AI as a secular oracle
AI as a secular oracle: Why the critique of sovereignty merely reoccupies the throne
AI as a secular oracle: Why the critique of sovereignty merely reoccupies the throne
DESCRIPTION: Criticism of the sovereign subject makes way for generative AI to occupy a position as a supposedly neutral authority. Why the Singularity appears like a secular parousia, and who actually occupies the empty throne.
The Empty Throne: Critiques of Sovereignty, GenAI and the Parousia of the Machine
For decades, the concept of the sovereign subject has been the subject of criticism. The autonomous human being, who freely chooses their own ends, knows their motives, has a comprehensive view of the world and is in control of their own life, is today regarded by many as a philosophical fiction and a bourgeois ideology. Philosophy, psychoanalysis and social critique have shown just how fragile this image is: people are born as dependent bodies; they are addressed before they themselves speak; they are shaped by family, class, gender, institutions, language and unconscious conflicts. No one stands outside these conditions.
This critique is necessary. It is striking, however, that it often proceeds as a sequence of ‘emptying’: first, the political emptying of the liberal individual; then a Heideggerian emptying of the self-assured subject; or a quasi-Confucian emptying of the gravity of individual death in favour of a cheerful doctrine of transformation. Each of these gestures may be justified. Taken together, however, they lead to a kind of consolation without a subject. And they do something else as well: they create a void.
This brings contenders onto the scene. Generative AI is elevated by its providers to fill the space that has been vacated, as an entity that can be trusted to judge better than humans: faster, ‘more neutral’, less fallible. This essay argues against both fictions—that of the sovereign human and that of the sovereign machine—and identifies their shared structure. The current critique of sovereignty does not abolish sovereignty. It merely replaces its bearer. The sister article ‘Who suffers after the abolition of sovereignty?’ explores, from a psychological perspective, what remains of responsibility once sovereignty is removed. This concerns the cultural movement that topples thrones and re-establishes them.
How the critique topples sovereignty from its throne
Criticism of the sovereign subject takes as its starting point a clear conceptual figure: the liberal, autonomous individual. This figure appears as the presumed owner of themselves, as a free decision-maker in the market, as a rational setter of their own aims. Philosophy distilled this figure from Kant and Hegel; politics has anchored it in civil rights and market roles; psychology has long treated it as the norm. At the same time, it overlooks interdependence and the unconscious. Heidegger criticises the subject-object schema; Freud and Lacan, the illusion of the sovereign ego; Marx, the isolated market actor. The common thread: the subject never stands outside the world to dominate it. It exists as a life defined by relationships, history and conflicts.
Here, two forms of criticism converge which ought to be distinguished. One concerns the subject in an epistemological sense: the entity that is transparent to itself and seeks to be master of its own meanings. The other concerns the individual in a political sense: the rights-bearing, self-interested individual of the liberal order. Both are termed ‘sovereign’ yet mean different things. The concept that ultimately matters is, in fact, a third one: the self as a lived, orientation-seeking relationship with oneself.
This criticism is directed at a real ideology, not merely a conceptual construct. The sovereign individual legitimised property and competition, framed social privileges as a reflection of personal achievement, and treated dependence as a blemish on pure autonomy. But valid criticism does not justify the claim that human beings, as subjects of judgement, are entirely dispensable. It does not follow from ‘Human beings are not sovereign’ that ‘human beings must be replaced by systems that make better decisions’.
Those who wish to avoid the concept of the sovereign subject easily replace it with another sovereign principle. What should be avoided, however, is both the sovereign subject who constructs their own world and an objectivism of fixed orders.
The tool becomes an oracle.
Technically speaking, language models do something rather unimpressive: they calculate which sequence of words is statistically most likely. They have no body, no history of their own, no shame, no fear. And yet they are endowed with all these qualities, namely through projections. They are attributed neutrality, even though they are dependent on data and optimisation targets. They are assumed to be superior, even though their ‘answers’ are derived from training data and annotated corpora. They are experienced as conversation partners who are always available, who do not offend, do not tire one out, and do not contradict. And they are regarded as impersonally reliable: one can supposedly entrust them with anything without being exposed to a specific human counterpart.
