Bounded Rationality & Satisficing

You are standing in a supermarket aisle in front of forty jars of pasta sauce. The “rational” thing, the one your economics textbook quietly assumes, is to weigh every jar against every other on price, taste, sugar, brand, and the memory of last Tuesday’s dinner — build the full ranking, and lift the single best jar into your basket. Nobody alive has ever done this. What you actually do is scan a handful, spot one that looks fine, and drop it in. You didn’t find the best sauce in the aisle. You found one that was good enough, and then you stopped. And here is the uncomfortable part: that lazy- looking move was almost certainly the smart one.

This is the shape of nearly every real decision. The textbook picture — a rational optimiser who knows all the options, computes the payoff of each, and maximises expected utility — is a fiction. Not because people are stupid, but because optimising is often impossible in principle: the options are too many to list, the future too uncertain to score, and the computation too long to finish before the decision is due. A perfect chess player would need to search more board positions than there are atoms in the universe. A perfect career-chooser would have to interview every job that will ever exist. Real agents have bounded time, information, attention, and computing power — so they cannot optimise even if they wanted to.

Herbert Simon, who won the 1978 Nobel Prize in Economics for exactly this idea, gave the alternative a name: satisficing — a blend of satisfy and suffice. Instead of hunting for the unknowable best, you set an aspiration level — a bar for “good enough” — search until an option clears it, and then stop and take it. Satisficing is not a failure to be rational; it is what rationality actually looks like once you count the cost of thinking. Simon’s deepest image is the mind as a pair of scissors: one blade is our cognitive limits, the other is the structure of the environment. You can’t understand a cut by studying one blade alone.

This course builds the whole model, rung by rung. It opens with why optimisation fails — combinatorial explosion, unlistable option sets, and the sheer cost of computation, so you feel why the optimiser is a myth. It then defines satisficing precisely: the aspiration level, the search-till-good-enough rule, and how it is really a reservation- value stopping rule — the same machinery as the optimal-stopping course, with the aspiration playing the role of the bar. It turns to the psychology — Barry Schwartz’s paradox of choice, and why maximisers get marginally better outcomes yet end up less happy, drowning in regret and counterfactuals, while satisficers decide faster and stay content. It builds up fast-and-frugal heuristics and Gerd Gigerenzer’s ecological rationality — simple rules like take-the-best, the recognition heuristic, and 1/N that exploit the structure of the world and often match or beat complex optimisation out of sample, because they don’t overfit. And it closes on transfer — how to pick and adjust an aspiration level, and how to tell which decisions deserve the full optimisation and which are better satisficed.

Because the model can lie to you if you take it lazily. Satisficing is not an excuse for sloppy thinking: it is optimal subject to real constraints, and the aspiration level still has to be chosen well — set it too low and you settle for junk, too high and you never stop. It does not say optimisation is always wrong: for small, well-defined, high-stakes, repeatable problems, you should optimise. “Good enough” can quietly entrench mediocrity and hide the fact that a problem deserved the full search. And a heuristic that is brilliantly adapted to one environment can misfire badly when the world shifts — a fast-and-frugal rule is only ever as good as its fit to reality. Hold those edges, and what you keep is a genuine thinking tool: decide first whether a choice is optimise-worthy or satisfice-worthy, set an explicit bar, and commit to stopping when it’s met — good enough, fast, on the many; deep optimisation reserved for the few that earn it.