Abstract

The concept of ‘distributed cognition’ is routinely invoked to unify heterogeneous collective epistemic systems, including prediction markets, open-source software development, deliberative bodies, digital platforms, and regulatory institutions. These systems are often treated as interchangeable instances of ‘crowd wisdom’, whose epistemic virtues are presumed to arise naturally from decentralisation and aggregation. This article argues that this assumption rests on a category error: it conflates epistemic coordination architectures with epistemic closure architectures and treats descriptive claims about cognitive distribution as if they entailed claims about epistemic reliability. Distributed cognition is not a unitary epistemic good but a family of structurally distinct socio-technical systems that perform different epistemic functions under different closure mechanisms, incentive structures, and governance-mediated conditions. The article develops a formal taxonomy of distributed epistemic systems and shows that each class distributes a different epistemic resource, stabilises closure in a different locus, and exhibits characteristic failure modes. Through a paradigm contrast between prediction markets and open-source communities, the article demonstrates that epistemic failures in distributed systems are governance-mediated distortions of closure and salience rather than merely cognitive accidents. Distributed cognition is therefore best understood as an engineered epistemic infrastructure whose reliability depends on architectural design rather than on decentralisation alone.