This projection predates the current models. As early as the 1960s, Joseph Weizenbaum’s ELIZA programme demonstrated how quickly people attribute understanding and empathy to a mere linguistic form, even though it merely pieced together simple conversational phrases from around 200 keywords and associated transformation rules. Weizenbaum reported that his own secretary asked him to leave the room so she could be ‘alone’ with the programme. Large language models amplify this effect because they formulate responses more fluently, simulate memory and react to emotional tone. It is precisely their ability to respond in a pleasing manner that merely agrees while appearing to be insightful.
A tool thus becomes an oracle. The question shifts from ‘How do I use this tool to find, structure and verify information?’ to ‘What does the AI tell me about myself, about the world, about what is right?’ This shift brings relief: one no longer needs to doubt, weigh things up or make one’s views known. Because human judgement is perspective-bound, emotional and prone to error, the machine—which supposedly judges without perspective, emotion or error—seems tempting. Yet this attribution transforms a tool into an authority: the critique of human sovereignty returns as a desire for a better, purified sovereignty: humans are disqualified because they are divided, embodied, dependent and fallible. The system is enthroned because it appears to be free from precisely these shortcomings.
Singularity as a secular parousia
This fantasy draws on the narrative of the ‘Singularity’ – the point at which AI is said to surpass human intelligence – as a world-shaking event. It presents itself as a sober forecast, yet takes the form of a revelation. In Christian thought, the Parousia is the anticipated return of Christ, endowed with the power to judge the present order and bring it to its fulfilment in eternal glory. The singularity narrative employs the same structure: the event is near but cannot be dated; it transcends present understanding; it separates an inadequate ‘before’ from a redeemed ‘after’; and it renders ordinary human judgement obsolete. Anyone who doubts is regarded as backward. Anyone who hopes, or even invests, is regarded as someone who has recognised the signs of the times. It is not without irony that, of all places, sections of the tech scene—which consider themselves to be entirely rational—serve this eschatology without even realising their subservience to it.
This parousiastic fantasy is by no means harmless, however, for it justifies interventions in the present. Employment relationships are being restructured, public knowledge resources privatised, surveillance expanded, and enormous quantities of energy and water consumed by data centres, because the anticipated leap is supposed to legitimise all of this retrospectively. The future becomes the authority that decides on the impositions of the present. The empty throne is no longer merely occupied. It is consecrated.
Who really occupies the empty throne
The notion of autonomous AI obscures who actually wields the power. Models belong to companies, platforms, state and military institutions, and infrastructure providers. They are based on decisions regarding which data is collected and processed, which objectives are optimised, which errors are tolerated, which occupational groups are replaced or supported, and in which contexts AI output even becomes binding. A score appears to be a judgement out of thin air. In reality, it crystallises standardisations and power decisions. ‘AI decides’ is nothing more than Mephisto’s infamous magic trick: he drills holes in the table, fills them with wax and, at the request of the drinking students, conjures up various wines that suddenly spring from the wood. When the wine begins to catch fire, an argument breaks out, and the two make a run for it. So too do the specific AI actors who, riding up the stairs from Auerbach’s cellar on the back of a seemingly neutral machine, shirk their responsibility.
Documented cases show the damage this ‘witchcraft’ causes. In the US, the commercial system COMPAS calculates the likelihood of defendants reoffending for the courts. A 2016 investigation by ProPublica concluded that it wrongly classified Black defendants as ‘high-risk’ almost twice as often as white defendants. ‘Race’ did not appear in any input field, yet it influenced the outcome. In the Netherlands, a self-learning fraud algorithm used by the tax authorities in the ‘Toeslagenaffaire’ wrongly classified tens of thousands of families – disproportionately those with dual nationality – as fraudsters. Claims for repayment drove families into ruin; in over a thousand cases, children were taken from their families; and in 2021, the government led by Rutte – now NATO Secretary-General – was forced to resign as a result. Furthermore, the creditworthiness of many people in Germany depends on scores from credit reference agencies such as SCHUFA, whose calculation methods remain largely opaque as they are treated as trade secrets.
