public debate in recent years. Following the release of generative tools such as ChatGPT to the wider public, AI ceased to be the exclusive domain of experts, programmers, and research laboratories. It entered everyday language and became part of conversations about work, education, medicine, law, culture, politics, and the future of society.
Along with this shift, the way AI is discussed has also changed. Increasingly, we no longer speak of it as a technology designed, implemented, and controlled by people, but as an independent participant in social processes. AI “suggests,” “analyses,” “understands,” “predicts,” “recommends,” “decides,” “discriminates,” “collaborates,” and even “knows better.” At first glance, these may appear to be ordinary linguistic shortcuts. The study shows, however, that their consequences may be far more significant.
The article “The Discursive Construction of Artificial Intelligence in the Media: In Gazeta Wyborcza from 2022 to 2025,” written by Mariusz Wszołek, Michael Fleischer, Wojciech Jastrzębski, and Annette Siemes, reconstructs the ways in which artificial intelligence is represented in Polish media discourse. Rather than analysing the technology itself, the authors examine the language used by the media to describe it.
They are primarily interested in the point at which AI remains a tool and the point at which it begins to be constructed as a quasi-subject — someone or something to which knowledge, intentions, decisions, responsibility, or agency are attributed.
What Was Studied?
The study was based on a corpus of articles published on Wyborcza.pl between 2022 and 2025 and labelled with the tag “artificial intelligence.” During this period, 1,149 texts concerning AI were published. From this material, the researchers extracted 2,781 passages referring directly to artificial intelligence and subjected them to a qualitative discourse analysis.
An important point of departure was the observation of a rapid increase in the number of publications about AI after 2022. At that time, artificial intelligence ceased to function as a specialist subject and became a permanent element of media discourse. In other words, AI was not only discussed more frequently, but also in increasingly established and recurring ways.
The study was qualitative and exploratory in nature. Its purpose was not to generalise statistically about all Polish media, but to reconstruct recurring models of speaking about AI within one important segment of public communication.
From a Tool to a Quasi-Subject
The analysis reveals a clear shift in the way AI is constructed in media discourse. In 2022, artificial intelligence was primarily described as a tool supporting human activity. It helped to analyse data, accelerated processes, and supported doctors, users, institutions, and organisations. It formed part of the technological infrastructure — a useful extension of human capabilities.
In 2023, AI increasingly appeared as a mediator of everyday choices. The media described it as a system that recommends content, suggests solutions, selects information, and organises access to the digital world. Within this model, the human being is still present, but increasingly as the recipient of decisions made or prepared by the system.
In 2024, the image of AI shifted further. Artificial intelligence began to be presented as a regulator and manager of social processes. It not only recommended, but also corrected, controlled, organised, and established standards. The language used to describe it began to resemble the language of institutions: education, public administration, management, and supervision.
The most developed model appears in material from 2025. There, AI is sometimes constructed as an epistemic authority, an environment in which people function, and a meta-institution that organises the social order. It is no longer merely a technology that we use, but something within which we begin to operate — a framework, a context, and sometimes an overarching point of reference for decisions, knowledge, and the organisation of social life.
Six Ways of “Humanising” AI
The researchers identified six strategies for anthropomorphising artificial intelligence.
- The first is cognitive anthropomorphism, in which AI is presented as a system that “knows,” “understands,” “learns,” “analyses,” or “predicts.”
- The second is intentional anthropomorphism, which involves attributing intentions, preferences, and decisions to AI — AI “wants,” “chooses,” or “pursues a goal.”
- The third strategy is emotional or quasi-emotional in nature. AI is described through references to emotions: as empathetic, cold, devoid of emotion, or capable of recognising emotions.
- The fourth strategy concerns morality. AI may be presented as fair, unfair, biased, ethical, or discriminatory.
- The fifth strategy is relational: artificial intelligence appears as a human assistant, partner, helper, competitor, or collaborator.
The most advanced form is ontological, or quasi-subjective, anthropomorphism. In this model, AI ceases to be described as a tool and begins to function as an independent actor. It acts, organises, establishes frameworks, and influences social processes.
Importantly, this often happens without explicitly identifying the people, companies, institutions, and design decisions behind the technology.
Why Does This Matter?
The most important conclusion of the study concerns responsibility. The way we speak about technology affects how we understand its social consequences.
When we say that “AI made the decision,” “the algorithm chose,” “the system assessed,” or “artificial intelligence knows better,” we can very easily lose sight of the fact that every technological system is backed by people: designers, programmers, company owners, institutions, regulators, business models, and specific organisational decisions.
The problem, therefore, is not that the media use metaphors. Language always relies on shortcuts, comparisons, and simplifications. The problem begins when metaphors start to replace questions about agency and responsibility.
When AI is presented as an autonomous actor, it becomes easier to accept its presence as something natural, inevitable, and independent of social decisions.
When we say that “AI made the decision,” “the algorithm chose,” “the system assessed,” or “artificial intelligence knows better,” we can very easily lose sight of the fact that every technological system is backed by people: designers, programmers, company owners, institutions, regulators, business models, and specific organisational decisions.
“We did not examine whether artificial intelligence really thinks, understands, or makes decisions. We examined how media language causes us to begin speaking about it as though it were an independent participant in social life,” explains Mariusz Wszołek of SWPS University.
As the authors emphasise, the anthropomorphisation of AI is not merely an everyday manner of speaking. It may perform a normalising function: it makes technology seem familiar and easier to understand, while at the same time potentially shifting responsibility away from people and institutions and towards supposedly “acting” systems.
“If we say that AI made a decision, it is worth asking: who designed the conditions under which that decision was made, who supplied the data, who implemented the system, and who is responsible for its consequences? Without these questions, it is easy to mistake a technology for an independent subject,” adds Wszołek.
The Media Do Not Merely Describe AI. They Co-Create Its Social Image
The study forms part of a broader reflection on the role of the media in shaping perceptions of new technologies. The media are not simply a mirror reflecting technological reality. Through their choice of words, metaphors, topics, and interpretative frames, they co-create the way society understands AI, what it expects from it, what it fears, and what forms of regulation it considers necessary.
The question of the language used to describe AI is therefore also a question of democracy, responsibility, and control over technology.
As artificial intelligence exerts an increasing influence on education, work, information, healthcare, public administration, and culture, the question of how we speak about AI becomes just as important as the question of how AI works.
The article “The Discursive Construction of Artificial Intelligence in the Media: In Gazeta Wyborcza from 2022 to 2025” has been accepted for publication and will appear in the second issue of the journal Prace Językoznawcze in 2027. Its authors are Mariusz Wszołek, Michael Fleischer, Wojciech Jastrzębski, and Annette Siemes.
Mariusz Wszołek 