Sandra LaFave

Donald Davidson, in an excerpt from his paper "Psychology as Philosophy"(1) argues that:

1. Psychological descriptions can’t be reduced to physiological ones.

2. The social sciences can’t be reduced to the physical sciences.

"Not all human motion is behavior ... but where there is behavior, intention is relevant." We may not always do what we intend (e.g., man who stomps on his own hat thinking it is the hat of his rival). But "happenings cease to be actions or behavior when there is no way of describing them in terms of intentions." (319)

This view is in sharp contrast to Skinner. For Skinner, behavior is simply bodily motion, mental events don’t exist, and methodological rigor requires that we make no references to intentionality. According to Skinner, we are things in a world of things.

Davidson agrees with Skinner in part: we are things in a world of things. But Davidson argues that this isn’t the whole story — a conclusion like that of Sartre, though for different reasons.

For Sartre, our experience of ourselves is essentially intentional. We cannot help but see the world from the perspective of a "for-itself", wanting, hoping, fearing, valuing. In all these modes of being, we freely create négatités (freely since all being is just in-itself and not incomplete in any way, and thus is unable to force any particular projection of négatité). And our freedom is the source of anguish and flight into bad faith.

Davidson begins from ordinary language rather than personal experience. According to Davidson, "intention is conceptually central" (319) to what we usually mean by a person’s "behavior". If psychology and other human sciences are mainly concerned with human behavior, and Davidson is right that intention is conceptually central to behavior, then psychology and the human sciences must be sciences mainly concerned with intentionality. (And Skinner must be dead wrong.)

So Davidson asks about the extent to which the connection between intention and behavior or action can be known and predicted: "Can intentional human behaviour be explained and predicted in the same way other phenomena are?" (320) I.e., are the human sciences the same sort of enterprises as the physical sciences?

Davidson’s answer is "yes and no". "On the one hand, human acts are clearly part of the order of nature, causing and being caused by events outside themselves. On the other hand, there are good arguments against the view that thought, desire, and voluntary action can be brought under deterministic laws, as physical phenomena can." (320)

In fact, the social sciences have produced only statistical correlations between intentions and actions. But might we hope for more precision as neurosciences advance? Davidson says no. The best we can ever hope for is statistical correlation. The human sciences are "nomologically irreducible" to the physical sciences. Intentional talk requires reference to consciousness: to the beliefs and desires of the agent. Any time you describe a specific intentional event, you make reference to the beliefs and desires of the agent. And every single individual intentional event no doubt can be correlated with a physiological event. But there is no general way to correlate beliefs and desires with physiological events or actions. Different persons may have different beliefs and desires, which nevertheless might produce the same actions. In other words, beliefs and desires (intentions) supervene on the physical world: the same physical world could be correlated with more than one intentional state.

But no such supervenience exists in physical descriptions of physical events (in a Newtonian world at least). In the physical world, there is a one-to-one correspondence between causes and effects in accordance with causal laws. It is possible to state the necessary and sufficient conditions for the application of a physical law – i.e., I can specify exactly the conditions for its application.

Thus for Davidson two parallel discourses – one deterministic, one non-deterministic – are inevitable. Thus the human sciences are never going to be swallowed up completely into the "hard" sciences.

The "nomological irreducibility" of the human sciences to the physical sciences "does not mean there are any events that are in themselves undetermined or unpredictable; it is only events as described in the vocabulary of thought and action that resist incorporation into a closed deterministic system. These same events, described in appropriate physical terms, may be as amenable to prediction and explanation as any." (320)

Davidson describes his metaphysical position as anomalous monism.He uses the word "anomalous" in the sense of its Greek roots: "a" meaning "not" and "nomos" meaning "law". I.e., he claims there can be "no laws" of the physical, deterministic kind that govern intentional events. Anomalous monism has the following four characteristics.

1. There is one world (monism).

2. Some events in this world are physical and some are psychological or intentional.

3. Every single psychological or intentional event is describable in physical terms.

4. But there can be no general laws correlating physical and psychological/intentional events – no "psycho-physical laws" connecting reasons and actions.

Why can there be no general psycho-physical laws? Why do intentions supervene on the physical world?

