Petter Nafstad Explanations The explanatory context An explanation could be regarded as an answer to a question being: Why x? Why is x the case? Why did x occur? If we look at this way we may come to regard explanations as elements of certain discourses. We may also come to see that the relevance or adequacy of an answer is partly determined by the kind of question it is an answer to. The elementary form of any explanation: x because y We therefore have three elements: 1. An x holding the place of something to be explained (explanandum), or rather a sentence describing some state of affairs to be explained. 2. An y representing a description of some other state of affairs that is assumed to explain x (explanans). 3. An inference from y to x. This extremely simple form opens up for a number of questions and possible problems: 1. The correct identification and description of an x. 2. The independent identification and description of an y. 3. The justification of a mechanism that is such that “if y then x”, or rationalisation of the statement “if y then x” or “x because y” We might add a forth problem/question (that might be conceived as a specification of 3.): 1 4. Is the inference from y to x one of deduction or one of some degree of probability? (Cf. the relationship between explaining and predicting). There is also a possible fifth question (that might be a bit idiosyncratic on my part): 5. Is the relationship between y and x a logical relationship between sentences or an empirical relationship between states of affairs) We will go straight to 3 and return to the others later. Traditionally we distinguish between three types of explanations, causal explanations, functional explanations and intentional explanations. Causal explanations x is assumed to be an effect of a certain cause, y. To justify such an assumption in a specific case, certain conditions must be fulfilled. 1. y must occur temporarily prior to x. 2. There must be some reason to assume that there is a necessary connection between y and x, given certain conditions (causal conditions). 3. It must be shown that these “certain conditions” really obtain. 1 and 3 poses no problems of principal interest. The question whether or not these conditions are fulfilled is one that can only be answered through empirical investigations. The more intricate problems arise in connection with 2 - how can we demonstrate the existence of a necessary relation? The standard way to do this is to construe the relationship between y and x as elements in a logically valid inference (What is a logically valid inference?). The inference is justified as an assumed description of a causal relation. P1. Whenever y (under conditions c) then x P2. y 2 P3. c _______________ Therefore x Explanandum is here considered explained. This kind of explanation if also called deductive-nomological. The obvious question here will of course be this: How do we justify P1? Some problems in this connection: The problem of induction. (assumptions of essentialism, essential features, natural kinds) The probability problem - probability of cause vs. probability of chance. The problem of identifying the relevant features of a certain state. In what respects are different cases similar? Logical and empirical necessity The kind of necessity that is demonstrated above is a logical one. That means that it is not possible, without self-contradiction to confirm the two premises and negate the conclusion. This should however not be confused with empirical necessity. Logics concern relationships between linguistic entities (terms and sentences). Empirical relations (for instance “necessity”) concern relationships between phenomena and events that we intend to describe or name through such entities. A rather fundamental question might be this: What does “necessity” mean? When the question relates to logical necessity it is quite easy to answer. For useful purposes it means exactly what 3 is stated above. Or, put otherwise: Necessity is here a property of logically valid inferences. The property is such that if the premises are true, then it is impossible that the conclusion is false. But when it comes to empirical necessity the matter is not so straightforward. I believe that we get the most fruitful conception of empirical necessity if we consider it to be a relationship between two phenomena/events y and x that is such that: whenever y (under conditions c) then x. I believe that this is what we intelligibly can mean when we use the term “necessary” as a property of assumed empirical relationships. If we believe that it is any more to it, then we should leave it to theologicians to decide. I take this to be what “necessity” means, the semantical meaning of the term, when applied to empirical phenomena. Another question is whether or not there really is sch a thing as the intended reference of the term. A note on laws and co-variation It might be said that social research is not really concerned with whether or not the connection between two phenomena (operationalized as variables) if a necessary or law-like one or merely one of observed co-variation. Therefore one might think that the condition of necessity (every time y then x) is in some way inadequate when it comes to explanations in social research. It is certainly true that social statistics are engaged in establishing measures of covariation being as accurate as possible, and that such co-variations should not be confused with necessary relations. But on the other hand, they should not be confused with explanations either. An established co-variation between two variables is a piece of data that we might or might not want to explain. Such data (the co-variation) belongs to the x-term in the explanatory scheme. To explain a co-variation is to take a step further. If we were to take that step, we would probably first ask ourselves what variable is the independent and what is the dependant one. If we call the independent 4 a and the dependent b, we would probably look for some explanans (the y-term) that would allow us to formulate an hypothesis like: x (a co-variates with b) because y. This y would then probably mention some characteristics of a that could be argued to (under the assumption of some law-like mechanisms) to cause the corresponding variation of b. Another thing: The question of generalizing co-variations from limited cases does not resemble any problem of induction when the population from which the cases are samples are always numerically finite. Functional explanations y is an effect of x when: 1. x is in some way advantageous to some subject (in the logical sense of subject). 