Heuristics, History of

The study of heuristics is an interdisciplinary and time-honored enterprise, with heuristics being examined across a wide range of fields, some focusing on professionals’ decision-making. In psychology, two influential research traditions have investigated which cognitive heuristics people use across various tasks. One tradition, which originated in the 1970s, has focused on the circumstances under which heuristics cause deviations from classical norms of rationality. The other, starting in the 1990s, has emphasized that heuristics are neither good nor bad per se, but that their success depends on how they are matched to environmental structures. This view suggests that heuristics, rather than leading to irrationality, enable ecological rationality. What Is a Heuristic? A heuristic is a simple but useful method for problem solving, decision-making, and discovery. The origin of the term goes back to the Ancient Greek verb heuriskein, which means ‘to find out’ or ‘to discover.’ Heuristics are sometimes also referred to as ‘mental shortcuts’ or ‘rules of thumb.’ One of their key functions is to reduce the complexity of a problem by ignoring part of the available information or searching only a subset of all possible solutions. Traditionally, heuristics have been regarded as necessary and efficient tools, but ones that produce only second-best solutions. However, more recent psychological research has shown that – under conditions that are ubiquitous in the real world, namely, limited knowledge and uncertainty – heuristics can in fact outperform more complex strategies. Heuristic methods have been considered in various scientific disciplines (though what exactly is meant by a heuristic varies across fields). Two key treatments of heuristic methods can be distinguished. In philosophy,mathematics, operations research, and artificial intelligence (AI), heuristics have primarily been investigated as prescriptive procedures specifying how a reasonable solution can be found given constraints such as computational intractability and limited time. In biology and psychology, by contrast, heuristic principles have also been used as descriptive models, that is, as models that describe how people and other animals sample information from the external and internal (memory) world, and how they make decisions based on that information. The next section gives a historical overview of the discussion of heuristics within these two contexts. A Short History of Heuristics Heuristics as Prescriptive Procedures Heuristic methods were first developed in philosophy and mathematics as a solution to the problems of algorithmic approaches to complex problems. To illustrate the algorithmic tradition, let us take the mechanical device developed by the Catalan philosopher Raimundus Lullus (1232–1315) in the thirteenth century. The device was able to automatically generate all combinations of religious and philosophical attributes that could be used in a debate (Figure 1). It consisted of six concentric discs representing the basic classes of arguments; each of these classes in turn had nine further attributes. Rotating the discs against each other produced every possible combination of attributes, thus automatically generating different arguments and potentially producing new ones. The German philosopher Gottfried Wilhelm Leibniz (1646–1716) had a similar, though even grander goal in the seventeenth century, namely to develop an algorithm for solving any conceivable problem, using a universal language that would allow every possible problem to be represented. It became increasingly clear, however, that such algorithmic approaches could easily lead to combinatorial explosion. Subsequent scholars therefore explored heuristics as means to bring Figure 1 A sketch of one of the discs of Lullus’ system for deriving all combinations of arguments. Source: http://www.medienkunstnetz.de/ works/ars-magna/. International Encyclopedia of the Social & Behavioral Sciences, 2nd edition, Volume 10 http://dx.doi.org/10.1016/B978-0-08-097086-8.03221-9 829 problem solvers to a solution without the need to explore the space of possibilities exhaustively. René Descartes (1596–1650), the French polymath of the seventeenth century, had formulated simple rules that were supposed to guide the problem solver toward the relevant aspects, rather than to all parts of a problem. In the nineteenth century, the mathematician and philosopher Bernard Bolzano (1781–1848), who proposed various heuristics for problem solving in his renowned Theory of Science (Wissenschaftslehre), further developed this approach for epistemic agents (e.g., the method of attempting to find truths by means of something that is not yet known to be true, as opposed to deducing truth from known truths). These scholars’ heuristics consisted of rather general procedures that were described in relatively vague terms. Their main goals were to offer a road to knowledge beyond proper deduction, to avoid an arbitrary solution process, and to foster creative thinking. For instance, many heuristics aimed at finding useful problem representations by way of analogy or metaphor. Heuristic methods for problem solving and discovery received