(This essay was written in 1979 and not published. I think it was intended to be longer.)

Here are some of the points:

1. The mathematically most tractable rationality is the optimization of a transferable utility in a single transaction. Often we wish to behave this way, but often neither our wishes nor our actual behavior can be described in this way.

2. Rationality is an approximate theory in the sense of LOGICAL THEORIES WITH APPROXIMATE CONCEPTS,

3. We want to be rational at least in the sense of transitivity of preference, but often we want total ordering. That is a person wants his own preferences to be totally ordered.

4. Rationality is most feasible in the short run. It is not easy to define what humanity wants between now and the heat death of the universe. It is even difficult to say what a person wants over his whole life.

5. Consider an animal that has certain drives. When it is hungry, it tries to eat; when it is sleepy, it tries to sleep. It is not plausible to ascribe a utility function to the life of a dog let alone a bacterium. This raises some questions:

What does it mean to do good to a dog? How is the welfare of a dog defined?

Much is apparent about the welfare of an animal in the short run. It harms a fly to swat it and a dog to kick it. It is good for a dog to be fed when hungry.

Consider a baby chick in the chicken house. Is it better for it to escape into the woods> It won't last long there - probably less than an hour. However, if it stays in the chicken house it will be fed and taken care of at the price of being slaughtered after six or seven weeks. In some sense, that's a good deal for the chick.

7. One of our human desires is to be rational. We also have other desires about our desires.

8. The simplest model of rationality occurs when one has a linear utility such that the utility of a sequence of events is the sum of the utilities of their individual outcomes and the events and the actions affecting them are independent. Then we can be sure that a person who adopts the strategy of maximizing the expected return from each event will with probability as close to one as desired end up with more utility than someone who adopts any significantly different strategy. When used as a betting model it requires infinite borrowing capability, because if losses can drive one out of the game, the outcomes of the events are no longer independent.

9. When each independent event is a distribution of one's wealth among investments with different probabilities of payoff but where there is no limit (upper or lower) on the amount that can be put into each investment, then one should maximize the expected value of the logarithm of one's wealth. A person adopting this strategy will end up wealthier than someone adopting a significantly different strategy.

10. A person sometimes can regard himself as being like an animal. He forms desires and seeks to satisfy them without fitting them into an overall life plan or measuring them numerically. Some people suppose that in principle, all desires should be numerically measurable, but this may not be true.

11. There is a view that one's lifetime utility is an integral over one's life experience. This seems very unlikely to me. In the first place people often form purposes whose achievement extends beyond their lifetimes.

12. John Rawls in his Theory of Justice suggests that a person should lead his life so that at its end, he will approve of what he has done. What do I care what that senile old codger (me not Rawls) will think?

13. An empiricist attempts to reduce all reality to experience. We will adopt a more general view in which a person's desires (or utility function if there is one) can be formulated in terms of events that occur in the world and not just in his experience. Thus, when we discuss a soldier who wants his country to win, we will not try to reduce this to his desire for certain experience, e.g. the experience of believing that his actions will contribute to the victory. We will take his willingness to sacrifice his life for this goal at face value - not as a construct in terms of the experience involved in making the decision to take the risky action.

14. The view that a person's desires are not merely an integral over his lifetime and that there are desires about desires admits views other than "materialism" in the sense of greed or sensuality as legitimate human desires. Of course, it doesn't dictate other values, but it seems to me that many people adopt "materialistic" values, because they can't logically conceive of others, and the religious appeals don't seem correct. Reading Solzhenitsyn's Harvard Commencement address reminds me that I share some of his religiously motivated concerns without believing any of the religion. To be precise, ... [In 1996 I can't remember what I was thinking about this in 1979].

15. Some people denounce rationality. This can have two motivations, one bad, one not so bad. The bad one is willfullness. It looks bad, it computes bad, but I want to do it anyway, e.g. the brain-washing of the communist and other cults. The other is an objection to the conclusions that other people claim to get from rationality, coupled with an inability to find the mistake. It is perfectly legitimate to say, "I am convinced there is a mistake in your argument, because the results conflict with beliefs or intuitions that seem as firmly based as the plausible assumptions from which your arguments started. Unfortunately, I'm not smart enough to find the mistake". Of course, this is somewhat of a self-indulgence is often mistaken.


Most humans want to behave rationally, and that's good. Rational people do more good and less harm than others even with a limited understanding of rationality and its limits. Most likely they are also happier.

However, it is important to understand the varieties of rational behavior and their limitations. Sometimes rational behavior is very well defined, and everyone can agree. In other cases, it is very subjective and murky.

An excessive confidence in one's own rationality and in one's model of the world can lead to great harm. The political fanaticism that has killed millions in the twentieth century is characterized by overconfidence in oversimplified views of the world.

In ascribing the bad actions of some people who claim rationality to error rather than to rationality, I am disagreeing with a prominent school of thought that attacks rationality itself. A common example given, is that of generals who believed computer output of enemy deaths in Vietnam and thought we were winning when we weren't. Believing the output of computers was taken as the essence of rationality, and so rationality was discredited by their failure. My point of view is that if the generals wrongly believed in the computer programs or the significance and reliability of the input to the computers or in a wrong model of the war that was embodied in the computer programs, then they were unsuccesfully rational.

We must make one concession to the critics of rationality. A person or group wanting to be rational and scientific may be inclined to wishful thinking about it. He may believe that he has a correct scientific theory when he doesn't. He may believe in the results of long chains of reasoning within a theory even when the results counter common sense and have immediate bad consequences. This may cause him to act worse than a person without a theory who may refrain from an action on the basis of its immediate and apparent bad consequences. Proper rationality requires a certain modesty about one's theories.

Not to be coy about it, the theories that come to my my mind as having led to the most bad actions recently have been variants of Marxism, although religious theories have had similar bad results in the past. Thus the Inquisition sometimes used a theory of demonic posession which justified torturing a person to drive the devils out. Further, since any earthly tortures were infinitesimal compared to those of hell, forcing a confession and repentance by torture was doing the victim a favor.

Likewise the Stalinist and Maoist theories of the "increasing intensity of the class struggle under socialism", justified widespread arrests and imprisonments without evidence. This was supported by the concept of "revolutionary justice" which didn't require definite charges or proof of guilt. Closer to home, Bruce Franklin's theory of the "revolutionary potential of the lumpenproletariat" justified associations of radical youth and violent criminals that caused dozens of deaths. [Franklin was a Maoist Stanford professor who headed an organization called Venceremos in the early 1970s]. I don't mind if the reader who is so inclined, makes up his own examples of the bad consequences of other wrong theories based on the crimes that are most salient in his own mind.

Up to: McCarthy home page

Send comments to mccarthy@stanford.edu. I sometimes make changes suggested in them. - John McCarthy

The number of hits on this page since 1996 April 20.