2. y contributes to the maintenance or continuations of x. 3. there is no conscious or intended relationship between x and y. This implies that the temporal sequence between x and y is different in functional compared to causal explanations. The following basic question therefore arises: How can something, x, be explained by reference to an effect, y, that occurs temporally after the same x? Perhaps the question might seem easy to answer if we hadn't added condition 3. But again, if we hadn't added this condition we wouldn't have a functional, but rather an intentional explanation. The standard, or rather the most unproblematic use of functional explanations is to be found in biology rather that in the social sciences. 5 “x: Religion is a phenomenon that are occurs in every society because y: it contributes to the integration of society (y).” Is it possible to make the “because” intelligible? Is it anything more than a word? Intentional explanations x is typically a human action or the product of such. The y-term consists of some actors 1. Desires (intentions) 2. Beliefs These explanations rely heavily on “folk psychology” and they might seem so obvious that the question of justification might seem a bit far-fetched. A second thought might however reveal that things are not that simple. Striking problems are these: 1. How can we identify someone's intentions (desires) independent of their actions? Another, related problem is this: 2. How can we formulate an x-term independent of assumptions about the y-term? 3.The adequacy of an answer is, among other things, a function of the language-game within which the question is asked! I will now address the first and the second of these questions. 1. How can we identify someone’s intentions independent of their actions? According to the ideal of giving intentional explanations to human action, a question about why a person is reading a newspaper or is running towards the subway-station is supposed to be answered in intentional terms. Intentional explanations are of course based on the assumption that the ways we conceive of our own mental states have explanatory force as to how we in fact act. This in turn is based on the assumption that descriptions of mental states in terms of propositional attitudes might be true descriptions. For instance; If x is running towards the subway-station and we explain this behaviour by saying that she does so because she wants to reach the train, then the following statements are supposed to give true descriptions of her mental state: 6 1. x desires to reach the train. 2. x believes that the train is leaving soon. One more condition applies: It should be possible to identify the references of 1 and 2 independent of the fact that it claims to explain. However, we should consider it an open question whether or not this is possible. How should we for instance identify “a desire”? Three possible answers immediately present themselves. (1) A desire is a tendency or disposition to bring about a state described in a proposition, namely the proposition that together with the attitude mentioned (desire) makes the propositional attitude in question. In that case our explanation of x's behaviour would go like this: x runs towards the subway-station because x is in a mental state that tends to make her run towards the subway-station. Does this explain anything? Hardly! It is much like saying that water is refreshing because it has a refreshing capacity. (2) The other possibility is to define a desire in terms of antecedent conditions: A desire is something that people get under certain conditions. Given that someone depends on the subway to get to work, then this someone would desire to reach the subway. The explanation would then go like this: x runs toward the subway-station because x depends on the subway to get to work (and x is late). This seems to be a perfectly reasonable explanation. However, it works perfectly well without the use of any mentalistic term (desire). If the proposition expressing the explanation is true and if we accept it as an explanation, then x's behaviour can be explained without reference to her mental state at all. But then again, this would hardly be an intentional explanation. Consider how easy it is in daily life encounters to predict, or rather anticipate, the behaviour of other people without any reference to their mental states or propositional attitudes. You walk into a grocery store and ask for milk. You anticipate that the man behind the counter is going to get you some milk, which he does. Why? Well, because you asked for milk! You enter a bus and anticipate that the driver is going to drive to the destination where the sign in the front says that the bus is going. Surely he does just this. Why? Well, because that is where his schedule tells him to go! You stop your car at an intersection and behave towards the intersecting driver in 7 a way showing that you are going to wait. He enters the intersection. Why? Well, because you showed him that you were going to wait! These explanations, or rather explanatory fragments, work perfectly well, and they make no use of mentalistic terms. (3) A desire (intention) is something that people report to be having when asked the right questions. The content of the desire is identical with the propositional content of the report. To this several things could be said. First: A report of the kind in question is a response to a question, not to the situation that preceded the action. Second: This response could be conceived as the respondent’s interpretation/presentation of her own action addressed to another person/or institution. We should expect that such a presentation will be adapted to current norms and codes for reasonable behaviour. Third: There are reasons to believe that we often deceive ourselves about our own motives. As a response to this it might be said that we can construct questionnaires that will reveal motives that the respondent didn’t intend to reveal. A possible answer to this could be that such motives (hidden to the actors themselves) would reveal just as much of the assumptions of the author of the questionnaire as of the intentions of the actors (This would have to be specified in each single case). 