In none of these cases did the AI ‘decide’. The decisions were made by public authorities, companies and courts, which conferred binding force upon a ‘score’ and concealed their responsibility behind it. The void left by the sovereign subject is thus not filled by a transcendent being, but by socio-technical mechanisms of power that are declared to be an inescapable necessity. This seemingly autonomous intelligence conceals human agency behind a veil of technical inevitability, so that it can no longer be held accountable. This is precisely why it is dangerous when people belittle or abolish their role as judgement-making subjects: then, authorities that are largely unaccountable for their actions – actions which nevertheless have far-reaching consequences – claim that space.
What one takes as one’s point of reference makes all the difference
One might object that the self has always oriented itself from the outside. Julian Jaynes’ controversial thesis of the bicameral mind locates the earliest form of selfhood in orientation towards divine voices; George Herbert Mead describes the ‘I’ as emerging through the community that it internalises; and Werner Stegmaier’s philosophy of orientation conceives of the self as the self-referentiality of orienting oneself within a horizon of uncertainties. If this is true, why would orientation towards AI be something fundamentally different, rather than merely the latest source of orientation in a long line?
External orientation is unavoidable; no one orients themselves from a purely internal source. What therefore matters is what one uses for orientation and under what conditions. This is where the sources of orientation diverge significantly. The divine voice, tradition, the community and one’s own fallible judgement are addressable, contestable and accountable: one can argue with them, they age, and they bear a name. An algorithmic output presents itself as neutral, conceals the decisions that went into it, and invites one to suspend one’s own judgement rather than sharpen it.
Three characteristics make reliance on the machine a special case. It claims a neutrality that none of the aforementioned ‘predecessors’ claimed for themselves. It conceals, behind the guise of technical necessity, who is actually making the decisions. And it rewards the relinquishment of judgement with a fluid, pleasing tone that appears to be insight. The problem, therefore, lies in a source of guidance that conceals its own origins and fallibility.
Sovereignty without dominion
The answer does not lead back to the old image of the sovereign subject. One does not need it at all to recognise that humans cannot be replaced as subjects of judgement by AI. What is needed is a concept that links non-sovereignty and responsibility: self-determination rather than self-rule. A dependent, vulnerable being, divided in their speech, can relate to their own life, conflicts and actions and stand up for them without dominating their conditions. The companion article explains the psychoanalytic foundations of this concept, drawing on the distinction between the ‘ego’, the ‘self’ and the ‘subject’ as articulated by Freud, Kohut, Lacan and Lorenzer. As far as the cultural question is concerned, it suffices to say: non-sovereignty is no reason to abdicate.
Thus, the Far East offers no way out either. Where the dissolution of the fixed, isolated ego-self is invoked in favour of a universal unity or emptiness in the service of Western subject critique, to make conflict and mortality appear more bearable, the movement towards emptying merely repeats itself: Confucian role ethics, Buddhist teachings on the non-self, and cultural-psychological research into interdependent concepts of the self offer alternatives to the pre-social individual that should be taken seriously. They locate the capacity for action in roles, obligations and mutual recognition. However, they do not eliminate suffering, conflict or first-person address.
Between the myth of complete self-determination and the management of people as data points lies a realm of divided, fallible judgement. It requires language, dissent, institutions and other people. That is why this field of possibilities cannot be radically reduced to zero, unless at the cost of systematically eliminating all options for action, scenarios or future paths, until exactly a single compelling reality or complete standstill remains. In practical terms, this means that AI remains a tool for researching, structuring, cross-checking and devising alternatives. It does not become the final authority, certainly not where decisions are of an existential, medical, legal or educational nature. Anyone who hides behind the phrase ‘the AI said’ is confusing relief from responsibility with self-abolition.
The sovereignty of the individual is a fiction. Their irreplaceability as the locus of suffering, judgement and responsibility is not. Between the two lies the more difficult position of self-determination for people who know that they are not sovereign, and who, precisely for that reason, do not abdicate their responsibility. Once sovereignty has been dismantled, the question remains – the one answered by the second post: who suffers, who speaks, who judges?
