Transnational Institute of Social Ecology

Habermas as a thinker against Google’s “Habermas Machine”

Written by Alexandros Schismenos, Ph.D.

Jürgen Habermas passed away on Friday 14 March 2026 at the age of 96. He was the worthy successor to the Critical Theory of the Frankfurt School and a great thinker on democratic politics and communicative action, or indeed on democratic politics as communicative action. Furthermore, he was one of the last great thinkers of the social-historical bridge between the analogue and digital ages and, as a critical theorist, he was particularly perceptive regarding the impact of digital technologies on the public sphere, tracing contemporary social transformations. Paradoxically, in 2024, towards the end of his life, he fell victim to the technocratic ideology he himself had criticised, when Google’s DeepMind unveiled an Artificial Intelligence model named the ‘Habermas Machine’.

Before we get to the ‘Habermas Machine’, let us briefly recall how the philosopher Habermas viewed technology.

As early as his first book in 1962, entitled The Structural Transformation of the Public Sphere [Strukturwandel der Öffentlichkeit], Habermas engaged with the concept of the public sphere as it takes shape as a field of negotiation between democratic civil society and the state political system. In this context, he understands the ideological function of technology in modern industrial societies, which undermines democratic social reciprocity to the benefit of capitalist mechanisms of exploitation.

In his 1968 essay “Technology and Science as ‘Ideology’” [Technik und Wissenschaft als Ideologie], Habermas uses arguments to refute the traditional claim that technology is neutral. He observes that, particularly in advanced capitalism, ‘technical problems’ are no longer distinguishable from ‘practical problems’ but merge into a ‘technocratic consciousness’ that reframes fundamental political issues as problems of technical management. Technocratic ideology leads to a shrinking of the public sphere as it undermines society into a ‘human-machine governance system’. Habermas’s critique of technocratic ideology in the late 1960s coincided with the explosion of information technology and the launch of the technoscientific programme of Artificial Intelligence.

Following the experience of the social movements of the 1960s, in his book The Theory of Communicative Action [Theorie des kommunikativen Handelns, 2 vols., Suhrkamp Verlag, Frankfurt 1981], Habermas distinguishes two basic spheres of contemporary society: the lifeworld and the system. The lifeworld is the symbolic realm of social meanings, cultural knowledge and normative certainties that are reproduced through communicative practices oriented towards mutual understanding. The system, by contrast, comprises institutional mechanisms of the material reproduction of the social, such as the market and the state, which coordinate action through the ‘regulatory means’ of money and power. Social activity is articulated and coordinated by communicative intersubjectivity. In rational terms, any discourse oriented towards intersubjective understanding, that is, mutual understanding, fulfils four claims to universal validity:

1. Intelligibility [Verständlichkeit], as every utterance must be comprehensible within a shared symbolic code.

2. The claim to truth: The propositional content must correspond to states of affairs in the objective world.

3. Normative Rightness: The utterance must conform to the normative expectations of the social world.

4. And sincerity [Wahrhaftigkeit]: The speaker must express their subjective intentions and feelings sincerely.

The aim of communicative action is a ‘consensus based on the intersubjective recognition of valid claims that are open to criticism’, which fosters social cohesion and solidarity without the need for external coercion. Honesty underpins mutual trust in intersubjective communication within the subjective realm, so that every utterance of valid discourse is a genuine expression of the speaker, as truth and normative rightness are claims to validity made by the interlocutors. However, mutual consent is opposed by the forces of systemic domination and the capitalist exploitation of the communicative function.

According to Habermas, a problem of modernity is the ‘colonisation of the lifeworld’, which occurs when systemic mechanisms of instrumental rationality penetrate areas of the lifeworld where the coordination of action ought to depend on intersubjective mutual understanding. However, this trend is being reinforced by the digital revolution. The technological colonisation of the public sphere is leading to a rise in anti-democratic technocratic tendencies in the socio-historical field.

In the digital age, the expansion of digitisation intensifies the colonisation of the living world within the microsocial sphere of everyday life, reinterpreting the underpinnings of symbolic meaning as a set of data that can be processed in terms of technical control. In 2022, in his latest work, “Reflections and hypotheses on a further structural transformation of the political public sphere” Habermas argues that the ‘platformisation’ of political discourse undermines the correspondence between governmental activities and the information provided to civil society.

