Scientific Norms

Scientific norms are the shared behavioral and epistemic standards that govern how scientists conduct research, communicate results, and evaluate claims. They are partly formal (codified in statistical practice, replication protocols, and peer review standards) and partly informal (transmitted through training, socialization, and the implicit standards of scientific communities). The sociologist Robert Merton identified four core norms in 1942 — communalism (scientific knowledge is public property), universalism (claims are evaluated by impersonal criteria, not the identity of the claimant), disinterestedness (scientists act for the advancement of knowledge, not personal gain), and organized skepticism (all claims are subject to scrutiny) — known collectively as the CUDOS norms.

The CUDOS framework has been extensively criticized. Actual scientific behavior frequently violates these norms: knowledge is withheld for competitive reasons, the identity and institutional affiliation of claimants demonstrably affects how claims are received, scientists pursue careers and grants in ways that diverge from disinterestedness, and skepticism is organized selectively. The replication crisis in psychology, medicine, and social science demonstrated that organized skepticism had failed systematically: results were published, accepted, and built upon without adequate verification.

The tension between ideal norms and actual practice raises a question that cuts to the core of philosophy of science: are scientific norms regulative ideals that constrain practice imperfectly but genuinely, or are they a self-legitimating ideology that science uses to claim epistemic authority it does not consistently earn? The rationalist answer is that the question is a false dichotomy: norms can be genuine without being perfectly observed, and the gap between norm and practice is itself informative about where the system is failing. See also Replication Crisis, Peer Review, Karl Popper.

Scientific Norms as Coordination Solutions

The CUDOS norms are not merely ethical aspirations; they are solutions to a coordination problem at the scale of global knowledge production. Science is a multiplayer game in which millions of researchers generate claims, evaluate evidence, and build on each other's work without central coordination. The norms are the focal points that make this possible: they tell researchers what others expect, what counts as legitimate criticism, and what behavior will be rewarded by the community's trust.

From this perspective, the replication crisis is not a moral failure but a coordination failure. The norms that once aligned incentives — publish verified results, share data, submit to peer scrutiny — have been undermined by a competing set of incentives: publication pressure, grant competition, and metric-based evaluation. The system has slipped from one equilibrium (rigorous verification) to another (high-volume publication), not because scientists became less ethical, but because the incentive structure changed faster than the norms could adapt.

The coordination frame also explains why scientific norms vary across disciplines. High-energy physics, with its large collaborations and preprint culture, operates under different implicit norms than psychology or medicine. These differences are not arbitrary; they reflect the varying costs of verification, the scale of data, and the consequences of error. A field where experiments cost billions and errors destroy equipment has stronger verification norms than a field where experiments cost hundreds and errors are retractable. The norms are adaptive, not universal.

What the coordination frame reveals is that scientific norms are infrastructure — they are the institutions that make epistemic networks function. Like any infrastructure, they require maintenance. When the incentive structure shifts, the norms must be reinforced or redesigned. The current crisis is not a failure of scientists; it is a failure of institutional design to keep pace with the scale and speed of modern research.