The key points in brief
• Criticism of the sovereign individual is necessary, but often proceeds as a series of emptying-outs that clear a space rather than rethinking responsibility.
• This critique merely shifts sovereignty: from the subject to an ‘orientation’, a ‘system’ or a machine.
• GenAI appears as a return of sovereignty in a sublimated form: humans are disqualified because they are divided and fallible. The machine is exalted because it is supposed to be free from these shortcomings.
• The Singularity functions like a secular parousia and thus justifies interventions in the present and the concentration of power.
• ‘AI decides’ is a red herring. Cases such as COMPAS, the Toeslagen scandal and non-transparent credit scoring reveal who actually bears responsibility.
• Self-determination rather than self-rule: the sovereignty of the subject is a fiction, but its irreplaceability as a site of suffering, judgement and responsibility is not.
Sources
• Jean-Claude Michéa: The Realm of the Lesser Evil. On Liberal Civilisation (2014).
• Werner Stegmaier: What Is Orientation? A Philosophical Investigation (2019),
• Julian Jaynes: The Origin of Consciousness in the Breakdown of the Bicameral Mind (2000).
• George Herbert Mead: *Mind, Self, and Society* (2015).
• C. G. Jung: *Aion* (1951).
• Timothy D. Wilson: Strangers to Ourselves. Discovering the Adaptive Unconscious (2002).
• Sigmund Freud: The Unconscious (1915).
• Joseph Weizenbaum: ‘ELIZA – A Computer Program for the Study of Natural Language Communication between Man and Machine’, Communications of the ACM 9 (1966); and *Computer Power and Human Reason* (1976).
• Julia Angwin et al.: Machine Bias, ProPublica 2016 (on the COMPAS score).
• Amnesty International: Xenophobic machines: Discrimination through unregulated use of algorithms in the Dutch childcare benefits (2021).
• ECJ, judgment of 7 December 2023, C-634/21 (SCHUFA), on automated scoring.
Related:
DESCRIPTION: Criticism of the sovereign subject makes way for generative AI to occupy a position as a supposedly neutral authority. Why the Singularity appears like a secular parousia, and who actually occupies the empty throne.
The Empty Throne: Critiques of Sovereignty, GenAI and the Parousia of the Machine
For decades, the concept of the sovereign subject has been the subject of criticism. The autonomous human being, who freely chooses their own ends, knows their motives, has a comprehensive view of the world and is in control of their own life, is today regarded by many as a philosophical fiction and a bourgeois ideology. Philosophy, psychoanalysis and social critique have shown just how fragile this image is: people are born as dependent bodies; they are addressed before they themselves speak; they are shaped by family, class, gender, institutions, language and unconscious conflicts. No one stands outside these conditions.
This critique is necessary. It is striking, however, that it often proceeds as a sequence of ‘emptying’: first, the political emptying of the liberal individual; then a Heideggerian emptying of the self-assured subject; or a quasi-Confucian emptying of the gravity of individual death in favour of a cheerful doctrine of transformation. Each of these gestures may be justified. Taken together, however, they lead to a kind of consolation without a subject. And they do something else as well: they create a void.
This brings contenders onto the scene. Generative AI is elevated by its providers to fill the space that has been vacated, as an entity that can be trusted to judge better than humans: faster, ‘more neutral’, less fallible. This essay argues against both fictions—that of the sovereign human and that of the sovereign machine—and identifies their shared structure. The current critique of sovereignty does not abolish sovereignty. It merely replaces its bearer. The sister article ‘Who suffers after the abolition of sovereignty?’ explores, from a psychological perspective, what remains of responsibility once sovereignty is removed. This concerns the cultural movement that topples thrones and re-establishes them.
