Philosophy
If you can’t escape algorithmic sorting, you can still disrupt it by becoming illegible to machines, not invisible to society.
Illegibility, Friction, and Collective Refusal
If we can’t get outside the system, we can at least make ourselves harder to sort. This is what is meant by illegibility: practices that interfere with algorithmic categorisation. Not invisibility (which is only available to the privileged), but strategic noise. Ways of presenting that resist easy classification. Ways of being that overflow the categories prepared for us. Some of this is technical: tools that obscure tracking, poisoning data profiles with misleading information, and refusing consistent patterns enabling prediction. Security researchers call this “adversarial input” data designed to confuse machine learning. What would it mean to live our lives as adversarial input to algorithmic purpose assignment?
But most illegibility is not technical. It is social, cultural and aesthetic. It is refusing coherent identity performance in contexts designed to extract coherent identity as data. Social media platforms encourage coherent self-presentation. Your profile should communicate clearly what kind of person you are. Your posts should be consistent with that persona. The algorithm rewards coherence because coherence is predictable and predictability is monetizable. To be illegible is to refuse that coherence. To post things that do not fit together. To like content spanning incommensurable categories. To perform multiple contradictory selves rather than a single stable identity. This is not being fake or strategic, that is still legible. This genuinely contains multitudes, genuinely being inconsistent, genuinely refusing to add up to a targetable demographic profile. Women already know versions of this. The “good girl” who drinks whiskey. The femme presenting woman who does construction. The mother who doesn’t want to talk about her kids. Every refusal to perform assigned gender creates small frictions in sorting systems. Marginalised communities know this too. Code-switching, strategic invisibility, playing dumb, performing compliance while practising refusal are survival strategies honed over generations of living under surveillance.
The question is whether individual illegibility can scale to collective resistance. Collective illegibility might look like, first, Organised opacity: groups collectively refusing to provide data, sharing accounts and devices making individual tracking impossible, coordinating action through channels, leaving no readable traces. This already happens in activist contexts, but could expand to broader practice. Second, Strategic incoherence: deliberately flooding platforms with content resisting categorisation, making data so noisy that prediction becomes unreliable. Third, Refusal of algorithmic authority: not treating algorithmic recommendations as authoritative or building social practices centring human judgment over computational judgment.
The challenge is that these practices are exhausting. They require constant vigilance, collective coordination, willingness to sacrifice algorithmic conveniences. They also risk romanticizing marginalization, making a virtue of illegibility without acknowledging its costs to those who can not afford visibility.
So illegibility cannot be the whole answer. We also need friction.
Friction is different from illegibility. Illegibility is about being hard to categorise. Friction is about making categorisation itself slower, harder, and more accountable. It insists that algorithmic sorting can; t happen at computational speed, that it needs human judgment, that it needs to justify itself at each step. The GDPR creates certain frictions. The “right to explanation” means companies can’t just sort you; they need to explain why. The “right to deletion” creates friction in prediction systems. These are not perfect, often weakly enforced, easily circumvented, but they establish that algorithmic sorting should face resistance.
What other frictions might we create? Temporal friction, requiring waiting periods for algorithmic decisions affecting life chances, creates time for appeal and review. Collective audit, communities have the right to examine and challenge algorithms that sort them. Platform liability, making platforms legally responsible for harms created by their sorting systems. Public alternatives, building non-profit, collectively governed platforms meeting the same needs as commercial platforms but with different purposes. Wikipedia is one model. What would Facebook look like, structured like Wikipedia?
These structural changes require political power: power to regulate platforms, redistribute resources and build public alternatives. Which brings us back to the problem: how do we build that power when infrastructure for building power runs through the platforms we are trying to change?
There is no easy answer. But there are examples: labour organising in Amazon warehouses, gig worker strikes coordinated through the apps controlling their labour, and content moderator unions forming at the platforms employing them. These movements don’t escape algorithmic power; they organise within it, against it, using it where they can and refusing it where they must. Resistance to algorithmic purpose assignment looks less like individual escape and more like collective struggle over infrastructure. Less like deleting your apps and more like fighting for public control over digital infrastructure. Less like opting out and more like fighting to opt everyone in to systems designed for human flourishing rather than capital accumulation.
Credits: TCA, LLC.