wider recognition inmodernmathematics through the work of the mathematician George Pólya (1887–1985). Pólya’s (1945) procedures consisted of simple rules, such as dividing the process toward a solution into simple steps by, for instance, finding an analogy to a problem, finding a more specialized problem, or decomposing and recombining the problem. Procedures inspired by Pólya later also informed the new field of AI. Relatedly, heuristics proved to be of great practical value in operations research, an applied field of mathematics. Here, they have been implemented as computer-based tools for planning and aiding decision-making in industry. Methods for simplifying decision-making have also been discussed in economics. Many decisions are made under uncertainty, with key aspects of the problem remaining unknown. In the absence of complete knowledge of the option space and the probability distributions, normative principles of rational choice are difficult or even impossible to apply. Heuristics for making decisions under uncertainty have therefore been proposed (Savage, 1954). The maximin rule (Coombs et al., 1970), for example, selects from a set of options the one that would yield the most attractive outcome under the worst-case scenario, ignoring both the outcomes to be expected under better conditions and the (unknown) probability distribution. The satisficing principle is a heuristic for sequential decision-making (Simon, 1956). Instead of aspiring to identify the globally best option, this principle evaluates each option according to whether it meets a certain minimum aspiration level, and the first option encountered that satisfies that requirement is chosen. Another research field of economics in which simple, heuristic decision strategies have received attention is game theory. One of the most frequently studied strategies is titfor-tat. This strategy, which can be applied in repeated games (e.g., the iterated prisoner’s dilemma), determines whether or not a player cooperates with his/her opponent. At the first encounter with the other player, tit-for-tat cooperates; in subsequent encounters, it simply copies the other’s behavior (cooperate vs defect) in the previous encounter. In a classical computer simulation tournament (Axelrod, 1984), tit-for-tat proved to be the most successful – though simplest – of the strategies submitted. As well as informing attempts to improve conflict resolution, the study of simple strategies for cooperation within game theory offers insights into how cooperative societies might have evolved in the first place. The advent of the computer as a computational tool and metaphor of the mind in the 1950s spawned attempts to simulate intelligent behavior in machines. The goal of the new field of AI has been to develop computer programs that can perform tasks such as playing chess, proving logical theorems, or understanding language – and a common challenge is, again, to find ways of limiting boundless and therefore prohibitively expensive search of the problem space. In contrast to previous treatments of heuristics, which often portrayed them in rather vague terms, heuristic rules in AI research have been formulated precisely, often in terms of computational models. A prominent heuristic method for limiting search is means‒ends analysis, which was developed in the context of Alan Newell and Herbert Simon’s (1972) General Problem Solver system. To move from a current state in the problem space toward a goal state (representing the solution), means‒ends analysis reduces the distance between the two by first addressing themost important dimension on which the two differ, followed by the second most important dimension, and so on. Heuristics as Descriptive Models of the Mind On the assumption that heuristics, though not perfect, are often effective tools for dealing with a complex and uncertain world, it seems reasonable to suppose that the mind employs them naturally and spontaneously. Indeed, the psychologist Karl Duncker (1903–40) argued that human problem-solving strategies rest on heuristic principles (Duncker, 1935). Similarly, Gestalt psychologists conceived of perception in terms of heuristics. Max Wertheimer (1880–1943), for instance, identified a set of simple principles that organize sensory input to yield object perception – such as the ‘laws’ of proximity, closure, and similarity (Wertheimer, 1923/1938). As in other contexts, these principles of perceptual organization represent best guesses that usually can be trusted, although they do not work all of the time. They are useful because they are adapted to certain regularities in the environment; for example, elements belonging to the same object are typically in close vicinity to each other (the law or heuristic of proximity; see Figure 2). Presently, cognitive heuristics are most commonly studied in research on how people make decisions. Inspired by Herbert Simon’s (1916–2001) seminal work on ‘bounded rationality’ – the study of how people reason and make decisions with limited computational and informational resources, and when Figure 2 The Gestalt ‘heuristic’ of proximity: objects positioned close together are perceived as belonging together. 830 Heuristics, History of