Implicit Inferences

Psychologists distinguish between the deliberative thinking that underlies deduction and the implicit, automatic, and largely unconscious inferences that help people to make sense of the world and its descriptions. Consider, for example, the following passage:

The pilot put the plane into a stall just before landing on the strip. He just got it out of it in time. It was a fluke.

Readers have no difficulty in understanding the passage, but every noun and verb in the first sentence is ambiguous. Also the search for the referents for the three occurrences of the pronoun "it" in the passage defeats even the most advanced computer programs for interpreting natural language. Humans have no difficulty with the passage because they are equipped with a powerful system that uses general knowledge to make implicit inferences. Readers should also have no difficulty in understanding the following passage:

Apart from her husband, a hairdresser, Eve was the only woman among 52 men on the tour. As a costumier, she filled a much needed gap, because when a company of actors is putting on a play in a different town each night, no damage to the costumes is too trivial not to be mended.

In fact, most people do not notice that the passage contains three deliberate mistakes. It implies that Eve's husband is a woman. It states that what is needed is a gap rather than Eve. It also asserts that no damage to the costumes is too trivial not be mended instead of what it surely means—no damage to the costumes is too trivial to be mended. The system of implicit inferences overrides the literal interpretation of the sentences and makes sense out of nonsense. The inferences resolve the senses of words and determine the references of pronouns and other such expressions. They enable individuals to construct a single model of the situation described in a passage, and the implicit system does not attempt to search for alternative models unless it encounters evidence for them. The process is therefore rapid, and it becomes as automatic as any other cognitive skill that calls for no more than a single mental representation at a time. For the same reason, implicit inferences lack the guarantee that their conclusions are valid. They are inductions rather than deductions. However, the implicit system is not isolated from the mechanisms of deduction. Normally, the two systems work together in tandem.

One consequence of implicit inferences is that people often jump to a conclusion, which later they have to withdraw. In logic, if a conclusion follows validly from premises, then no additional premises can invalidate it. Logic means never having to be sorry about a conclusion. As new premises are added to existing premises, then increasing numbers of logical conclusions follow (i.e., logic is "monotonic"). However, in daily life, conclusions are often withdrawn in the light of subsequent information. These inferences are "nonmonotonic". The original conclusion may have been based on an assumption made by default that turned out to be false. For instance, I tell you about my cat Hodge, and from your knowledge of cats you infer that Hodge has fur and a tail. You withdraw your conclusion, however, when you learn that Hodge is bald and tailless. Your knowledge contains various assumptions that you can make in default of information to the contrary. The whole purpose of these default assumptions is to allow you to make useful inferences that you can withdraw in the light of contrary evidence.

A more problematic sort of nonmonotonic reasoning is illustrated in the following example. You believe the following premises:

If Viv has gone shopping then she will be back in an hour.

Viv has gone shopping.

It follows, of course, that Viv will be back in an hour. However, suppose that Viv is not back in an hour. You are in a typical everyday situation in which there is a conflict between the consequences of your beliefs and the facts. At the very least, you have to withdraw your conclusion. You also have to modify your beliefs, but in what way? Should you cease to believe that Viv went shopping or that if she went shopping she will be back in an hour, or both? Philosophers and students of artificial intelligence have made various proposals about these puzzles. Unfortunately, the understanding of nonmonotonicity in human reasoning lags behind.

Reasoning in daily life often calls for the generation of explanations and diagnoses. For example, in the case of Viv's failure to return, you do not merely modify your beliefs, you try to make diagnostic inferences about what happened:

Possibly, Viv met a friend and went for a coffee.

Possibly, Viv felt ill on the way to the shops.

One possibility leads in turn to further explanatory possibilities, for example,

Possibly, Viv couldn't get the car to start after shopping.

.'.Possibly, the car's battery is dead. Possibly, Viv left the headlights on.

You use your knowledge and any relevant evidence to generate possibilities. Human reasoners easily outperform any current computer program in envisaging putative explanations. Given two sentences selected at random from different stories, such as

Celia made her way to a shop that sold TV sets.

She had recently had her ears pierced.

they readily offer such explanations as Celia was getting reception in her ears and wanted the TV shop to investigate, or Celia had bought some new earrings and wanted to see how they looked on closed-circuit TV. This propensity to generate explanations underlies both science and superstition. The difference is that scientists test their explanations empirically.

Inferences in real life are often not deductively ''closed''—that is, there is not enough information to draw a valid conclusion. Reasoners must therefore make inductions, that is, they use their knowledge to draw conclusions that go beyond the information given and that therefore may be false. There is no normative theory of induction and no comprehensive psychological theory of it, either. What does exist are a number of well-established heuristics, which were identified by two pioneers, Kahneman and Tversky. One heuristic is the availability of relevant knowledge. Most individuals, for example, judge that more people die in automobile accidents than as a result of stomach cancer. They are wrong, but the media publish more stories about auto accidents than about stomach cancer. Similarly, people rely on the representativeness of evidence. If you are told that Bill is intelligent but unimaginative and lifeless, then you are unlikely to judge that he plays jazz for a hobby, though you may find it more likely that he is an accountant who plays jazz for a hobby. If so, you have violated the principle that a conjunction (being an accountant and playing jazz) cannot be more probable than one of its components (playing jazz). The description of Bill, however, is more representative of an accountant than of a jazz musician. It has therefore led you to overlook a simple principle of probability.

Unraveling Alzheimers Disease

Unraveling Alzheimers Disease

I leave absolutely nothing out! Everything that I learned about Alzheimer’s I share with you. This is the most comprehensive report on Alzheimer’s you will ever read. No stone is left unturned in this comprehensive report.

Get My Free Ebook


Post a comment