Mechanism Realism

The Causal Talisman: When a Cause-Name Replaces an Explanation

What if a causal claim names something real — but still does no explanatory work? This episode of Mechanism Realism introduces the causal talisman: a morally protected cause-name used to discharge explanation-pressure without producing contribution accounting. The named cause may be real. Premature birth, poverty, demographic change, global technological shifts, trauma, austerity, screen time, or neoliberalism can all matter. The pathology is not that the cause is false. The pathology is that the cause-name is deployed as if it carried the explanatory weight of a bounded analysis. A causal talisman has a specific shape. It is plausible at the local level, distant from the accountable institution, morally difficult to challenge, and left undecomposed at the aggregate level. It shifts attention away from the question that matters: what share of the phenomenon does this cause explain, against which rival causes, with what evidence, changing which decision, and under what conditions would confidence fall? The episode walks through the diagnostic kit: share, rivals, discriminator, decision, and defeater. A cause becomes talismanic when it cannot answer those questions but is still offered with the rhetorical weight of explanation. This is not anti-causality. It is stricter causality. The cure is not to stop naming causes. The cure is to stop treating morally charged cause-names as substitutes for mechanism analysis. Once you see the shape, you see it everywhere: across political coalitions, institutional defenses, media narratives, expert discourse, and everyday arguments where the need for an explanation arrives faster than the willingness to compute one. https://kunnas.com/articles/causal-talisman [https://kunnas.com/articles/causal-talisman]

The Stack: Where Goal-Directed Systems Fail

When a system fails, where exactly did it fail? This episode of Mechanism Realism introduces The Stack: a twelve-layer map for diagnosing goal-directed systems. Institutions, policies, programs, markets, cultures, organisms, and AI systems do not fail in one generic way. They fail at different layers. Sometimes the purpose is unowned. Sometimes the mechanism was never real. Sometimes actors respond through the cheapest available channel. Sometimes the harmed party has no carrier. Sometimes the metric replaces reality. Sometimes hidden capital stocks are depleted. Sometimes the public decision frame was never compiled. Sometimes the frame exists but nobody computes with it. Sometimes diagnosis exists but no one is forced to decide. Sometimes implementation capacity is missing. Sometimes feedback produces reports but not correction. Sometimes the output survives while the generator-chain dies. The Stack is not a claim that reality has exactly twelve layers. It is a diagnostic surface: name the system, name the reference telos, then walk the layers until the binding failure appears. The episode gives a map for moving from vague criticism — “the system is broken” — to precise diagnosis: broken where, relative to what purpose, through which missing primitive, and with what repair path? ⁠https://kunnas.com/articles/the-stack⁠ [https://kunnas.com/articles/the-stack]

Stand Alone Complex: Convergence Without Coordination

Why do institutions sometimes converge on the same behavior when no one is coordinating them? This episode of Mechanism Realism examines Stand Alone Complex: coordinated-looking convergence without coordination. The pattern looks organized, but there is no organizer. It looks like conspiracy, but no command structure appears. It looks like imitation, but there is no traceable copying chain. It looks like independent discovery, but reality did not force the same conclusion on everyone. The missing mechanism is the reward gradient. Funding, status, career advancement, legitimacy, audience attention, peer approval, and regulatory safety can select similar behavior from independent actors. Each actor responds locally. The aggregate becomes structurally coherent because the gradient is coherent. The episode separates Stand Alone Complex from conspiracy, distributed coordination, mimetic copying, fashion, and genuine independent discovery. The key test is ex ante: can the reward gradient be identified before the convergence is observed? If the gradient is only inferred after the pattern appears, the diagnosis collapses into storytelling. The intervention logic follows from the diagnosis. Refuting the pattern does little if adoption was never driven by belief in its truth. Replacing individual actors does little if the gradient selects the same behavior from their replacements. The real levers are gradient change, carrier disruption, counter-gradient creation, and architectural redesign. https://kunnas.com/articles/stand-alone-complex [https://kunnas.com/articles/stand-alone-complex]

Causal Scope Laundering: When Evidence Closes the Wrong Question

What if the evidence is real, the citation is accurate — and the argument is still invalid? This episode of Mechanism Realism examines causal scope laundering: the public-discourse failure where a bounded study is used to settle a larger policy question that the study was never designed to answer. A treatment-effect study can tell us what happened to a defined population, under a defined intervention, against a defined counterfactual, over a defined window. That is valuable evidence. But policy arguments often smuggle that narrow estimate into a much larger claim: whether a whole accountability system, labor-market structure, school architecture, or incentive regime should exist. The episode walks through examples like body-worn camera trials, minimum-wage studies, charter-school lotteries, class-size experiments, and grade-retention research. In each case, the study may be rigorous and correctly summarized. The problem appears when the citation becomes a stopping rule for a composite policy question whose decisive mechanisms were not varied by the study. The repair is not anti-science. It is stricter evidence discipline. Before a citation closes a dispute, ask three questions: what exactly did the study estimate, what exact policy action is the citation being used to close, and which action-relevant mechanisms or contexts were outside the study’s scope? A study cannot close a question it did not identify. https://kunnas.com/articles/causal-scope-laundering [https://kunnas.com/articles/causal-scope-laundering]

The Mandate Trap: Why Problem-Solvers Become Problem Managers

Why do organizations created to solve problems so often become institutions for managing them? This episode of Mechanism Realism examines the mandate trap: the structural pattern where an organization’s mission and its effective telos diverge. The mission says: solve the problem. The telos says: preserve the mandate, funding, staff identity, status, and legitimacy that exist around the problem. This is usually not hypocrisy. The people may be sincere, the reports accurate, the campaigns useful, and the work genuinely necessary. The failure is architectural. Most organizations are authorized to act on a slice of a problem: document it, evaluate it, advise on it, serve its victims, or campaign around it. But when the real repair lies upstream of that mandate, every adjacent organization can truthfully say: not our job. The episode distinguishes downstream organs from the missing upstream organ. Ambulance services do not redesign roads. Shelters do not redesign housing markets. AI evaluation institutes do not automatically own deployment authority. These organs can be valuable and still not own lifecycle repair. The deeper missing function is mechanism lifecycle ownership: testing a mechanism before installation, monitoring it after deployment, detecting failure, triggering repair-or-explain, and forcing the political system to respond. The mandate trap is the condition where every organization can truthfully say “not our job” while the job remains undone. https://kunnas.com/articles/the-mandate-trap [https://kunnas.com/articles/the-mandate-trap]