C Hasty Generalization

Each of the following is a general proposition, or generalization:

1. Water always flows downhill.

2. Large breeds of dogs have shorter life spans than smaller ones.

3. People in France are not very friendly to tourists.

4. Italians are quick-tempered.

5.3 Fallacies of Context 119

5. Stockbrokers drive BMWs.

6. Bill never gets anywhere on time.

General propositions make a claim about a category of things (water, dogs, Italians, etc.), attributing some characteristic to all members of the category. As statement 6 il-lustrates, however, a generalization may be about a particular individual; what makes it a generalization is that it attributes a general pattern of behavior to that individual.

General propositions play a vital role in reasoning. They allow us to reap the advan-tages of having concepts: We form a concept for a type of thing, and then generalize about the attributes and patterns of action shared by things of that type. This saves us the trouble of having to discover from scratch the features of every particular thing we encounter. Knowing about the life spans of large versus small dogs, for example, can help us in choosing a pet.

General propositions are normally supported by observing a sample of particular cases. But we often draw conclusions too quickly, on the basis of insufficient evidence.

This fallacy, known as hasty generalization, can take many forms. A single bad experi-ence while traveling can prejudice our view of an entire city or country (as in 3 above).

Most of us have stereotypes about ethnic groups, professions, or people from different regions of the country, based on our exposure to a few individuals (as in 4 and 5 above).

Even a judgment about the character or personality of an individual is a generalization drawn from our observation of that person on specific occasions; here, too, we often jump to conclusions. And we can jump to conclusions about ourselves: We make a mis-take, fail a test, have a problem in a relationship, and then draw a sweeping conclusion about our inadequacies.

In Part 3, we are going to study the proper methods and precautions for general-izing from particular cases. But the essence of the method is to ask ourselves whether we have considered a wide enough context. Two contextual factors are especially impor-tant if we want to avoid hasty generalization: the number and the variety of cases in the sample from which we generalize. It is rarely if ever possible to draw a legitimate gener-alization from a single instance, or even from a few; we should consider a larger sample.

I may have met a few stockbrokers who drive BMWs, for example, but if I asked around I would quickly find some who don’t. We should also look at a variety of instances. For ex-ample, we might observe that Bill is late for work not just on one occasion but on many.

It would still be hasty to conclude that he never gets anywhere on time. Is he also late for parties, dates, ball games? If so, we have much stronger evidence that his problem has to do with time in general rather than with work.

Of these two factors, variety is usually the more important one. If you liked a cer-tain novel and wanted to know whether you might like all the author’s works, it would obviously be useless to read another copy of the same novel; you would want to sample other novels by that author. To take a more realistic situation, suppose you are buying a car and want to know whether a certain dealer offers good service on the cars it sells.

You might ask other customers of that dealer. If you ask people who are very similar, you run the risk that they are not representative of the dealer’s clientele. If those cus-tomers are all friends of the dealer, for example, or if they are all lawyers who are good at negotiating service contracts, then the generalization that this dealer offers good service might well not be true for other people—including yourself. To avoid hasty gen-eralization, we want to base our generalization on a sample of particular cases that are

representative of the wider group. Seeking variety in our sample is the best way to get a representative sample.

5.3D Accident

Hasty generalization is a fallacy that can occur in moving from particular cases to a generalization. There is also a fallacy, called accident, that can occur when we move in the opposite direction by applying a generalization to a particular case. This fallacy consists in applying a generalization to a special case without regard to the circum-stances that make the case an exception to the general rule. Accident is the fallacy of hasty application.

Hasty generalization

Generalizing from too few particular cases or from nonrepresentative ones

Abstract generalization

Particular instance(s)

Applying generalization to particulars in disregard of special features

Accident

Abstract generalization

Particular instance(s)

For example, matches are made to light when they are struck correctly, so I can infer that the next match I strike will perform as advertised. But what if the match is wet? If I disregard that fact and expect the match to light, I am committing the fallacy. Again, consider the generalization that birds fly. What does this imply about penguins? If we infer that penguins must be able to fly because they are birds, or conversely that they can’t be birds because they can’t fly, we commit the fallacy of accident. Penguins are properly classified as birds for anatomical and evolutionary reasons, even though the adaptation of their wings to swimming, and the heavy layer of body fat to insulate them from the cold, prevents them from flying.

In everyday speech, “accident” refers to something that is unexpected or happens by chance. In logic, however, the word retains the older meaning: “a nonessential prop-erty . . . of an entity or circumstance” (Merriam-Webster’s Collegiate Dictionary, 10th ed.).

Generalizations typically apply to things in virtue of the things’ essential properties.

When a generalization has an exception, it’s usually because of some accidental (i.e., nonessential) feature of that particular thing. Matches are designed to be used dry; if my matches got soaked in the rain, that’s an accidental feature. Penguins evolved from birds and they adapted to finding food in the cold waters of the Antarctic; their inability to fly is a by-product, and thus in a sense is accidental.

The classic examples of accident involve the application of moral principles to par-ticular situations. Take the principle that one should always tell the truth. What if a mugger on the street asks you where you live? Applying the principle mechanically to this situation by answering the question truthfully would be a rather dangerous case of fallacious thinking. Even if you lie but feel guilty about doing so, the feeling of guilt

5.3 Fallacies of Context 121

presumes that you should have told the truth to the mugger. Drawing that conclusion is fallacious whether you act on it or not. As an exercise for yourself, you might take other moral principles and imagine exceptional cases in which applying the principle would commit the fallacy. When we apply a given principle to a particular situation, we need to take account of other principles that may also apply and take account of any unusual consequences that might arise from applying the principle. It will not always be clear what the best decision is, and reasonable people may disagree. The point is not to apply the principle mechanically without considering the context.

Another common form of accident is extrapolating a generalization beyond the range in which we established its truth. For example, we have all learned that water boils at 212ºF. But that measurement was made at sea level. It would be fallacious to assume that the generalization is true for all locations. That’s not the case, as you have probably also learned. At higher altitudes, where the atmospheric pressure is lower, water boils at lower temperatures.

Not all generalizations have exceptions. In mathematics, there is a principle that if you have two equal quantities and subtract the same amount from both, the quantities will still be equal. That principle is true in all cases. In any area, on any topic, a strictly universal generalization has the form “All P are Q ,” and “all” means all, without excep-tion. If even one P is not Q, then the generalization is false. But if the generalization is true, we can safely infer that any particular instance of P will be Q. (This is one of the major forms of deductive inference, and we will examine it at length in Part 2.) If a generalization does have exceptions, moreover, and if we know exactly how and why the exceptions occur, we can qualify the generalization and restore it to strict universality.

For example, we can qualify the generalization “water boils at 212ºF” by adding “at sea level.”

For many generalizations, however, we may not know all the qualifications that would be needed to make them exception-proof. Or the exceptions may be so rare that we don’t bother to add the qualification. The fact is that in wide areas of knowledge—

including ethics, law, and politics and the properties and behavior of living species, to name a few—many of the generalizations we rely on have exceptions, and we need to avoid the fallacy of accident.