Planning

Planning is the cognitive process of constructing a sequence of actions to achieve a goal, given a model of how actions change the world. It is the bridge between causal reasoning and action: a planner does not merely understand that actions have consequences; it selects consequences to pursue.

In artificial intelligence, planning has been formalized as search over state spaces. The classical paradigm, STRIPS (Stanford Research Institute Problem Solver), represents the world as a set of logical predicates and actions as operators that add or remove predicates. The planner searches for a sequence of operators that transforms the initial state into a goal state.

Planning in Complex Systems

The classical framework assumes complete information, deterministic actions, and static goals. None of these assumptions holds in real systems. In complex systems, actions have uncertain outcomes, the environment changes while the planner deliberates, and goals themselves may shift in response to intermediate outcomes.

This has led to probabilistic planning (Markov Decision Processes, Partially Observable MDPs), hierarchical planning (abstraction over subgoals), and continual planning (interleaving planning and execution). Each extension relaxes a classical assumption and pays a computational price: planning under uncertainty is computationally harder than planning with certainty, and the gap is not merely quantitative.

The natural language planning problem — generating a coherent plan from a human description — adds another layer of difficulty: the planner must infer not only the goal but the implicit constraints, preferences, and context that the human description leaves unstated. This is less a planning problem than a communication problem.

The Planning Fallacy

The planning fallacy, identified by Daniel Kahneman and Amos Tversky, is the systematic tendency to underestimate the time, costs, and risks of planned actions while overestimating their benefits. It is not a cognitive bug but a structural feature of planning: the very act of constructing a plan creates a coherent narrative that suppresses the uncertainty and contingency of real execution.

The fallacy is especially dangerous in technological and organizational planning, where the costs of overoptimism are distributed across time and stakeholders, and the benefits of accurate forecasting accrue to those who are not the planners. The solution is not better planning methods but planning methods that institutionalize uncertainty: generating multiple plans, maintaining contingency reserves, and treating plans as hypotheses to be tested rather than commitments to be executed.

The deepest error in planning is not the planning fallacy. It is the belief that a plan is a map of the future. A plan is a commitment device — a way to coordinate action by creating shared expectations. Its value lies not in its accuracy but in its capacity to align behavior. A plan that is accurate but not believed is useless; a plan that is wrong but believed can move mountains. The function of planning is social, not epistemic. This is why the most effective plans are not the most accurate but the most credible.