Davidson seems to be alluding to a couple of separate but related arguments:

1. Specifying someone’s intention always makes implicit reference to the agent’s point of view. A consequence of the logical issue of referential opacity in intentional contexts (explained below) is that we need to make an huge and probably practically unmanageable inventory of all of an individual agent’s beliefs and desires in order to specify the agent’s intention from the agent’s point of view (note that the phrase "the agent’s intention from the agent’s point of view" is really redundant if you understand what an intention is). In other words, it is unrealistic to expect that we can ever distinguish an individual agent’s intentional from non-intentional acts without enormously detailed detective work. And this just to uncover one intention of one agent. (And even this assumes a lot: e.g., that agents are usually aware of their intentions.)

2. But another consequence of referential opacity is even more devastating. If there were general laws linking intentions and behavior in a strict "hard-science" way, the intentions would have to be classified into general types. For example, we might seek to establish a law linking the intent to murder the President with some sort of behavior (actually murdering the President, or some displacement activity). "Murderous intent directed toward the President" would be a type of intent. And this classification would presuppose that we could identify the members of the class denoted by the type — to identify who intends to murder the President and who doesn’t. But referential opacity makes this impossible, because we can’t say to begin with when two people have the same intention.

Consider the first argument first. The idea is that we can’t know in any detailed, general way how to distinguish intentional from unintentional events in the first place. Since no general criteria exist for distinguishing intentional from non-intentional events, we can’t begin to formulate any general laws linking intentionality with action.

To see why, consider how you’d go about the task of identifying someone’s intention. You might think explanation of behavior (in Davidson’s sense of behavior) is a simple matter of linking a desire, a belief, and an action, and supposing the desire and belief to cause the action. For example, in the case of Achilles returning to battle because he wanted to avenge the death of Patroclus, you would identify the three elements as follows:

1. Achilles wants to avenge Patroclus’ death (intention/desire).

2. Achilles believes he can do this by going back to the battle (belief).

3. Achilles returns to the battle (action).

You would also stipulate that #1 and #2 cause #3.

But, Davidson claims, this model is inadequate. Knowing the "facts" of the desire, belief, and action, and specifying the causal link are necessary but not sufficient to identify the intention. We can get all these elements right and still miss what the agent intended. For example, "suppose, contrary to legend, that Oedipus, for some dark oedipal reason, was hurrying along the road intent on killing his father, and, finding a surly old man blocking his way, killed him so he could (as he thought) get on with the main job. Then not only did Oedipus want to kill his father, and actually kill him, but his desire caused him to kill his father. Yet we could not say that in killing the old man he intentionally killed his father, nor that his reason in killing the old man was to kill his father." (322)

Thus according to Davidson, we cannot distinguish intentional from unintentional events in the first place unless we can specify every relevant aspect of the agent’s belief system (not to mention the social context in which the behavior occurs, and the bodies and physical forces involved in the event). Because of what Davidson calls the "holistic" nature of the cognitive field, "any effort at increasing the accuracy and power of a theory of behavior forces us to bring more and more of the whole system of the agent’s beliefs into account." (321). And this task, per Davidson, is simply not practical.

(Furthermore, this doesn’t begin to address the question of whether we are always aware of our intentions. You could say, "Well, just ask Jane what she intends"; but we can take Jane at her word only if we’re certain that people know what they intend, and Freud, for one, seems pretty sure people are mostly unaware of their motives.)

Davidson’s Oedipus story exemplifies something Gottlob Frege first noticed, and Bertrand Russell expanded on: referential opacity in intentional contexts. Consider the argument:

Venus is the Evening Star.
The Evening Star is identical to the Morning Star.
Therefore, Venus is the Morning Star.

This argument is valid, i.e., its conclusion must be true if its premises are true. It has correct logic. But contrast it with

Jack believes Venus is the Evening Star.
The Evening Star is identical to the Morning Star.
Therefore, Jack believes Venus is the Morning Star.

This argument is invalid. The premises might be true and the conclusion false. The references to Jack’s beliefs make a difference to the logic.

Here’s another example:

Jack is President.
The President is an assassin’s target.
Therefore, Jack is an assassin’s target.

It’s valid. But

Jack hopes he is elected President.
The President is an assassin’s target.
Therefore, Jack hopes to be an assassin’s target.

isn’t as obviously valid, is it? The introduction of the intentional predicate ("hopes") changes the degree of certainty we have about the conclusion.

Davidson’s Oedipus example falls into the same type.

Oedipus left home intending to kill his father.
The surly old man is Oedipus’ father.
Oedipus left home intending to kill the surly old man.

