bounded rationality
In plain wordsThe idea that real decision-makers, whether people or machines, cannot consider every option or compute the perfect choice, so they make reasonable decisions within limits of time, information and thinking power.
The view that the rationality of real agents is limited by the information they have, the computation they can afford and the time available, so that they rely on heuristicsheuristic, n.: A rule of thumb: a quick, usually good-enough way to make a decision or to estimate how close you are to a goal, without any guarantee of finding the best answer., simplified models and stopping rules rather than optimizing. Herbert Simon introduced it against the all-knowing 'economic man' of classical economics, proposing satisficingsatisficing, n.: Choosing the first option that is good enough instead of searching for the very best one — a sensible strategy when time, information or brainpower are limited. as a realistic decision procedure; later work studied heuristics and biases (Daniel Kahneman and Amos Tversky) and ecological rationality, the success of simple heuristics in particular environments (Gerd Gigerenzer). In AI, bounded rationality motivates agents that decide well under resource limits: anytime algorithms and metareasoning that decide how much to compute, Stuart Russell's bounded optimality (the best program a given machine can run), and resource-rational models of human cognition. It also bears on AI safetyAI safety, n.: The effort to keep AI from doing serious harm, whether because it pursues aims its makers never intended, is turned to bad uses or simply malfunctions — a concern that grows as systems become more capable., since idealized expected-utility maximizers are a poor model of trained systems, and on embedded agencyembedded agency, n.: The puzzle of how an agent should reason when it is itself part of the world it is trying to understand and change, rather than standing outside it..
Under bounded rationality, the robot stops planning once further deliberation would cost more than it could gain.
Origin
Coined by Herbert A. Simon, who used the phrase in Models of Man (1957) after developing the idea in Administrative Behavior (1947) and 'A Behavioral Model of Rational Choice' (1955). His 1978 Nobel memorial prize in economics cited this research on decision-making in organizations.
Usage note
Bounded rationality is not irrationality: it describes doing as well as possible within real limits. Contrast the idealized agents of classical decision theorydecision theory, n.: The study of how to make the best choice when outcomes are uncertain, by weighing how likely each result is against how much it matters. and AIXIAIXI, prop. n.: A mathematical description of a perfect learning agent that weighs every possible explanation of its experience and acts to get the most future reward. It cannot actually be built, but it defines an ideal., which assume unlimited computation.
Related
- satisficingsatisficing, n.: Choosing the first option that is good enough instead of searching for the very best one — a sensible strategy when time, information or brainpower are limited.
- Herbert A. SimonHerbert A. Simon, prop. n.: An American scientist of decision-making who, with Allen Newell, wrote some of the first AI programs and argued that minds and computers both think by manipulating symbols.
- heuristicheuristic, n.: A rule of thumb: a quick, usually good-enough way to make a decision or to estimate how close you are to a goal, without any guarantee of finding the best answer.
- decision theorydecision theory, n.: The study of how to make the best choice when outcomes are uncertain, by weighing how likely each result is against how much it matters.
- embedded agencyembedded agency, n.: The puzzle of how an agent should reason when it is itself part of the world it is trying to understand and change, rather than standing outside it.
- intelligent agentintelligent agent, n.: In AI theory, a decision-maker, such as a robot, a self-driving car, a chess program or a helpful piece of software, that notices what is going on around it and responds in ways that serve its goals.
- logical uncertaintylogical uncertainty, n.: Uncertainty about math or logic facts that are settled in principle but that you haven't worked out yet — like not knowing a distant digit of pi until you compute it.
- AIXIAIXI, prop. n.: A mathematical description of a perfect learning agent that weighs every possible explanation of its experience and acts to get the most future reward. It cannot actually be built, but it defines an ideal.
Sources
- Simon, H. A. (1955). A Behavioral Model of Rational Choice. Quarterly Journal of Economics 69(1).
- Simon, H. A. (1957). Models of Man: Social and Rational. Wiley.
- Russell, S. J. & Subramanian, D. (1995). Provably Bounded-Optimal Agents. Journal of Artificial Intelligence Research 2.
Mentioned in
satisficingsatisficing, n.: Choosing the first option that is good enough instead of searching for the very best one — a sensible strategy when time, information or brainpower are limited., logical uncertaintylogical uncertainty, n.: Uncertainty about math or logic facts that are settled in principle but that you haven't worked out yet — like not knowing a distant digit of pi until you compute it.