2. How can we formulate an x-term independent of assumptions about the y-term? Perhaps one would say that intentional explanations do not presuppose the possibility of independent identification of motives or intentions. The motive, one might say, is considered to be part of the action of any actor as far as this actor is considered rational and capable of being ascribed an action at all. The reason would be that the concept of action in some way incorporates the concept of an actor, which, in turn, presupposes the concept of an I who has taken some decision. It seems to me that this concerns the way other people are able to relate to someone's actions and only indirectly how the actor herself is able to do this. It is, in my opinion, indeed true that partakers in human communication and interaction relate to their own actions through some kind of anticipation of the likely response of other people. But that also entails 8 that the explanandum of an intentional explanation is an action as it seen from at least one more perspective than that of the actor i.e. an action at the level of social mediation. The basic problem that is raised in the headline question is, in my opinion this: It seems difficult to describe an explanandum (the x-term) independent of certain assumptions about the intentions that, in the next turn is considered as explanans (the y-term). But if so, it seems to be the case that so-called intentional explanations resemble descriptions rather than explanations. How do intentions enter into action-descriptions? It is obvious that if you describe someone’s action as one of revenge, as a lie, as a result of a rational choice as a theft, as a piece of benevolence etc. such descriptions (or classifications) will include some assumptions about intentions (or motives). In such cases intentions will be considered from a certain perspective, the one that relates them to institutional facts. Some statements might seem to be descriptions while they in fact will be better understood if considered as ascriptions. An ascription may be contested with reference to the institutional context within which the action has occurred.1 An institutional fact has this basic form2: x counts as y in c (This peace of paper counts as money in certain contexts. This gesture counts as a confirmation. Etc.). Conditions for an x to be y (when y is an institutional fact): 1. We must believe intent that x be y. 2. The existence of y is observer-dependent. 1 See H. L. A. Hart: The Ascription of Responsibility and Rights. Proceedings of the Aristotelian Society (1948-49). 2 Cf. John Searle: The Construction of Social Reality, 1995 and Mind, Language and Society, 1998. 9 3. The intention mentioned I y must be collective (we intend that …). 4. The must be some assignment of function expressed in the intention. 5. The assignment of function must work through rules that are constitutive (as opposed to regulative). In addition to this we must consider intentions as something having certain conditions of satisfaction (or conditions of success). The bottom line of this is that intentions are (often) entangled in social institutions and should be understood relative to the conditions that these represents with respect to the characteristics of action. If x is an action and we want to describe x, we must also understand the institutional fact to which x relates. Should we then disregard intentions? So why not use Ockham’s razor and say that the explanation is done when we have described the antecedent conditions? The obvious answer would be that we wouldn’t be able to rationalise the mechanism presupposed in the “because”. The intentions are supposed to provide the missing link between antecedent conditions and the action. But, if we conceive of such intentions as reasons, we can still ask: Why did x act on those reasons rather than do something else? Admittedly, the examples used above are very simplistic, in fact to such a degree that they perhaps don’t catch the real problem at hand. It might be objected that they doesn’t instantiate behaviour that need to be explained at all. If we really should mean that intentional explanations don’t explain anything, we should try this assumption against cases where it is not obvious why people act as that do, i.e. cases that need to be explained. It might be the case that two different desires, a and b, might lead to the same action and we don’t understand the action because we don’t know whether the actor acts from a or b. Or a certain desire might lead to different actions, c or d, and we don’t understand why the actor chose c rather than d. In the first case it might seem that an explanation should be supported by arguments that the actor had a specific desire, in the second that she had a certain belief. 