Bounded Rationality: A Short History

For a long stretch of the twentieth century the formal model of a decision-maker was a creature that knew its options, knew what each was worth, and picked the best. The model was never meant as a description of anybody. It was a normative standard. The trouble began when it started being used as though it were a description.

Satisficing

The first sustained alternative came from work in administrative decision-making, which pointed out that real decisions are made by agents with limited time, limited information and limited capacity to compute. Such an agent does not survey the option space. It searches it in some order and stops when it finds something good enough against a standard it carries into the search. The word coined for that stopping rule was satisficing, and the general position became bounded rationality.

The important move here is that the bounds are not merely a defect. A stopping rule is what makes deciding possible at all in an environment where exhaustive search would never terminate.

The heuristics tradition

The second wave came from experiments showing that judgements under uncertainty follow a small number of shortcuts, and that those shortcuts produce systematic rather than random departures from the formal standard. Judging likelihood by how easily examples come to mind, or judging membership by resemblance to a type, gets a great deal right cheaply and gets particular things reliably wrong.

Systematic is the operative word. Random error averages out and tells you little. Error that leans consistently in one direction is a signature of the process that produced it, which is why the biases were interesting rather than merely embarrassing.

The standing dispute

A long-running argument follows from this. One reading treats the shortcuts as flaws measured against the formal standard. Another treats them as well-matched to the environments in which they evolved and in which they are normally used, and treats the laboratory tasks that expose them as unrepresentative. The dispute is unresolved, and both sides agree on the underlying phenomena while disagreeing about what counts as the right yardstick.

The recurring misquotation

The version that circulates most widely, that people are irrational, is the one the literature least supports. Bounded rationality is a claim about the fit between a reasoning procedure and the conditions it runs under. It is not a verdict on the reasoner.

Permalink: /notes/bounded-rationality-a-short-history