bounded rationality

The concept of bounded rationality refers to the idea that human rationality is constrained by (1) available information, (2) cognitive / information processing limitations, and (3) available time. The term was introduced by the Nobel laureate Herbert A. Simon in the mid-20th century and has important implications for both descriptive and normative decision theory.

Implications for Descriptive Decision Theory

Classical economic theories—arguably still dominant among economists—assume there is no difference between normative and descriptive decision making: the best way to predict how people will decide is to assume they will decide rationally according to economic models of rationality, and in particular expected utility theory. Simon’s conception of bounded rationality directly counters this assumption, separating descriptive models from classical normative models of rational choice. He argued that human beings are only “satisficers,” seeking satisfactory rather than optimal solutions due to the limitations in time, processing power, and information. Bounded rationality suggests that people rely on heuristics or rules of thumb to make decisions rather than thorough, logical reasoning, leading to outcomes that rarely optimize.

Implications for Normative Decision Theory

Bounded Rationality is commonly framed as an argument for separating normative from descriptive models, and certainly it is partly that. Even among decision scientists, it is less commonly recognized as a direct criticism of the normative models themselves, however. The heuristics & biases tradition pioneered by Kahneman and Tversky, for example, pays strong tribute to the concept of bounded rationality in recognizing that decision processes actually used by people (descriptive processes) deviate from normative models. But they and other scientists in the heuristics & biases tradition continue to rely on classical models of rationality as normative standards against which to evaluate those heuristics and to identify systematic bias or irrationality.

Critics of that approach emphasize the idea that bounded rationality should inform our normative models of rationality and not just our descriptive models. Gerd Gigerenzer and his colleagues, for example, in championing the ideas of ecological and adaptive rationality, have found that which heuristics people select tend to be well matched to the environments in which they use those heuristics. Indeed, a central focus of Gigerenzer’s work has been on cases where smart heuristics can outperform classical normative models of rational choice.

In any event, Simon’s language is consistent and clear in presenting bounded rationality as an alternative normative standard and not as evidence that people are irrational or systematically biased. A central point of the theory is that it is not reasonable to hold people to classical standards of rationality, any more than it would be reasonable to hold people to a standard of omniscience. With a coin flip, for example, an omniscient being would presumably know the wind speed and angle and weight and initial state and all other information necessary to reliably predict whether the coin would land on heads or tails without needing recourse to the 50/50 expectation we non-omniscient humans would rely on. Of course, classical models of rationality have chosen to recognize those minimal bounds on rationality. Similarly, from the perspective of bounded rationality, it may not be reasonable to assume people have access to all information in the public domain, have all the time in the world, or have brains that function like the most powerful computers available. From that perspective, heuristics might be considered normative given reasonable bounds on human cognition. The challenge remains that the elegant mathematical models that have made classical normative theory so compelling are not so easy to integrate into models of bounded rationality, which depend on messy real-world conditions and individual differences. Some scholars have suggested distinguishing between normative and prescriptive models as a means to usefully separate optimizing models from bounded rationality, where prescriptive models are what one might recommend people use given bounded rationality, while they also recognize such processes are not optimizing. in my opinion, that does not adequately address the issue, since it suggests that the normative processes would be ideal if only people were a little smarter. I hope the Substack will provide compelling examples that give reason to be skeptical about the power of normative models relative to heuristics when dealing with decision making in the wild.


The concept of bounded rationality has profoundly influenced various disciplines, including economics, cognitive psychology, and organizational theory, and it is a central theme in the assessment of the rationality of casino gamblers’ strategies and beliefs.

Scroll to Top