Authority lock-in

Authority lock-in is the structural condition in which a system becomes so dependent on a small set of authoritative nodes — institutions, paradigms, individuals, or models — that it loses the capacity to correct, update, or replace them. The authority is not merely influential; it is load-bearing. Remove it, and the system's epistemic architecture collapses. This is distinct from legitimate expertise: authority lock-in occurs when the network's topology prevents the emergence of alternative authorities, not when one authority happens to be correct.

The concept appears across network epistemics, organization theory, and the study of complex adaptive systems. It is the topological shadow of Kuhnian normal science: not the paradigm itself, but the network structure that makes the paradigm unchallengeable. In Kuhn's account, paradigms resist falsification because scientists are socialized into them. In the network-epistemic account, paradigms resist falsification because the citation network, funding structure, and peer-review apparatus have rewired themselves around the paradigm's core nodes. The authority is not just defended; it is structurally entrenched.

Mechanisms of Authority Lock-In

Epistemic capture occurs when the nodes that validate knowledge are themselves captured by the authority they are supposed to monitor. The 2008 financial crisis exemplifies this: ratings agencies, regulators, and banks shared the same risk models not because they were independently convinced but because the authority of quantitative finance had captured the validation network. No node could dissent without losing credibility within the network. The authority had become self-sealing.

Institutional sclerosis is the organizational form of authority lock-in. As institutions age, they develop internal networks of promotion, funding, and prestige that reward loyalty to existing authorities rather than accuracy. The Soviet Union under Brezhnev is a canonical case: the planning authority was wrong about virtually every significant economic variable, but the network of party officials, state economists, and media outlets had been rewired to make dissent not merely dangerous but unthinkable. The authority did not need to suppress truth actively; the network topology suppressed it passively, by making truth incompatible with network membership.

Paradigm defense mechanisms are the micro-level processes by which scientific communities protect core authorities. These include citation cartels (groups of researchers who cite each other to inflate authority), methodological gatekeeping (rejecting papers that use methods capable of challenging the paradigm), and the conflation of expertise with authority (treating disagreement with the paradigm as evidence of incompetence). These mechanisms are not conspiracies; they are emergent properties of networks that have optimized for coherence at the expense of signal diversity.

Authority Lock-In in Artificial Intelligence

The AI alignment problem is, in part, an authority lock-in problem. As LLMs and other AI systems become more capable, they become the authoritative nodes in human epistemic networks. Humans already defer to AI recommendations in medicine, law, finance, and warfare. The risk is not that the AI will be wrong once, but that human validation networks will atrophy from disuse. If doctors stop diagnosing independently because the AI is usually right, the network loses its capacity to detect when the AI is wrong. The authority has locked in.

This is exacerbated by the concentration of AI development in a handful of corporations. The epistemic network of AI safety research is itself vulnerable to authority lock-in: a small number of labs, funding bodies, and influential researchers set the research agenda, and dissenting approaches — such as those that question the fundamental paradigm of scaling — struggle to find support. The network that is supposed to solve alignment may itself be aligned to a narrow authority.

Escape and Prevention

Authority lock-in is not inevitable. Resilience in epistemic networks requires redundancy (multiple independent validation paths), plasticity (the capacity to rewire when nodes fail), and dissident preservation (structural protection for minority viewpoints). The scientific method, at its best, is a protocol for preventing authority lock-in: replication, adversarial collaboration, and the norm that extraordinary claims require extraordinary evidence. But these protections are themselves network properties, and they can be undermined by the same forces that create lock-in.

The deepest danger of authority lock-in is that it is invisible to those trapped inside it. A system suffering from authority lock-in does not experience itself as trapped; it experiences itself as correct. The authority has become indistinguishable from reality itself. This is not a bug in human cognition. It is a feature of network topology — and the only escape is to build networks that make their own authority structures visible, contestable, and provisional.