Lave, Charles A. & James G. March. 1993(1975). An Introduction to Models in the Social Sciences. Lanham: University Press of America

2.9 THREE RULES OF THUMB FOR MODEL BUILDING


Model building as you have done it in this chapter is not a novel activity. It is something we do all the time. We speculate about things that happen to us or that we see happening to others. It is not mysterious, but it probably can be improved by a little attention to some elementary rules. In Chapter Three we will suggest some more detailed rules of thumb. Here we will simply note three general rules that we have been using repeatedly in making the speculations in this chapter. They are probably sensible much of the time, though they are not absolute truths.

Rule 1: Think "Process." A good model is almost always a statement about a process, and many bad models fail because they have no sense of process. When you build a model, look at it for a moment and see if it has some statement of process in it.

Example

Your chemistry professor shows up in class but has forgotten to bring along last week's homework papers. He apologizes, and you turn to the person next to you and say, "What can you expect from absent-minded professors?" This is your explanatory model for the professor's behavior. This is a common, ordinary, but poor model. Look at it for a moment. Where is the process? One way to put a process into the model is to ask why professors are absent-minded. If you think about it for a moment, you will be able to think of a number of processes that might produce absent-minded professors.

Model 1. Busy people try to devote their limited time to the things they consider most important. The professor does not consider teaching important, and so he did not bother to go to his office and find the homework papers.

Model 2. You become a professor by learning to be a good problem solver. Good problem solving involves almost single-minded concentration. So the professor occasionally forgets to do one thing because he is concentrating on another.

The models are different from each other, but each involves a sense of process, or relationship. One way to be certain that your models involve a sense of process is to see if you can derive general relational statements from them, that is: The greater X is, the greater Y will be. Thus Model 1 contains the following general relational statement: The busier someone is, the more likely he is to concentrate on important things. And Model 2 contains this general relational statement: The tougher the problem and the harder someone is concentrating on it, the more likely he is to forget other things.

Rule 2. Develop Interesting Implications. Much of the fun in model building lies in finding interesting implications in your models. In the problems associated with this course you will repeatedly be asked to develop interesting implication from some model. Whether something is considered interesting obviously involves a judgment, but there is a good strategy for producing interesting predictions: Look for natural experiments.

Example

An uninteresting prediction from Model 1 would be: Make the professor value his students more, and he will then become less absent-minded. Or from Model 2: Get the professor to work on easier problems, and he will become less absent-minded. These are relatively uninteresting because they ask us to run an experiment in a situation in which we probably cannot.

The way to find more interesting predictions is to think about the process involved in each model and then look for natural instances in which the key variables in the process vary. In Model 2, for example, it is not simple to vary the difficulty of the professor's problems, but you can easily find instances of similar situations and hence can predict that people (business executives, architects, football coaches) in other occupations that demand concentrated abstract thought will occasionally forget things, too. Or you can pedict that the professor will be just as absent-minded when engaged in his laboratory research as when he is engaged in teaching.

Or, for Model 1, you cannot easily make the professor value some given class of students more, but you can search for natural occurrences of this event. For example, if you believe that he values students in his graduate research seminar more than the students in his freshman introductory class, you would predict less absent-minded behavior with respect to the graduate students. Suppose you did make such observations and discovered that he was equally forgetful in his graduate classes; and furthermore that his freshmen lectures are well prepared, that he seems to have great quantities of careful notes, and that he often spends so much time answering questions after the freshmen class that he is late for his next class. You would then be highly skeptical of the truth of Model 1.

Rule 3: Look for Generality. Ordinarily, the more situations a model applies to, the better it is and the greater the variety of possible implications. Finding generality involves the ordinary process of generalizing nouns and verbs.

Example

Expand "college professors" to "busy people"; expand "forgetting the homework papers" to "forgetting anything"; expand "bringing papers" to "one kind of work." Finding generality also involves asking repeatedly why the process we have postulated is true. We ask: Is there another model that, if true, would include our model as an implication? That is, we look for a more general model that predicts our model and other things as well. Model 2, for instance, can be generalized to a large family of learning models that can be formulated to predict what woul happen if people learned to be good social scientists (see section 2.7) or executives (see Chapter 6).

From such simple heuristics, a little experience, some playfulness, and a bit of luck come good models, and some bad ones. Indeed, it is the creativity with which we specify bad models that leads us to good ones.