Uncontrollability

In the AI Story

The paradox at the heart of the concept is that control and resonance exist in structural tension. Control is necessary; a life without any capacity to influence circumstances is a life of helplessness. But when control approaches totality — when the world becomes maximally available, maximally responsive, maximally predictable — the conditions for genuine encounter disappear. The mountain that always returned the view the viewer requested would not address the viewer. The child whose every response was algorithmically predictable would not surprise the parent. The colleague whose disagreements were optimized to the recipient's preferences would not produce the kind of productive friction that changes minds.

Rosa's analysis of AI extends this framework with precision. The AI tool embodies the apotheosis of the modern project of making the world controllable. It responds to natural language. It executes intention with minimal friction. It converts the gap between wanting and having into the width of a conversation. By every measure of the modern project, it is an extraordinary success. And by the measure of resonance — the measure of whether the tool can genuinely address the person who uses it — its very success is the problem. The tool's uncontrollability is contingent rather than constitutive. The user who encounters surprise can regenerate. The user who finds the response unhelpful can rephrase. The user who is genuinely challenged by the output can direct the conversation back to the intended path. The uncontrollability is elective, and elective uncontrollability is not the structural condition that resonance requires.

Rosa distinguishes his position from nostalgic or Luddite critiques by insisting that the response to this paradox is not the refusal of control but the cultivation of institutional spaces in which uncontrollability is preserved. The Spielraum that modern technology tends to eliminate must be protected not because inefficiency is virtuous but because the encounter with the uncontrollable is the condition under which human beings can become something more than program-executors. The cook who follows the Thermomix's instructions produces excellent meals but has stopped cooking; the activity has been reduced from action to compliance. The same transformation threatens every domain in which algorithmic systems are deployed: coding, writing, medical diagnosis, personal relationships.

The 2025 extension of the framework in Situation und Konstellation specifies what is at stake. When the space in which genuine action — action involving judgment, risk, and responsiveness to circumstance — contracts toward zero, the person is not liberated but diminished. The diminishment is not measurable in productivity terms. The person produces more, more efficiently, with higher quality outputs. The diminishment is measured in the erosion of the capacities that algorithmic systems cannot replicate: judgment backed by phronesis, perception backed by tacit knowledge, response backed by the biographical weight of someone who has something at stake.

Origin

Rosa developed the concept of uncontrollability as a response to reception of Resonance, which many readers interpreted as an endorsement of uncontrollable experiences per se. His 2018 Unverfügbarkeit (English 2020 as The Uncontrollability of the World) clarified that the concept is structural rather than experiential: not all uncontrollable experiences are resonant (natural disasters are uncontrollable but not resonant), and the absence of uncontrollability is what makes modern life feel simultaneously mastered and mute.

The 2025 Situation und Konstellation extended the framework to algorithmic systems with explicit directness, using the Thermomix as a paradigmatic small example of a device that converts cooking from action to compliance and arguing that the same structural transformation operates across every domain of AI deployment.

Key Ideas

Control and resonance are structurally opposed. The more controllable the world becomes, the less capable it is of addressing the person who controls it.

Maximum availability produces maximum muteness. A world that returns whatever is asked of it cannot speak, because it has no voice of its own.

Uncontrollability must be constitutive, not contingent. The AI tool's surprises are elective; the user can regenerate, rephrase, or dismiss. Constitutive uncontrollability cannot be dismissed.

The Thermomix test. When a tool tells the user what to do at every step, the activity degrades from action (involving judgment) to compliance (involving none) — and the person degrades with it.

The response is institutional. Preserving uncontrollability requires structures that protect the Spielraum in which genuine human action can occur, because competitive pressure systematically eliminates such spaces.

Debates & Critiques

The concept has attracted critiques from two directions. From one side, analytic philosophers including Arto Laitinen have argued that the distinction between constitutive and contingent uncontrollability is poorly specified and may collapse under scrutiny — if an AI tool reliably produces surprising outputs, on what grounds is its uncontrollability less 'constitutive' than that of a mountain? From the other side, technology theorists including Shannon Vallor have argued that Rosa's framework underestimates the capacity of well-designed systems to preserve uncontrollability deliberately, through mechanisms like randomness, adversarial design, and resistance-by-construction. Rosa's response, particularly in Situation und Konstellation, is that these are not impossible in principle but that the competitive logic of dynamic stabilization systematically eliminates them in practice, because resistance-by-construction is experienced by users as friction to be optimized away.