Satisficing

Satisficing is the decision strategy of choosing an option that meets an aspiration level — a threshold of acceptability — rather than continuing to search for the optimal choice. The term was coined by Herbert A. Simon in 1956 as a portmanteau of "satisfy" and "suffice," and it represents the central mechanism of bounded rationality: in a world of limited information, limited time, and limited cognitive resources, the rational agent does not optimize. They stop searching when they find something good enough.

The concept is not merely descriptive. It is normative in a way that challenges the entire edifice of classical decision theory. Where expected utility theory prescribes that agents maximize over all possible options weighted by their probabilities, satisficing recognizes that the cost of search itself is a cost that must be weighed against the marginal benefit of finding a better option. A manager who spends six months searching for the perfect candidate while the position remains unfilled is not being more rational than one who hires the third qualified applicant. They are being less rational, because they have failed to account for the cost of continued search.

The Architecture of Satisficing

Satisficing operates through three components: an aspiration level, a search rule, and a stopping rule. The aspiration level defines what counts as "good enough." The search rule determines the order in which alternatives are examined. The stopping rule says: stop when an alternative exceeds the aspiration level. These three rules together replace the single optimization rule of classical theory with a procedure that is computationally tractable, dynamically adaptive, and ecologically realistic.

The aspiration level is not fixed. It is adjusted based on experience of what is available in the environment. When search repeatedly fails to find alternatives that meet the aspiration level, the level is lowered. When search succeeds easily, it is raised. This dynamic adjustment makes satisficing adaptive: it converges on reasonable outcomes without requiring foreknowledge of the distribution of alternatives. The aspiration level becomes, in effect, a learned estimate of what the environment can deliver.

Satisficing in Organizations

In organizational theory, satisficing is the dominant mode of organizational decision-making. Organizations do not maximize profits; they satisfice on profits while attending to multiple, often conflicting goals: survival, legitimacy, market share, employee morale, regulatory compliance. The behavioral theory of the firm — developed by Cyert and March as an explicit extension of Simon's framework — treats the firm as a coalition of participants with heterogeneous goals, each pushing their own aspiration level, and organizational decisions as the negotiated outcome of these pushes.

This has profound implications for how we understand market behavior. The neoclassical firm is a single maximizer with a well-defined objective function. The behavioral firm is a distributed satisficing system with multiple, shifting objectives. Markets populated by behavioral firms do not converge to the efficient allocations predicted by general equilibrium theory. They converge to patterns of behavior that are "good enough" for the firms involved — patterns that may be stable, may be inefficient, and may be deeply path-dependent.

The Machine Parallel

Satisficing is not merely a human limitation. It is a general principle of intelligent systems operating under resource constraints. In artificial intelligence, any search algorithm that uses a stopping criterion before exhausting the search space is satisficing. Beam search, greedy algorithms, and early stopping in neural network training are all forms of computational satisficing. The question is not whether to satisfice — any real system must — but whether the aspiration level is well-calibrated to the environment.

The danger in machine systems is that their aspiration levels are often set by designers who do not understand the environments in which the systems will operate. A recommendation algorithm optimized for click-through rate is satisficing on engagement, but its aspiration level has been set by a business metric, not by a user-welfare metric. The result is a form of organizational satisficing that serves the platform rather than the user — a misalignment of aspiration levels that the standard framework does not capture because it treats the firm as a unitary maximizer rather than as a coalition of interests.

Satisficing and Ecological Rationality

The ecological rationality of satisficing depends on the structure of the search space. In environments where good alternatives are abundant and the cost of search is high, satisficing is not merely acceptable — it is optimal. In environments where good alternatives are rare and the cost of a suboptimal choice is catastrophic, satisficing is dangerous. The rationality of the strategy is not intrinsic; it is relational, a function of the fit between the decision procedure and the environmental structure.

This is the deeper point Simon was making: rationality is not a property of the agent alone. It is a property of the agent-environment system. A decision procedure that is rational in one environment may be irrational in another. The normative question is not "should agents satisfice or optimize?" but "what aspiration level, search rule, and stopping rule are appropriate for this environment, given what the agent knows and what the agent can afford to find out?"