In On Photography (1977), Susan Sontag described the modern world as “one vast quarry”, a place where former structures of meaning have faded, allowing the intellectual to salvage worthy pieces from the rubble. That may have been accurate for the 1970s. Nowadays, I’d say the world is no longer a quarry, but a dump. We are no longer scavengers struggling through rare fragments, but instead surfers of an endless landfill, overwhelmed by infomania. Our problem is not a lack of meaning, but being smothered by it.
In this environment, the act of (political) judgement becomes filtration: deciding what we focus on, what we ignore, what we discard.
The next question is then: Who performs the filtering? Societies have long outsourced such tasks to religious figures, politicians, editors and teachers. Today, people increasingly turn to non-human curators. Find any viral political post on X/Twitter and underneath you will find the same reply: “@grok is this true?”. (Grok is X’s inbuilt AI chatbot).
Initially, this outsourcing doesn’t seem extraordinary. Humans have always, to an extent, delegated judgement for a multitude of reasons. No-one can feasibly read or deliberate on every issue; we struggle with temporal, intellectual, linguistic or spatial limitations. People have spent decades outsourcing judgement to, say, CNN or FOX News — surely AI, with its greater access to data, is a “better” judge?
But there is a crucial difference here, one which lies in appearances. When one hands authority to a news outlet, a party leader or a role model, there is an inherent knowledge of context framing the judgement. You expect Fox to lean right, or The Guardian to lean left. Your brain can (typically) filter their filters, often subconsciously. I’m more likely to take a BBC News headline at face value than one from Infowars.
AI and algorithms have, for many, a mechanical sheen of impartiality. We are meeting with something digital, unconscious, detached. Their digital nature means we struggle to imagine its perspective in a concrete sense. But the appearance that its judgement is somehow unbiased and more accurate is precisely the source of danger.
Norbert Wiener, the father of cybernetics, warned against this fiction long before the rise of AI. In The Human Use of Human Beings (1950), he noted that “external messages are not taken neat, but through the internal transforming powers of the apparatus, whether it be alive or dead.”. No signal is ever neutral; every system has an apparatus that shapes and alters it.
Think of trying to have a conversation with your friend at a crowded party. Their voice is the signal; the background conversations and music is the noise. Your brain uses an apparatus to filter away noise in order to better follow the signal. Machines also perform this.
The key point is that every system, whether human or mechanical, imposes filters. There is no such thing as “raw” data. To decide what is recorded, what is amplified, and what is deemed “irrelevant” is already an act of judgement. What is deemed as “signal” is never without motive or purpose.
Acknowledging this, the fiction of AI’s objectivity collapses. AI does not cancel the informational noise of our age to uncover the “true” signal. Instead, AI is performing a task similar to modern search engines: ranking information according to a hidden criteria. To ask an AI model for “the truth” is not to find signal among noise, but to accept its ranking as reality—and to accept its apparatus as the filter through which the world is comprehended.
In Discipline and Punish (1975), Michel Foucault wrote that “power and knowledge directly imply one another; that there is no power relation without the correlative constitution of a field of knowledge”. Filters don’t merely sort but produce knowledge, deciding what is considered “truth” in the first place. Later, in Power/Knowledge (1980), Foucault described this boundary drawing as the creation of “regimes of truth”, a general politics of truth which establishes “the types of discourse which it accepts and makes function as true”.
This threat—that filters are framed as “neutral” curation rather than a form of control—was captured two decades ago in an underappreciated medium: video games. Metal Gear Solid 2 (2001) contained one of the most prescient speeches of the digital age. The AI antagonists (the “Patriots”) declared: “What we propose to do is not to control content, but to create context.”
The temptation is to let AI sort for us, not only for ease but also for their seeming neutrality. But coherence is not the same as truth, and computation is not the same as judgement.
This uncanny line appears alongside their diagnosis of a (then fictional) world drowning in “trivial information”, splintered into gated communities, each seduced by convenient “half-truths”. Their solution is to curate meaning from above. Such a vision is no longer abstract. Our present reliance on algorithmic filters moves us ever closer to the Patriots’ design, not through coercion, but through convenience and the voluntary provision of information—we exist in societies where it seems unique to not use social media. The comfort of frictionless coherence, whether from newsfeeds or chatbots, is tempting humanity to hand over the very work of political judgement itself.