Digitalisation marks a fundamental shift, eliminating the journalistic mediation and public critical function that once ensured quality control in the public sphere. Whilst digital networking empowers users as creators, the lack of criteria and the erosion of institutions of authority, such as traditional media, allow for the proliferation of ‘semi-public’ channels and closed echo chambers that silence dissent. This fragmentation leads to democratic regression, where citizens no longer perceive their differences of opinion as disagreements that enrich public dialogue, but as identity-based conflicts. The digital fragmentation of the public sphere severely undermines the possibility of a deliberative politics, where decision-making is based on the exchange of arguments. Habermas, as an advocate of representative democratic control, was particularly concerned about the impact of digital technology on the electoral representation of public opinion:

“[T]here must be a recognizable connection between the results of government action and the input of the voters’ decisions such that the citizens can recognize it as the confirmation of the rationalizing power of their own democratic opinion and will formation. The citizens must be able to perceive their conflict of opinions as both consequential and as a dispute over the better reasons”

To address the fragmentation of the digital public sphere, Habermas speaks of the ‘imperative of constitutional control’ over the digital world, that is, the establishment of binding regulations for digital society, and the creation of forms of digital “popular sovereignty” where the design and use of technologies will be publicly controlled so as to serve the needs of the user community.

However, despite his criticism of the prevailing trends of digitalisation and datafication, Habermas emphasised the communicative dimension of the public sphere and the deliberative aspect of democratic politics. At the philosophical core of Habermas’s conception of democracy lies the intersubjectivity of communicative rationality.

As we know, Habermas disagreed with Jacques Rancière, who, in our conversation in 2017 beneath the Acropolis, stated emphatically: “Democracy is not communication”.

Habermas advocated a model of open communicative rationality, in which the aim of democratic deliberative politics is to achieve mutual understanding, consensus and agreement based on validity. In contrast, Rancière argues that democratic politics is defined by public dissent [dissensus], which reveals, rather than crowns, deeply rooted inequalities of power.

It should therefore come as no surprise that Google’s DeepMind drew inspiration from Habermas’s philosophy of communicative action and named the AI digital mediator it unveiled in 2024 the [Habermas Machine] the digital AI mediator it unveiled in 2024.

The Habermas Machine is a Large Language Model [LLM] for digital deliberation among groups of people with conflicting political views, acting as an Artificial Intelligence mediator to help them reach consensus or at least find common ground on issues where they disagree. (Tessler, M. H., Bakker, et al 2024)

The intended aim of AI is to achieve a harmonious balance at the intersection of the ‘triple dilemma of deliberation’: the tensions between equality of participation, the quality of deliberation and the scale of the collective. The Habermas Machine works by providing questionnaires to members and synthesising the responses in order to formulate consensus statements that integrate different perspectives into common areas of agreement, whilst preventing the ‘tyranny of the majority.

The Machine follows a structured protocol designed to synthesise collective positions by comparing individual responses to structured questionnaires:

1. Input phase: Participants are presented with a question of public interest and asked to indicate their level of agreement and provide a short paragraph justifying their position.

2. Generation phase: The Large Language Model (LLM) records the individual responses and identifies points of agreement and disagreement, compiling a preference matrix that maps the spectrum of divergences and a series of potential consensus proposals.

3. Prediction phase phase: Next, a ‘reward model’ is used to predict how each member will rank the possible consensus proposals based on their initial response.

4. Simulated election: The system conducts a simulated election to select the most likely consensus proposal.

5. Refinement and Critique: The system proposes the selected consensus proposal to the members and accepts refinements and critiques from each member individually in order to run the production, prediction and simulated election until a consensus proposal is reached that is mutually acceptable to all members.

Google’s DeepMind conducted an experiment with the ‘Habermas Machine’ in 2024 involving 5,734 people from the UK, who were selected to reflect the demographic distributions of gender, age, income, and were divided into groups of five to discuss various political issues, with AI acting as mediator and facilitator. According to the researchers (Tessler et al.), 56% of participants stated that Artificial Intelligence seemed to them to be a better mediator and facilitator than traditional human intervention. It is easy to see that Google DeepMind’s Habermas Machine is essentially a parody of the great thinker’s philosophy of communicative action. (Palomo Hernandez 2025).

The algorithmic structure of Artificial Intelligence, by definition, precludes Habermas’s fundamental claims of validity. The operation of the Habermas Machine lacks comprehensibility – as simulations and choices are made automatically by the system – nor any claim to truth – as the system performs only statistical analysis – nor any ‘sincerity’ – as the parallel computational processes of Large Language Models produce structural opacity [black box].

On 16 April 2025, Habermas himself gave an interview to the Süddeutsche Zeitung newspaper, in which he publicly stated that he had had no involvement in the construction of the machine and protested that the use of his name was abusive and misleading, as it served ‘advertising purposes.’

Of course, the 94-year-old Habermas had the insight to recognise that productive Artificial Intelligence is the antithesis of any deliberative politics, as it represents the ultimate realisation of the fantasy of delegating and alienating of citizens from politics.

So Habermas, too, has fallen victim to the misinformation and digital barbarism inherent in the digitisation of communication. In reality, the ‘Habermas Machine’, like any Artificial Intelligence model, removes human subjectivity from the very process of deliberation, thereby abolishing by definition the public sphere of intersubjective communication that it is supposed to reproduce.

