Exercises

Basic Instinct

Three centuries ago, Gottfried Wilhelm Leibnitz had a dream. The German philosopher and mathematician believed that all rational thinking could be described with a formula. He imagined that by inventing an alphabet of human thought—a system of characters for irreducible concepts—and then combining these in a calculus of reasoning, mathematicians would be able to solve all scientific and moral matters. He proclaimed that "a few selected persons might be able to do the whole thing in five years".

Leibnitz's grand scheme may seem naive, even ludicrous, but shades of it live on. From philosophers to economists, many modern researchers believe that rationality is objective and open to mathematical analysis. In their studies of reasoning and models of markets, they see rational decision-makers as supernatural beings with the logic of a supercomputer, boundless knowledge and all eternity in which to make a decision. Most of them readily accept that this view is unrealistic but, they argue, if there were no limits to our rationality we would be able to make the best possible choices. We may not think in this way, but we should.

This view is now being challenged. In an attempt to inject some realism into the study of rationality, Gerd Gigerenzer and his team at the Max Planck Institute for Human Development in Berlin are investigating the idea that evolution has endowed us with a set of mental shortcuts—tools for making quick decisions. Sure, the human mind can perform long-winded calculations and amazing feats of memory, they say, but in everyday situations we tend to use the shortcuts in our "adaptive toolbox". Gigerenzer and his colleagues have not only identified some of these mental shortcuts but also put them to the test.It may look like sloppy thinking when we jump to a conclusion or follow a gut feeling, but our mental shortcuts turn out to be astonishingly successful.

In the real world, a good decision is less about finding the best alternative than about finding one that works. Herbert Simon of Carnegie Mellon University in Pittsburgh was one of the first to recognise this in the 1950s when he coined the term "bounded rationality". He pointed out that the way any animal thinks depends on its cognitive limitations and the environment in which it lives. So a creature such as a field mouse, whose food is randomly distributed, needn't evolve complex foraging strategies, whereas one such as a lion, whose food sources are indicated by clues in the environment, will have an advantage if it can use sophisticated mental abilities such as planning.

Simon's ideas have become fashionable in recent years, and the Berlin researchers are leading the renaissance. They point out that our minds, like our bodies, have been shaped by evolution: we have inherited ways of thinking from those of our ancestors whose mental tools were best adapted for survival and reproduction. No time then for careful calculations—a cogitating ancestor would have risked losing dinner, a mate or even its life. Instead, our mental tools are fast and frugal. They allow us to make decisions based on very little information and using simple rules. Each tool, or heuristic, is designed to resolve a certain type of dilemma under certain circumstances. "There is no general tool," says Gigerenzer. "Our simple rules are problem-dependent."

Although they apply to different sorts of problems, heuristics have a common structure, which arises from the way humans make decisions. First, we search the environment for information, or cues, upon which to base a choice. A heuristic contains rules that direct the search. Next, we must stop searching. It's pointless trying to find out everything there is to know about a nut or berry if we starve in the process. Heuristics contain a stopping rule, often ending the search after only a few cures have been considered. Finally, we must make the choice—eat, run, mate, attack. But all the survival benefits of speed are lost if we make the wrong decision.

Perhaps the fastest and most frugal rule of thumb is the Recognition heuristic. Peter Todd, an evolutionary psychologist and cofounder of the research group, points out that, given Dr Seuss's famous menu of green eggs and ham, most people would opt for the ham. By choosing "the familiar" as the only cue worth considering, you get your calories without wasting time trying to discover whether green eggs are edible. Brown rats follow the same strategy, preferring to eat foods that they have smelt on the breath of other rats. But the recognition heuristic doesn't work only with food. Imagine you are a Stone Age man choosing a hunting party or a computer-age woman looking for business partners.Chances are you'll pick people you know, or have heard are good. The benefits are obvious.

In many situations, simply choosing what you recognise will work better than choosing at random. When Gigerenzer and his colleague Daniel Goldstein showed volunteers pairs of cities and asked them to identify the largest of each pair, people tended to choose the city whose name they recognised. When Americans were asked to distinguish between pairs of German cities, this strategy gave a 73 per cent success rate. Random guessing would have produced around 50 percent. What's more, the success rate fell to 71 per cent when the Americans were asked to do the same for cities in the US. This "more-is-less" effect happens because the Recognition heuristic does not work as well when you know too much. "There is wisdom in ignorance," says Gigerenzer. Though many choices will not succumb to such a simple approach, the researchers have identified several other shortcuts that make a decision based on a single reason. Heuristics called Minimalist and Take The Best, for example, search through a sequence of cues until they find one that distinguishes between the alternative courses of action. Minimalist is perhaps the natural progression from the Recognition heuristic. Forced to make a choice between two cities that you recognise but know very little about, you might consider a cue such as "Do the cities have an airport?" If only one does, then you assume that it is the larger city. If both or neither do then you consider another cue at random.

Take The Best, on the other hand, works well in situations where experience leads us to believe that we know which cues are most important. In choosing a mate, for example, many animals (including humans) have distinct priorities. Take The Best uses the cues in order of importance, stopping the search as soon as one cue distinguishes between the possible choices.

To see whether a single reason really can form a good basis for making decision, Goldstein, working with Max Planck researchers Jean Czerlinski and Laura Martignon compared Minimalist and Take The Best with two conventional analytical tools that use all available information—multiple regression and a simplified regression known as Dawes's rule. The researchers used all four algorithms to make predictions in 20 test areas. These included the dropout rates in various Chicago high schools, given such cues as ethnic composition and class size, and the incomes of academics, given cues such as gender, rank and years since graduation.

"The two fast and frugal heuristics always came close to, and often exceeded, the performance of the traditional algorithms," says Todd, "even though they only looked through a third of the cues on average." One reason for this success might be that in natural environments cues tend to be linked, so an exhaustive search may not provide much more useful information than a fast and frugal search.

In nature, one-reason heuristics seem to be used by parents to decide which of their offspring to invest in. Some birds, for example, always feed the largest chick in the nest, while others chose the hungriest or feed chicks at random. In Berlin, Todd and his colleague Jennifer Davis used computer simulations to show which single reason works best under various environmental conditions. They found that when food is scarce, feeding the largest offspring gives parents the greatest chance of getting their genes passed on. In times of plenty, however, the more egalitarian approach of choosing the hungriest or feeding at random is most successful. In the wild, most bird species do seem to follow such behavior patterns. Some, such as pied flycatchers and sparrow hawks, even change tack as the availability of food changes.

Davis and Todd also point out that human parents divide land between their children using similar reasoning. In cultures where resources are scarce, the eldest son tends to inherit the land, but where land is plentiful, it is divided more fairly among all the children.

Such single-reason heuristics may be very useful, but they do not work in every situation. Our adaptive toolbox has more complex equipment. Humans, like many animals, use body language to distinguish friend from foe. The way people move can tell you about their intentions—whether they want to fight, play or court, for example. "Some of the most obvious cues for intention can be assessed at a distance," says Todd. But it takes more than a one-reason heuristic to decide which intention the motion cues are pointing to.

Todd and Philip Blythe, another member of the Max Planck team, showed people virtual bugs on a computer screen that were programmed to suggest various intentions by exhibiting different cues—such as their speed and whether they moved in a straight line or meandered—to test a heuristic called Categorization By Elimination. This uses a succession of cues to whittle away the alternatives until only one remains. With just half the available cues, Categorization By Elimination correctly predicted two-thirds of the intentions—similar to the success rate of a trained human observer.

This match between the performance of a real person and a heuristic is common, and the researchers view it as evidence that we do indeed think in this way. But mental short cuts are not always the best option.The team has found that people tend to use more calculated reasoning when they can take their time, while heuristics come into their own when people are forced to think on their feet.

Even so, heuristics work in a broad range of situations. Not only do they allow us to choose between alternative courses of action, they also work when a choice doesn't come with all the options up front. If you're looking for a new frock, a new home or a new girlfriend you must search for the options—the available frocks, houses and women—as well as the cues with which to distinguish between them. How do you know when to stop looking and make a choice?

The answer is something that Simon calls "satisficing"—perhaps best thought of as a cross between satisfying and sufficing. He says that in these situations we set ourselves aspiration levels—which may alter over time—and stop looking only once these have been achieved. Todd and Geoffrey Miller from University College London, have used computer modeling to investigate Satisficing heuristics in mate selection. The most successful strategy for individuals seems to be  to learn their own rank in the mating hierarchy by looking at the quality of the partners who accept or reject them. They aspire only to those potential partners who match or exceed their assessment of themselves. Todd and Miller now aim to test how well this model matches the way people really search for partners.

But what about love? Our emotions, it turns out, can help us to make decisions too. Romantic love acts as a potent force, stopping the search for new partners. Love also highlights the importance of some cues above others. Indeed, all emotions seem to work in this way, so helping us to make decisions that effect our survival. Fear, for example, may narrow our options to just one: flight. Parental love leads us to care for our children regardless of the personal cost. And disgust keeps us from eating rotting food.

Survival and reproduction are the two cornerstones of evolution. And as evolution has shaped our adaptive toolbox, it is not surprising that it is chock-full of tools to solve problems such as finding food, avoiding predators, finding a mate and caring for offspring. You can imagine each tool labeled with a different sort of choice in a different environment. Gigenrenzer and his team are only beginning to reveal spanners and hammers, wrenches and saws.

What is clear, however, is that we are not born with a full set of shiny tools just waiting to be used. Instead, we seem to get a starter kit upon which to build, adapting tools and adding new ones as we learn about the world we live in. So, for example, different cultures conform to different social norms. By learning the rules, priorities and expectations of our culture, we may be able to take advantage of generations of acquired wisdom without really understanding why. We also learn that in some social situations unpredictability can be an advantage—giving you the edge over a competitor, for example. In such cases it is rational to be inconsistent, and heuristics can be adapted to allow for that.

Open the lid of the adaptive toolbox and you start to see that rationality is impulsive, emotional, flexible and inconsistent. It's a long way from the idealised decision-maker. As for a calculus of rational thought: dream on Leibnitz.

(2 239 words )

(From New Scientist, 4 September 1999 )

 Text