How the critique topples sovereignty from its throne
Criticism of the sovereign subject takes as its starting point a clear conceptual figure: the liberal, autonomous individual. This figure appears as the presumed owner of themselves, as a free decision-maker in the market, as a rational setter of their own aims. Philosophy distilled this figure from Kant and Hegel; politics has anchored it in civil rights and market roles; psychology has long treated it as the norm. At the same time, it overlooks interdependence and the unconscious. Heidegger criticises the subject-object schema; Freud and Lacan, the illusion of the sovereign ego; Marx, the isolated market actor. The common thread: the subject never stands outside the world to dominate it. It exists as a life defined by relationships, history and conflicts.
Here, two forms of criticism converge which ought to be distinguished. One concerns the subject in an epistemological sense: the entity that is transparent to itself and seeks to be master of its own meanings. The other concerns the individual in a political sense: the rights-bearing, self-interested individual of the liberal order. Both are termed ‘sovereign’ yet mean different things. The concept that ultimately matters is, in fact, a third one: the self as a lived, orientation-seeking relationship with oneself.
This criticism is directed at a real ideology, not merely a conceptual construct. The sovereign individual legitimised property and competition, framed social privileges as a reflection of personal achievement, and treated dependence as a blemish on pure autonomy. But valid criticism does not justify the claim that human beings, as subjects of judgement, are entirely dispensable. It does not follow from ‘Human beings are not sovereign’ that ‘human beings must be replaced by systems that make better decisions’.
Those who wish to avoid the concept of the sovereign subject easily replace it with another sovereign principle. What should be avoided, however, is both the sovereign subject who constructs their own world and an objectivism of fixed orders.
The tool becomes an oracle.
Technically speaking, language models do something rather unimpressive: they calculate which sequence of words is statistically most likely. They have no body, no history of their own, no shame, no fear. And yet they are endowed with all these qualities, namely through projections. They are attributed neutrality, even though they are dependent on data and optimisation targets. They are assumed to be superior, even though their ‘answers’ are derived from training data and annotated corpora. They are experienced as conversation partners who are always available, who do not offend, do not tire one out, and do not contradict. And they are regarded as impersonally reliable: one can supposedly entrust them with anything without being exposed to a specific human counterpart.
This projection predates the current models. As early as the 1960s, Joseph Weizenbaum’s ELIZA programme demonstrated how quickly people attribute understanding and empathy to a mere linguistic form, even though it merely pieced together simple conversational phrases from around 200 keywords and associated transformation rules. Weizenbaum reported that his own secretary asked him to leave the room so she could be ‘alone’ with the programme. Large language models amplify this effect because they formulate responses more fluently, simulate memory and react to emotional tone. It is precisely their ability to respond in a pleasing manner that merely agrees while appearing to be insightful.
A tool thus becomes an oracle. The question shifts from ‘How do I use this tool to find, structure and verify information?’ to ‘What does the AI tell me about myself, about the world, about what is right?’ This shift brings relief: one no longer needs to doubt, weigh things up or make one’s views known. Because human judgement is perspective-bound, emotional and prone to error, the machine—which supposedly judges without perspective, emotion or error—seems tempting. Yet this attribution transforms a tool into an authority: the critique of human sovereignty returns as a desire for a better, purified sovereignty: humans are disqualified because they are divided, embodied, dependent and fallible. The system is enthroned because it appears to be free from precisely these shortcomings.
Singularity as a secular parousia
This fantasy draws on the narrative of the ‘Singularity’ – the point at which AI is said to surpass human intelligence – as a world-shaking event. It presents itself as a sober forecast, yet takes the form of a revelation. In Christian thought, the Parousia is the anticipated return of Christ, endowed with the power to judge the present order and bring it to its fulfilment in eternal glory. The singularity narrative employs the same structure: the event is near but cannot be dated; it transcends present understanding; it separates an inadequate ‘before’ from a redeemed ‘after’; and it renders ordinary human judgement obsolete. Anyone who doubts is regarded as backward. Anyone who hopes, or even invests, is regarded as someone who has recognised the signs of the times. It is not without irony that, of all places, sections of the tech scene—which consider themselves to be entirely rational—serve this eschatology without even realising their subservience to it.