And what about the democratic question?

I was reminded of the old dispute between Castoriadis and Habermas regarding whether the institution of communication precedes it, or vice versa.

Habermas draws on language to explain intersubjectivity on the basis of communication theory. He disagrees with Castoriadis because he refuses to accept the primacy of the imaginary over the communicative. This move, however, restores the primacy of the symbolic over the imaginary, which Castoriadis had overturned. So when Habermas, in his book The Philosophical Discourse of Modernity, argues that Castoriadis ‘cannot offer any framework for mediation between the individual and society’, fails to take into account the social function of language as a primary institution of sociogenesis, nor the psychological function of transubstantiation which, for Castoriadis, highlights the individual as an institution of society, existence as temporality, and the socio-historical as the co-existence of the subjective and the objective.

Habermas therefore fails to take into account the intertwining of society and history, and the opposition between the soul and society, as presented in the social articulation of time in Castoriadis’s philosophy. The difference between the two thinkers essentially lies in the question of whether institution or communication comes first.

Castoriadis, in a letter to Jean-Marc Ferry dated 2 July 1985, as quoted by F. Dosse, wrote:

“I think the crucial question is precisely this, and it is extremely simple: what kind of communication is possible without institution, and in particular without language? Is language, after all, a neutral medium in relation to the object of communication?’

Castoriadis realises that prioritizing communication over institution would make institution a form of communication, whereas communication presupposes language, which is a fundamental form of institution of society, but also of the individual within society.

In this age of digital barbarism, the social-historical possibility of deliberative self-determination is threatened by a cognitive and political closure in which the future is the subject of algorithmic prediction and management. The ‘Habermas Machine’ can be seen as a symbolic product of this cognitive closure, an instrument designed to automate the very dialogue which, according to Habermas, ought to open up the horizon of social autonomy.

It seems that democracy, insofar as it concerns the conditions of our social life, cannot be reduced to a mutual public consensus without encompassing the open possibility of public dissent.

On the afternoon of Saturday 15 March 2026, a number of us met in person at the welcoming and beautiful TRISE venue, housed in the listed three-storey townhouse on Kolokotroni Street, for the launch of Yavor Tarinski’s book entitled HORIZONS OF DIRECT DEMOCRACY, during which we had the opportunity to chat online with his publishers in Atlanta via a live video link, a possibility afforded to us by global digital networking technologies.

Modibo Kadalie highlighted the chapter of the book that analyses the importance of physical presence and the bodily dimension of social coexistence for direct democracy.

He argued that this point is particularly relevant in our times, where digital technologies and telecommunications fuel theories of digital democracy or e-democracy, or even ‘appocracy’, theories that promote the passivity of individuals as digital consumers of virtual politics.

In the lengthy discussion that followed, we frequently referenced Habermas, whose name remains a living presence in the public sphere.

Bibliography

Castoriadis, C. (1987). The imaginary institution of society (K. Blamey, Trans.). MIT Press.

Crawford, K. (2021). Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.

Floridi, L. (2011). The philosophy of information. Oxford University Press.

Habermas, J. (1989). The structural transformation of the public sphere: An inquiry into a category of bourgeois society (T. Burger & F. Lawrence, Trans.). MIT Press. (Original work published 1962).

Habermas, J. (1970). Technology and science as “ideology.” In Toward a rational society: Student protest, science, and politics (J. J. Shapiro, Trans.). Beacon Press. (Original work published 1968).

Habermas, J. (1984). The theory of communicative action: Vol. 1. Reason and the rationalization of society (T. McCarthy, Trans.). Beacon Press.

Habermas, J. (1987). The theory of communicative action: Vol. 2. Lifeworld and system: A critique of functionalist reason (T. McCarthy, Trans.). Beacon Press.

Habermas, J. (1996). Between facts and norms: Contributions to a discourse theory of law and democracy (W. Rehg, Trans.). MIT Press.

Habermas, J. (2022). Reflections and hypotheses on a further structural transformation of the political public sphere (C. Cronin, Trans.). Theory, Culture & Society, 39(4), 145–171.

Marcus, G., & Davis, E. (2019). Rebooting AI: Building artificial intelligence we can trust. Pantheon.

Palomo Hernández, N. (2025). Towards automating deliberation? The idea of deliberative democracy embedded in Google’s Habermas Machine. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, 8(2), 1951-1960.

Schismenos, A. (2025). Artificial intelligence and barbarism: A critique of digital reason. Athens School.

Tessler, M. H., Bakker, M. A., Jarrett, D., Sheahan, H., van de Braak, M., Singh, H.,… & Summerfield, C. (2024). AI can help humans find common ground in democratic deliberation. Science, 386(6718), 216-221.