In the print age, effective lying was far more difficult, given the work one had to do to modify the “factual texture” of reality: rewriting textbooks, editing archives, suppressing problematic facts. Today, digital media has made this substitution of reality near effortless in comparison. Posts can be deleted, filter bubbles established, history redrafted. The pace at which information is provided too makes contextualisation unrealistic. And indeed, the same plasticity has too enabled forms of resistance, allowing quasi-anonymous communication among insurgents in authoritarian nations. But this potential for emancipation coexists with the ease with which reality’s texture can now be rewoven. Reality can now shift so readily that people increasingly lose their political orientation.
A resulting cynicism often takes hold. When the Overton window—the range of ideas considered politically ‘acceptable’ in the mainstream at a given time—shifts so constantly, many cease to believe in anything. Hannah Arendt warned in Between Past and Future (1968) that “the result of a consistent and total substitution of lies for factual truth is not that the lies will now be accepted as truth… but that the sense by which we take our bearings in the real world… is being destroyed”. Once this sense is gone, political judgement becomes near impossible.
For Arendt, this judgement was not about asserting “truths” but rather cultivating what Kant called the “enlarged mentality”: the ability to view an issue from the standpoint of other people. The Greeks called this phronesis, the practical insight of the statesman as opposed to the wisdom of philosophers. To judge well in a political sense is to rise above private interest, a sort of radical empathy. Politics requires judgements be made in full view of that plurality of perspectives, even those which unsettle us.
Much of modern political governance has lost this capacity. Since the emergence of the New Right of the 1980s, waves of hyper-individualism have encouraged individualistic tribalism. The resulting political discourse is reduced to personal grievances, dismissing pluralism as “wokeness”, and framing governance as a zero-sum war rather than the building of communal world. Though the 1990s and 2000s saw a moderation of this, the current movement is something of a resurgence — the political centre of mass seems to be shifting rightward but also towards both extremes.
The emergence of social media has only served to accelerate this decline by rewarding anger and conviction over scepticism and reflection. Digital architecture is typically designed less for orientation than for consumption. We do not reflect over “feeds”; we are fed. This language itself reveals the passivity of most individuals’ relationship: you are the commodity.
Byung-Chul Han captures the heart of this passivity in The Agony of Eros (2017): “There is no such thing as data-driven thinking. Only calculation is data driven. The negativity of the incalculable is inscribed in thinking”. Data can count, but it cannot think. It may tally bricks but it cannot wonder what humanity might build with them. Real thought requires a distance: hesitation, consideration, dwelling upon the ambiguity of values and meaning that cannot be reduced to metrics. Calculation abolishes this otherness (“negativity”), and with it, thinking itself. AI accelerates this abolition, presenting instant correlations, predictions, and answers. To use AI for judgement is to mistake pattern-recognition for insight and logic for truth.
I found myself struck by Andreas Malm’s declaration in How to Blow Up a Pipeline (2021): “To act politically is to reject probability assessment as a ground for action”. Malm writes on climate struggle, but the phrase resonates more widely to me. It sharpens Han’s critique that thought begins in the incalculable, Malm reminds us that politics must begin there too. If thought is reduced to the calculation of (pre-filtered) data, and politics is reduced to probability, both are stripped of their radical power. Politics is not about accepting the “probable”, but about making room for the improbable — the rupture. These are the acts that change the world—those which cannot be justified by metrics alone.
Like a muscle going unused, outsourcing this thought results in judgemental atrophy. There is a dual issue: Cynicism erodes trust, and convenience corrodes exercise. What we are losing is not merely the ability or habit of deliberation, but the very expectation that politics demands it.
In a time of informational excess, we need the capacity to wrestle with material more than ever. The temptation is to let AI sort for us, not only for ease but also for their seeming neutrality. But coherence is not the same as truth, and computation is not the same as judgement. Every filter is a frame, there is power inscribed into those who draw the boundaries of knowledge. There is only one real safeguard against this atrophy: to resist the ease of frictionless coherence. Judgement cannot be automated.
Leave a comment