Bounded Rationality

Sud Alogu
6 min readJun 15, 2023

Herbert Simon, a pioneer in the field of cognitive psychology, introduced the concept of “bounded rationality” — a concept that I cover in “The End of Wisdom: Why Most Advice is Useless.

Bounded rationality posits that individuals make decisions based on the information they have at hand, their cognitive limitations, and the finite amount of time they have to make a decision. Simon’s work laid the groundwork for the study of heuristics, as he suggested that individuals use these mental shortcuts to make decisions within their cognitive and temporal constraints.

Herbert Simon’s concept of “bounded rationality” has had a profound impact on various fields, including economics, psychology, and artificial intelligence. This theory acknowledges the limitations of human cognition and the constraints of the real world, suggesting that individuals make decisions based on the information they have at hand, their cognitive limitations, and the finite amount of time they have to make a decision.

Simon’s work was revolutionary because it challenged the prevailing assumption in economics and decision theory that individuals always make perfectly rational decisions. Instead, Simon proposed that individuals are only “boundedly rational,” meaning that their decision-making capabilities are constrained by their cognitive resources and the information available to them. This perspective has significant implications for understanding human behavior and designing systems that interact with humans.

Simon’s theory of bounded rationality also laid the groundwork for the study of heuristics. He suggested that individuals use these mental shortcuts to make decisions within their cognitive and temporal constraints. This insight has led to a rich body of research on heuristics and their role in human cognition.

One of the key insights from Simon’s work is that heuristics can often lead to good decisions, even when they are not optimal. This is because heuristics are designed to work well in the types of environments that humans typically encounter. For example, the “recognition heuristic” suggests that if one of two objects is recognized and the other is not, then the recognized object is inferred to have the higher value concerning the relevant criterion. This heuristic can be very effective in situations where recognition is a reliable cue for the criterion of interest.

Bounded Rationality and Economics

In economics, the traditional model of human behavior was that of the “rational actor,” who always makes decisions that maximize their utility. However, Simon argued that this model was unrealistic and did not accurately reflect how people make decisions in the real world. He proposed the concept of “bounded rationality,” suggesting that individuals make decisions based on the information they have at hand, their cognitive limitations, and the finite amount of time they have to make a decision.

Simon’s theory of bounded rationality has had a profound impact on economics. It has led to the development of behavioral economics, a field that combines insights from psychology and economics to provide a more accurate model of human decision-making. Behavioral economics acknowledges that people often make decisions that are not perfectly rational due to cognitive biases and heuristics, a concept that owes much to Simon’s work.

Imagine you’re at a grocery store, and you see a sign that says “Limit 12 per customer” for cans of soup. Now, you had only planned on buying one or two cans of soup, but suddenly you find yourself thinking, “Well, if the limit is 12, maybe I should buy more. There must be a soup shortage I don’t know about!” So, you end up buying six cans of soup, even though you don’t really need them.

This is an example of the scarcity heuristic at work, where we perceive items as more valuable when they are less available. The grocery store sign triggered a sense of scarcity, leading you to buy more soup than you initially intended. It’s not a perfectly rational decision (after all, you now have more soup than you know what to do with!), but it’s a common example of how cognitive biases and heuristics can influence our behavior in economic situations.

Heuristics and Psychology

Simon’s work also had a significant influence on psychology, particularly cognitive psychology. He suggested that individuals use heuristics, or mental shortcuts, to make decisions within their cognitive and temporal constraints. This idea has led to a rich body of research on heuristics and biases in human decision-making.

Simon’s work on heuristics challenged the prevailing view in psychology that human cognition is slow and deliberative. Instead, he proposed that much of our thinking is fast and automatic, guided by heuristics that often lead to good decisions. This perspective has significantly influenced the field of cognitive psychology and has led to the development of dual-process theories of cognition, which posit that human cognition consists of both slow, deliberative processes and fast, automatic processes. This should remind you of the work of Kahneman and Tversky, Thinking: Fast and Slow.’

Imagine you’re at a party, and you see a bowl of what appears to be delicious guacamole. Without a second thought, you scoop a generous amount onto your chip and take a big bite, only to realize that it’s actually a bowl of wasabi!

This is a classic example of Kahneman’s System 1 thinking in action. System 1 is fast, automatic, and often guided by heuristics or mental shortcuts. In this case, the “green paste in a bowl at a party” heuristic led you to believe that the substance was guacamole, not wasabi. Your System 2, the slower, more deliberate thinking process, didn’t have a chance to intervene and consider that the bowl’s contents might not be what they seem.

This might not be a perfectly rational decision (after all, a moment’s reflection could have saved you from a mouthful of wasabi!), but it’s a funny example of how our fast, automatic thinking can sometimes lead us astray. So, next time you’re at a party, remember to engage your System 2 before diving into the dips!

AI in Chess and Other Games

Simon’s work on bounded rationality and heuristics has also had a significant impact on the field of artificial intelligence. AI systems often face similar constraints to humans, such as limited computational resources and incomplete information. Simon’s work has inspired the development of AI algorithms that use heuristics to make efficient decisions under these constraints.

One of the most well-known applications of heuristics in AI is in game-playing algorithms. For instance, in chess, AI doesn’t calculate all possible moves up to the end of the game (which would be computationally infeasible) but instead uses heuristics to evaluate the desirability of different positions and guide its decision-making. These heuristics might include rules like “control the center of the board” or “minimize the potential for the opponent’s checkmate.”

AI in Route Optimization

In logistics and transportation, AI algorithms use heuristics to solve complex route optimization problems. For example, the “nearest neighbor” heuristic is often used in the traveling salesman problem, where the goal is to find the shortest possible route that visits a set of cities and returns to the origin city. The heuristic simply involves choosing the nearest unvisited city as the next city to visit. While this doesn’t always yield the absolute shortest route, it provides a good solution in a fraction of the time it would take to calculate all possible routes.

AI in Machine Learning and Pattern Recognition

Machine learning algorithms often use heuristics to guide their learning process. For instance, in decision tree algorithms, heuristics like “information gain” or “Gini impurity” are used to decide which attribute to split on at each node of the tree. Similarly, in clustering algorithms like k-means, the heuristic of minimizing the distance between data points and their assigned cluster centroids is used to group similar data points together.

The Impact and Limitations of Heuristics in AI

The use of heuristics in AI has enabled the development of efficient and effective algorithms that can tackle complex real-world problems. However, it’s important to note that heuristics are not infallible. They are designed to work well on average or in typical situations, but they may perform poorly in certain cases. For instance, the nearest neighbor heuristic for the traveling salesman problem can yield suboptimal results if the cities are arranged in a particular way.

Moreover, the use of heuristics in AI raises important ethical and transparency issues. Since heuristics simplify complex decision-making processes, it can be challenging to understand and explain why an AI system made a particular decision based on a heuristic. This lack of transparency can be problematic in high-stakes domains like healthcare or criminal justice, where understanding the rationale behind decisions is crucial.

--

--