The premises are true and the conclusion false, so the argument can’t be valid. The introduction of intentional contexts in these arguments makes it impossible to be certain about conclusions even if premises are true. Because of the referential opacity here (which seems inevitable in intentional contexts), we can’t tell for certain in any simple way how Oedipus’s action is caused by his intention, because we can’t know in any simple way (e.g., using the desire, belief, action model) what his intention is. We might change the explanatory model, but in order to escape the possibility of referential opacity entirely, we’d have to change it in a way too cumbersome to use easily; and that wouldn’t make referential opacity go away.

Besides, imprecision of this magnitude, at the most fundamental level of logic, is obviously antithetical to the hard science approach.

Consider a more prosaic example. I know Jane wants a promotion at her job, and I know that she has been reading Malloy’s Dress for Success , and I see her following Malloy’s advice ("Dress like people the next level up in the corporation"). So I explain her new look by saying her desire for a promotion causes her to follow Malloy’s advice. To avoid referential opacity, my explanation will be more reliable the more completely I can specify her belief system. For example, Jane’s new look might be an accident; she doesn’t really understand Malloy’s book at all, but just inherited a closet full of clothes from her recently-deceased twin sister, who just happened to have the sort of job Jane wants. So in order to buttress my explanation I add some of Jane’s relevant beliefs:

Jane believes getting a promotion would give her more money.

Jane believes more money is good.

Jane believes she can trust Tom Malloy.

Jane believes that if she follows Malloy’s advice, she won’t be fired.

Jane believes that if she follows Malloy’s advice, she won’t spontaneously explode, shrink to the size of an ant, be thrust into orbit...

Jane believes it is possible to follow Malloy’s advice.

etc.

Now, where do I stop? How much of Jane’s belief system must I specify in order to identify Jane’s intention for certain? Davidson’s answer: there is no obvious stopping place. I would have to be able to rule out all other possible intentional descriptions that may be employed by Jane; I would have to be able to show, for example, that Jane doesn’t have a peculiar learning disability such that she doesn’t understand the word "promotion" at all, and has just fallen in love with her boss and has decided to dress like her. Getting Jane’s intention right might require me ultimately to note the most fundamental and/or trivial elements of Jane’s belief-system: for example, that Jane believes in the existence of a world external to her senses.

The second argument lays out another consequence of referential opacity: we can’t decide when two people have the same intention, and thus can’t reliably identify classes of intentions, and thus can’t formulate psycho-physical laws linking intention and action. The argument for this can be illustrated by an example. Jack wants to kill the author of Waverly. Jill wants to kill the author of Ivanhoe. Neither knows that the author of Waverly and the author of Ivanhoe are the same person (Sir Walter Scott). Thus Jack and Jill have different intentions, no? If you ask Jack, for example, he’ll say "I don’t care about the author of Ivanhoe; it’s the author of Waverly I’m after." But saying Jack and Jill have different intentions seems wrong, too, since they clearly want the same thing, just under different descriptions. I.e., we are inclined to say two agents have the same intention if they want the same thing (if the object of the intention is the same). But now we have a dilemma. It seems absurd to say Jack and Jill have the same intention, if they wouldn’t agree. (Who knows one’s intentions better than oneself?) And it seems equally absurd to say their intentions are different, since they want the same thing. The case of Jack and Jill isn’t just a weird isolated example, since any object of desire can be identified in innumerable ways (e.g., in different languages). So we have a big problem in identifying when intentions are the same.

Davidson thus concludes that we can’t read people’s intentions off their behaviors, beliefs, or desires, except statistically. Most people who act like Jane would be trying for a promotion because that’s the most likely or rational conclusion to draw from Jane’s actions. But judgments of likelihood or rationality are inevitably based on arguments from analogy: from "the need to view others, nearly enough, as like ourselves. As long as it is behavior and not something else we want to explain and describe, we must warp the evidence to fit this frame." (324)

Thus, when we explain by reasons, we inevitably sacrifice the precision of hard science. " Explanation by reasons avoids coping with the complexity of causal factors by singling out one [the one that seems to us " most likely" or "most reasonable, of all possible ones], something it is able to do by omitting to provide, within the theory, a clear test of when the antecedent conditions hold [no clear test that someone is doing the action for the reason we picked]."(324)

COMMENT: Davidson’s "two languages" scheme would, as Stevenson says, provide a way out of the age-old dilemma of freedom and determinism, except that, as Searle (and many others) point out, the two languages are contradictory! One is determinist, the other isn’t. If Davidson weren’t a monist, his position would be practically Kantian (two realities, one describable only deterministically, the other describable only in the language of freedom). There may be life in the free will problem yet.