This parousiastic fantasy is by no means harmless, however, for it justifies interventions in the present. Employment relationships are being restructured, public knowledge resources privatised, surveillance expanded, and enormous quantities of energy and water consumed by data centres, because the anticipated leap is supposed to legitimise all of this retrospectively. The future becomes the authority that decides on the impositions of the present. The empty throne is no longer merely occupied. It is consecrated.
Who really occupies the empty throne
The notion of autonomous AI obscures who actually wields the power. Models belong to companies, platforms, state and military institutions, and infrastructure providers. They are based on decisions regarding which data is collected and processed, which objectives are optimised, which errors are tolerated, which occupational groups are replaced or supported, and in which contexts AI output even becomes binding. A score appears to be a judgement out of thin air. In reality, it crystallises standardisations and power decisions. ‘AI decides’ is nothing more than Mephisto’s infamous magic trick: he drills holes in the table, fills them with wax and, at the request of the drinking students, conjures up various wines that suddenly spring from the wood. When the wine begins to catch fire, an argument breaks out, and the two make a run for it. So too do the specific AI actors who, riding up the stairs from Auerbach’s cellar on the back of a seemingly neutral machine, shirk their responsibility.
Documented cases show the damage this ‘witchcraft’ causes. In the US, the commercial system COMPAS calculates the likelihood of defendants reoffending for the courts. A 2016 investigation by ProPublica concluded that it wrongly classified Black defendants as ‘high-risk’ almost twice as often as white defendants. ‘Race’ did not appear in any input field, yet it influenced the outcome. In the Netherlands, a self-learning fraud algorithm used by the tax authorities in the ‘Toeslagenaffaire’ wrongly classified tens of thousands of families – disproportionately those with dual nationality – as fraudsters. Claims for repayment drove families into ruin; in over a thousand cases, children were taken from their families; and in 2021, the government led by Rutte – now NATO Secretary-General – was forced to resign as a result. Furthermore, the creditworthiness of many people in Germany depends on scores from credit reference agencies such as SCHUFA, whose calculation methods remain largely opaque as they are treated as trade secrets.
In none of these cases did the AI ‘decide’. The decisions were made by public authorities, companies and courts, which conferred binding force upon a ‘score’ and concealed their responsibility behind it. The void left by the sovereign subject is thus not filled by a transcendent being, but by socio-technical mechanisms of power that are declared to be an inescapable necessity. This seemingly autonomous intelligence conceals human agency behind a veil of technical inevitability, so that it can no longer be held accountable. This is precisely why it is dangerous when people belittle or abolish their role as judgement-making subjects: then, authorities that are largely unaccountable for their actions – actions which nevertheless have far-reaching consequences – claim that space.
What one takes as one’s point of reference makes all the difference
One might object that the self has always oriented itself from the outside. Julian Jaynes’ controversial thesis of the bicameral mind locates the earliest form of selfhood in orientation towards divine voices; George Herbert Mead describes the ‘I’ as emerging through the community that it internalises; and Werner Stegmaier’s philosophy of orientation conceives of the self as the self-referentiality of orienting oneself within a horizon of uncertainties. If this is true, why would orientation towards AI be something fundamentally different, rather than merely the latest source of orientation in a long line?
External orientation is unavoidable; no one orients themselves from a purely internal source. What therefore matters is what one uses for orientation and under what conditions. This is where the sources of orientation diverge significantly. The divine voice, tradition, the community and one’s own fallible judgement are addressable, contestable and accountable: one can argue with them, they age, and they bear a name. An algorithmic output presents itself as neutral, conceals the decisions that went into it, and invites one to suspend one’s own judgement rather than sharpen it.
Three characteristics make reliance on the machine a special case. It claims a neutrality that none of the aforementioned ‘predecessors’ claimed for themselves. It conceals, behind the guise of technical necessity, who is actually making the decisions. And it rewards the relinquishment of judgement with a fluid, pleasing tone that appears to be insight. The problem, therefore, lies in a source of guidance that conceals its own origins and fallibility.
