Simon’s split, and five kinds of rationality

The notes ask whether policy decisions in the real world are rational, or whether they run up against limits on rationality. To look at that, they list a rational model, a satisficing model, an incremental model, and a garbage-can model, and they also name a cybernetics model, mixed scanning, an optimal model, and Allison’s multiple-perspective model. Cybernetics, mixed scanning, and the optimal model stay as names in that list; their inner workings are not invented here.

In policy research, they write, rationality often means maximizing a goal or optimizing a task. Simon splits that into substantive rationality, judged against a set goal, and procedural rationality, a product of the thinking process that puts the course of choice first and turns the question toward the capacity of the system that has to solve the problem. Reaching substantive rationality, the notes say, assumes an actor who already has a goal of utility or profit maximization, and who has the knowledge and ability to choose rationally, and so can pick a substantively rational alternative. People, Simon argues, drop that attempt to maximize and, pursuing procedural rationality, look for an alternative they can be satisfied with.

Diesing and Dunn, in a classification the unique body actually spells out, name five kinds. Technical rationality is about choosing an effective answer to a public problem: among alternatives, picking the means that best reaches the goal raises it. Economic rationality is about an efficient answer: it is higher when less cost yields a larger result. Legal rationality chooses by consistency with established law and precedent. Social rationality looks at an order of interdependence and solidarity among the parts of a society, and chooses by the capacity to promote institutionalization; that order of interdependence, they add, develops through mutual adjustment. Political rationality is the rationality of the decision structure itself, and appears when structural devices for better, sounder policy are in place.

Five general steps, and the rational-comprehensive model

A general procedure, as they lay it out, has five stages. First, define the problem to be solved through policy and, after checking the facts, estimate how far the present situation diverges from a desired one, in parallel with specifying the goal. Second, set a concrete goal that can be reached within the plan period and the available resources, and put it in operational, quantitative terms: the condition to be corrected or the state to be reached, the limited population in which that condition exists, the time needed, and the degree and direction of the intended change. Third, design alternative courses: develop and identify means that could reach the concrete aim. Fourth, predict the consequences of each course, both positive and negative. Fifth, choose the alternative the decision maker thinks most desirable for reaching the goal.

The rational, or rational-comprehensive, model is the process a decision maker is said to follow when seeking to maximize a goal or optimize a solution. At its core are two premises about the individual decision maker: an atomized actor who pursues self-interest, with interests and preferences already set before the decision situation; and an actor able to identify decision criteria and alternative courses clearly, and to rank them. The five steps that follow are: grasp the problem fully and define the goal clearly; search widely for alternatives; predict every result of those alternatives completely; have a clear standard for comparing them; and apply that standard to pick the best. Newer researchers, the notes say, still hold to the core assumptions of rational theory while recognizing uncertainty; the development of statistical decision theory and game theory helped put that uncertainty into the model as probability. As a descriptive, empirical theory, though, the rational model is criticized for not portraying real policy situations adequately. Incrementalists go further: because agreement among the parties matters, they say the model’s premises and procedures are not only impossible for human beings but, even as a prescription, not desirable.

Satisficing, incrementalism, a garbage can, and Allison on Cuba

Under bounded rationality the notes first place a satisficing model. Simon’s claim, they write, is that a policy maker stops at a satisfactory alternative rather than the one that maximizes the goal, because only a few alternatives are searched, at random and in sequence, and the search ends when one meets the test. Satisficing is the minimum standard a decision maker wants to apply; a person who satisfices is thought to have a sequential information-processing system. As a prescription the model is given a cautious review: it is hard to tell public officials to use it, and whether something satisfies depends on an aspiration level that is fluid. Its useful lesson, they say, is that human cognitive limits mean rational decisions need those limits to be supplemented.

The incremental model tries to explain actual policy making, and focuses on a setting with many decision makers. The alternative chosen is usually a small modification of existing policy; decisions are partial and sequential; goals and means are often chosen together, so a sharp goal-means analysis is a poor fit; and the test of a good alternative is agreement among those concerned. Comprehensive analysis is limited by intellectual capacity and information, and comparison proceeds by trial and error. The notes add that the model has relatively few precedents and a limited domain, and they record a critique: it tends to defend existing policy and to constrain the decision maker’s analytical effort.

The garbage-can model is offered to describe decisions in highly uncertain organizations. Organized anarchy, as they list it, has problematic preferences, a technology in which the link between goals and means is unclear, and fluid participation. The elements of a decision are problems, solutions, participants, and choice opportunities. Styles inside the can are named as problem solving, insertion, and putting off, or delay; problem solving, they say, is not the most common method, and insertion and delay are the most important features. The process is sensitive to the volume of work, and decision makers and problems both move from one choice opportunity to the next. Later fragments in the notes are left incomplete, and are not filled in here.

Allison, in a brief close, is said to have answered the Cuban Missile Crisis with three paradigms — a rational-actor model, an organizational-behavior model, and a governmental-politics model — and to have tried to explain the adoption of a naval blockade. That is as far as the unique body goes. The same Korean report, at fuller length, is on CampusDown.

Sources

1. Notes for a chapter on rationality and policy-decision models, asking whether real-world decisions are rational or bound by limits.
2. Simon’s split between substantive and procedural rationality; Diesing and Dunn’s technical, economic, legal, social, and political types; and a five-step general procedure, as the unique body spells them out.
3. The rational-comprehensive model and its critiques; satisficing and incrementalism; the garbage can in organized anarchy; and Allison’s three views of the Cuban Missile Crisis and a naval blockade, without added later detail.

Last updated: 14 November 2026