Cover of The Logic of Failure

The Logic of Failure

Dietrich Dörner

5 ideas

  1. Humans cannot intuitively extrapolate delayed dynamics

    People project change linearly and react to the current state of a system rather than to its trajectory, so they systematically underestimate exponential growth and misjudge lagged responses. In the refrigerated-storeroom thermostat task, subjects over-adjusted the dial because the temperature responded slowly, producing wild oscillations. Without an explicit model of time delays, corrective action arrives too late and too strong.

  2. Ballistic action versus monitored action

    Poor performers fire off decisions like cannonballs and never check where they land, treating an order as equivalent to a result. Good performers asked 'why' more often, tested hypotheses, and tracked the effects of each measure before acting again. Failing to gather feedback lets errors compound unseen, and it protects the actor's sense of competence from disconfirming evidence.

  3. Repair service behavior: fixing visible problems

    Under complexity, people fix whichever problems are currently salient or already familiar to them, rather than the ones that are structurally important. The resulting agenda is set by what happens to catch attention, so latent problems grow unnoticed until they become crises.

  4. Failure protects the feeling of competence

    Many characteristic errors serve to keep up the actor's sense of control rather than to solve the problem. Examples include hasty action, avoiding information that might disconfirm, dogmatic overgeneralized hypotheses, and retreating into small solvable subtasks. When things go badly, subjects often turn cynical, blame the system, or escalate into 'horizontal flight', where they narrow to a comfortable niche, or 'vertical flight', where they escape into abstract planning.

  5. Steps for acting in complex systems

    Good performance follows a cycle: formulate concrete, interim goals and resolve the contradictions between them; build a model of how the variables interconnect; gather information and forecast side-effects and long-term effects; plan and act; then review outcomes and revise both the model and the strategy. Complexity (many interconnected variables), intransparency (hidden states), internal dynamics (the system changes without you) and incomplete understanding each demand one of these steps. Skipping any step converts that property into a source of failure.

Save and mark ideas in the app