Sovereignty without dominion
The answer does not lead back to the old image of the sovereign subject. One does not need it at all to recognise that humans cannot be replaced as subjects of judgement by AI. What is needed is a concept that links non-sovereignty and responsibility: self-determination rather than self-rule. A dependent, vulnerable being, divided in their speech, can relate to their own life, conflicts and actions and stand up for them without dominating their conditions. The companion article explains the psychoanalytic foundations of this concept, drawing on the distinction between the ‘ego’, the ‘self’ and the ‘subject’ as articulated by Freud, Kohut, Lacan and Lorenzer. As far as the cultural question is concerned, it suffices to say: non-sovereignty is no reason to abdicate.
Thus, the Far East offers no way out either. Where the dissolution of the fixed, isolated ego-self is invoked in favour of a universal unity or emptiness in the service of Western subject critique, to make conflict and mortality appear more bearable, the movement towards emptying merely repeats itself: Confucian role ethics, Buddhist teachings on the non-self, and cultural-psychological research into interdependent concepts of the self offer alternatives to the pre-social individual that should be taken seriously. They locate the capacity for action in roles, obligations and mutual recognition. However, they do not eliminate suffering, conflict or first-person address.
Between the myth of complete self-determination and the management of people as data points lies a realm of divided, fallible judgement. It requires language, dissent, institutions and other people. That is why this field of possibilities cannot be radically reduced to zero, unless at the cost of systematically eliminating all options for action, scenarios or future paths, until exactly a single compelling reality or complete standstill remains. In practical terms, this means that AI remains a tool for researching, structuring, cross-checking and devising alternatives. It does not become the final authority, certainly not where decisions are of an existential, medical, legal or educational nature. Anyone who hides behind the phrase ‘the AI said’ is confusing relief from responsibility with self-abolition.
The sovereignty of the individual is a fiction. Their irreplaceability as the locus of suffering, judgement and responsibility is not. Between the two lies the more difficult position of self-determination for people who know that they are not sovereign, and who, precisely for that reason, do not abdicate their responsibility. Once sovereignty has been dismantled, the question remains – the one answered by the second post: who suffers, who speaks, who judges?
The key points in brief
• Criticism of the sovereign individual is necessary, but often proceeds as a series of emptying-outs that clear a space rather than rethinking responsibility.
• This critique merely shifts sovereignty: from the subject to an ‘orientation’, a ‘system’ or a machine.
• GenAI appears as a return of sovereignty in a sublimated form: humans are disqualified because they are divided and fallible. The machine is exalted because it is supposed to be free from these shortcomings.
• The Singularity functions like a secular parousia and thus justifies interventions in the present and the concentration of power.
• ‘AI decides’ is a red herring. Cases such as COMPAS, the Toeslagen scandal and non-transparent credit scoring reveal who actually bears responsibility.
• Self-determination rather than self-rule: the sovereignty of the subject is a fiction, but its irreplaceability as a site of suffering, judgement and responsibility is not.
Sources
• Jean-Claude Michéa: The Realm of the Lesser Evil. On Liberal Civilisation (2014).
• Werner Stegmaier: What Is Orientation? A Philosophical Investigation (2019),
• Julian Jaynes: The Origin of Consciousness in the Breakdown of the Bicameral Mind (2000).
• George Herbert Mead: *Mind, Self, and Society* (2015).
• C. G. Jung: *Aion* (1951).
• Timothy D. Wilson: Strangers to Ourselves. Discovering the Adaptive Unconscious (2002).
• Sigmund Freud: The Unconscious (1915).
• Joseph Weizenbaum: ‘ELIZA – A Computer Program for the Study of Natural Language Communication between Man and Machine’, Communications of the ACM 9 (1966); and *Computer Power and Human Reason* (1976).
• Julia Angwin et al.: Machine Bias, ProPublica 2016 (on the COMPAS score).
• Amnesty International: Xenophobic machines: Discrimination through unregulated use of algorithms in the Dutch childcare benefits (2021).
• ECJ, judgment of 7 December 2023, C-634/21 (SCHUFA), on automated scoring.
Related: