The Arithmetic of Human Behavior

Abstract | Introduction | Request Book (PDF) | Buy from Amazon

Download Introduction (PDF)

Social scientists are often criticized for leaving out much of what makes us human. They are apt to reply that if we hope to get at what is really going on, we have to cut away the incidental, disclose the underlying consistencies, and from that identify basic structures, forces, and mechanisms. Leaving things out isn’t only an analytic tactic forced on us by the complexity of the human world, it is central to a scientific approach.

This is a good reply given its assumptions, but what if the human world does not work the way it assumes? What if the patterns of human behavior we observe do not stem from underlying consistencies? What if the underlying consistencies, like the rules of a game of chess or computer hardware, just allow a wide variety of ways things can happen? In that case explanations are not restricted to consistencies leading to consistencies. Reliable patterns of behavior can arise because there are many ways the same thing can happen: inconsistent ways to reach consistent results. In this case behaviors are explained by the sum of the probabilities of the diverse ways they happen, and a science looking to tease out consistencies leading to consistencies is an exercise in misdirection.

Mathematically, the idea of different and even improbable inputs leading to a probable output is not problematic. It is simply the effect of a disjunction under uncertainty. But are disjunctions a plausible mechanism for explaining much of the human world?

Consider, for a start, how much ordinary observation, as well as scientific studies, find different motivations and histories leading to the same behaviors; or that linking multiple inputs to one output, which is generally hard to do in the physical world, is quite easy to do when the linkages are in the mind, so disjunctions can be an efficient way to produce consistent outcomes; or that disjunctive mechanisms explain how our diverse and inconsistent selves can produce consistent patterns of behavior without hypothesizing unseen underlying forces or taking recourse in high levels of abstraction. But the most compelling argument that disjunctions are pervasive in thehuman world is that, once we know what to look for, we can see that they are hidden in plain sight, and all but universal.

An example is the easiest way to see how this works. There are numerous reasons why people don’t leave their jobs. A quick ten: the job seems secure, starting a new job is stressful, they have friends at work, the boss is not too bad, the commute is short, the salary is good enough, jobs they’d rather have require training they can’t afford, the company provides subsidized daycare, the work is interesting, the health benefits are good. The ten generate 1024 combinations of reasons, each combination a reason, some better some worse, to stay on the job. For most people there will be a number of combinations that will keep them on the job. Consider someone who might stay if the boss is a pain, but the work is interesting, and the benefits are good; that person might also stay if the work is dull as long as the daycare is available and the job is secure; and further, that individual might also stay if the commute is short and friends work there, at least until they can afford training or a better job; and so forth—quite different combinations, but as long as one of them holds the person stays. The probability of that person staying on the job, then, is the probability of combinations A or B or C or… or N. In short, a disjunction.

Similar examples are easily found. Just list conditions that can influence the likelihood of a behavior. Odds are that they can produce that behavior in various combinations, so there is a disjunction. When we know what to look for, it becomes harder to find examples of reliable behaviors where disjunctions are not involved than when they are.

Once these disjunctive mechanisms are recognized we are faced with the possibility that the great intellectual apparatus associated with science and rationality is not well suited to making sense of the human world. Conventional foci — commonalities, central tendencies, distinguishing signal from noise, parsimonious and unifying theories and models, abstracting to general concepts and covering laws, theory testing, generalizability — might as well have been designed to overlook human disjunctions. Reduction to basic principles is not a powerful strategy if those principles (like the rules of chess) just tell us that things can happen in lots of different ways. A competition of ideas is not appropriate when many of the ‘competing’ ideas can be valid, ‘cooperating,’ parts of the disjunction.

In terms of basic logic, we are trying to understand disjunctive (A or B or C…) phenomena with conjunctive (A and B and C…) reasoning: a mismatch between the logic of the phenomena and the logic of our explanations. This is a poor foundation for a rational enterprise.

The practical problem is that we cannot readily replace conjunctive understandings with disjunctive ones. Our minds are not built to handle the myriad possibilities, probability estimates, calculations, and data handling that explanations depending on sums of probabilities require. We can handle this with the aid of computers, if we have the data and analytic tools designed to work with disjunctions, but this is only a practical solution if we can invest the resources required. For the most part we will have to rely on reasoning that is inadequate if not misleading: more conjunctive than disjunctive. But we can apply heuristics and strategies that improve how reasoning is interpreted, compensate for its distortions and limits, and take advantage of information conventional reasoning discounts.

This book is divided into four main sections. The first makes the argument for the role and pervasiveness of disjunctive mechanisms in the human world, it’s basic arithmetic, and how we distort evidence and logic to make conjunctive thinking appear credible. The second looks at practical strategies we can, and sometimes do, employ to cope with disjunctive mechanisms we cannot adequately grasp. These strategies utilize the predictive power of correlations between commonalities in a disjunction’s inputs and an output. More generally, however, they seek useful actions by capitalizing on disjunctive mechanisms. The third section presents a statistical approach to analyzing disjunctive human mechanisms, called Disjunctive Mapping, with a worked example. Analyses are based on computing the sums of the probabilities of the various ways outcomes occur, and measuring the influence of individual factors in the contexts of each of those ways. (An appendix provides JMP scripts for those who might want to try it.) The fourth, a conclusion, steps back from the particulars of the arguments to try to get a sense of where they leave us.

Top of Introduction