# The Unaccountability Machine **Dan Davies** | [[Prediction]]  --- > "The purpose of a system is what it does." This single sentence—Stafford Beer's axiom—demolishes most corporate mission statements. Organisations don't do what they claim to do. They do what their structure, incentives, and information flows make inevitable. The "unaccountability machine" isn't a moral failure; it's a structural property of complex systems that can't be fully controlled from the top. **You can't manage what you don't understand, and you can't understand complex systems by looking inside them.** Beer's cybernetic approach treats organisations as black boxes—what matters isn't the internal wiring but the observable behaviour. If your system produces terrible outcomes despite good intentions, the problem isn't bad people. It's bad design. Davies uses Beer's [[Viable System Model]] to show why organisations fail: they lose the capacity to process the variety (complexity) thrown at them by their environment. When information channels narrow—through outsourcing, rigid hierarchies, or metrics that ignore critical signals—the system becomes fragile. It can't adapt. And fragility, in cybernetic terms, means death. --- ## Core Frameworks ### [[POSIWID]] – The Purpose of a System is What It Does This is Beer's most subversive insight. Ignore mission statements, values posters, and executive speeches. **Judge systems by their actual behaviour**, not their stated intentions. If your hiring system consistently rejects diverse candidates, its purpose is homogeneity—regardless of what HR says. If your quality process creates mountains of paperwork but doesn't catch defects, its purpose is bureaucracy, not quality. If your strategy produces short-term profits and long-term decline, its purpose is extraction, not sustainability. POSIWID strips away comforting self-deception. It forces you to confront what your organisation *actually optimises for*, which is often very different from what you claim. ### [[Ashby's Law of Requisite Variety]] **"Only variety can absorb variety."** A system can only remain viable if its internal complexity matches the complexity of its environment. If the world throws 1,000 different problems at you, and your organisation can only produce 10 responses, you will fail. Organisations lose variety through centralisation (pushing decisions up reduces response options), outsourcing (contracts flatten information to metrics, losing nuance), standardisation (processes that work "on average" fail at extremes), and rigid hierarchies (each layer filters information, reducing variety). The solution isn't more control. It's **distributed decision-making** where frontline workers have the autonomy to handle local variety without escalating everything. ### [[The Viable System Model (VSM)]] Beer's model describes the five functions every viable system must have: **1. Operations** – The doing: frontline teams executing work **2. Coordination** – Ensuring operations work together without conflict **3. Control** – Optimising operations, resolving conflicts, maintaining stability **4. Intelligence** – Scanning the environment for threats and opportunities ("there and then") **5. Policy/Identity** – Balancing future vision with present operations Common failure modes: missing intelligence function (organisation becomes internally focused, blind to external change like Kodak missing digital photography), overactive control with weak coordination (micromanagement creates bottlenecks), short-term policy dominance (quarterly earnings obsession kills long-term intelligence). Viable systems need all five in balance. Most organisations are lopsided. ### [[Algedonic Signals]] "Algedonic" = pain/pleasure. These are urgent signals that bypass normal channels when something is critically wrong (or right). Examples: whistleblower hotlines (if they work), direct escalation paths for critical failures, customer complaints that reach leadership immediately. Most organisations suppress algedonic signals. Middle management filters bad news. Hierarchies slow urgent information. By the time leadership sees the problem, it's too late. **Cybernetic design principle:** Create channels for pain signals to reach decision-makers *fast*, without filtering. **"Every decision about what to measure is implicitly a decision about what not to measure."** Metrics narrow attention—what you don't measure becomes invisible. Information vs. data: only inputs that affect decisions count as information. Outsourcing degrades information flow because contracts reduce complexity to KPIs, losing critical context. **Viability requires accountability across functions, not just top-down control.** Decisions must be justifiable to those affected (reciprocal responsibility). Lack of perceived control damages health (Marmot's research)—inequality creates a variety mismatch where individuals can't regulate their environment. **The "Ricardian Vice"**: economists prove things in models and mistake them for reality. Financial accounts are control models but can distort reality (prioritising cost cuts over revenue growth). Corporations seek **stability, not maximisation** (echoing Herbert Simon's satisficing). Galbraith's point: corporations *plan* demand through advertising—they're not passive market participants. **Complex systems are black boxes**—you can't fully understand them from inside. Overconfidence comes from knowing details of a subsystem whilst missing broader causal dynamics. Management cybernetics sees organisations as decision-making systems, not machines to be programmed. --- ## Connects To - [[The Fifth Discipline]] – Both use systems thinking; VSM is operational, Senge's work is conceptual; both emphasise feedback loops and delays - [[Dead Companies Walking]] – Failing companies lose requisite variety—they can't adapt to environmental change - [[How Finance Works]] – Financial accounts as control models; Davies critiques overreliance on backward-looking metrics - [[Playing to Win]] – Strategy requires intelligence function (scanning environment) balanced with policy (long-term vision) - [[Nine Lies About Work]] – POSIWID challenges stated values; judge systems by behaviour, not rhetoric