10 However, all this admitted, it doesn’t follow that such explanation of complicated cases also applies to simple cases. The first kind of explanations implies some sort of elimination of possible alternative desires or beliefs. They can therefore be construed as a reduction of possible complexity. But what about cases where there are no conscious awareness of alternatives? If anyone would try, from these examples, to infer that we could, in general explain human action without the use of mentalistic terms, the obvious objection would be that such terms (or rather the phenomena to which they refer) are, in some way, presupposed in interaction of the sort mentioned, and that the phenomena thus presupposed should be part of the explanation of the action. And basically I believe that this is a correct and adequate objection. We should however ask ourselves what this objection does and does not imply. Being presupposed does not imply being object of conscious awareness at the moment of action. At most it implies the possibility of being the object of conscious focus. But if so, then mentalistic terms are not necessarily parts of the self-understanding of the agents as they act. If we still insist that they should be part of the explanation of action, we would therefore also have to say that such explanations might go beyond the conscious intentions of the agents. Now, this seems very reasonable; but would such explanations still be intentional explanations? If so, we would also have to say that “intentions” include more than what is part of the conscious mental content. This wouldn't necessarily be very problematic. We might say that human action is performed on the background of some intersubjectively shared convictions about what is the case, that convictions about mental states are among these and that such convictions might, if the demand is made, be mobilised as reason for the actions even though they are not at the moment in the focus of awareness. This distinction between smooth, unproblematic, linguistically mediated interaction and, on the other hand, the mobilising of reasons corresponds to Habermas' distinction between communicative action and discourse (Cf. Habermas, 1981). If I have understood Habermas correctly, he is partly formulating a norm for rational interaction and cooperation and partly describing a communicative practice and its foundation in a modern differentiated life-world. But does the possibility of discourse also explain the pre-discoursive interaction? I believe that it might do so in cases where this possibility is part of the conscious awareness when the action is initiated, like when you know that you can throw in more reasons if necessary without spending time figuring out in 11 advance what those reasons more precisely would be. But in many instances this is simply not the case. In many cases the discourses are actions distinctively different from the communicative actions (or other kinds of actions) that precede them. In such cases the mentalistic terms might be part of the explanation of the temporally last action, but not of the first. This doesn't, I believe, conflict with Habermas' theory, as this is set forth as a theory of the preconditions for certain kinds of rational action with a certain degree of consensus-orientation, not action in general. Reasons and causes “Why does S do A?” An answer might consist in showing good reasons for S to do A. Still, “why did he act on those good reasons?” To answer the why, the reasons must also be assumed to work as a cause. Four problems3 1. Reasons are states, not events. Causes are events. Therefore reasons cannot be causes! Answer: States are in fact implied in causal explanations, namely as causal conditions. If say that the window was broken because it was hit by a stone, we imply that the glass was in a certain state. If mental dispositions shall lead to actions, there must first be some event to which the action is a response. 2. Causes are logically distinct from effects. That is: description of causes do not imply certain descriptions of effects. On the other hand: Intentionalistic descriptions of action (performed from reasons) implies that the description of the reasons are correct. Therefore: Reasons cannot be causes! 3 After Davidson: Actions, Reasons and Causes. I Davidson: Essays on Action and Events. Oxford 1980. 12 Answer: At the bottom of this argument there is a confusion between descriptions and the phenomena described. A description can imply another description. A real state cannot imply another real state. Any state can be described in different ways. If John kills Paul because of revenge, his state can also be described otherwise (for instance as a certain neurological state). 3. Causal explanations are nomological explanations (explanations from laws). There are no known laws connecting reasons and actions. Therefore explanations from reasons cannot be causal! Answer: We actually consider relations as causal even if we don't know any law from which we can deduce certain events. If someone kicks a stone and the stone moves, people have known the causal connection before Newton formulated the three laws of movement. 4. Our actual experience of the relationship between our reasons and our actions are inconsistent with a theory of reasons as causes. Answer: This argument is founded on the presumption that our own feelings of the ways we act are also the most secure source to the real causes of action. But it is a fact that we often deceive ourselves as to why we do certain things. Validity and relevance Could the classical distinction between the validity and the relevance of arguments be applied to evaluation of explanations? Yes! The validity-question relates partly to the description of y as independent of assumptions about x, and partly on the empirical correctness of the description. The relevance-question focuses on the justification of the “because”. Explanation chains 13 We have been talking about explanations on a very elementary level, the elementary form of explanations. In real research and real projects you will build your explanations on other explanations that you assume to be correct. It you for instance want to explain the different rates of spread of the HIV in different areas of Africa and you locate the explanans in different degrees of trust to the health authorities, you will probably base your explanatory hypothesis on other explanations of the contraction of the virus (for instance that it spread through sexual intercourse). This explanation might be based on other explanations identifying an explanans on other physiological levels (molecular or genetic). There might be different ways to think of the relationship between the links in such explanatory chains. One might be to say that such presumed explanations help us formulate an x as accurately as possible so as to make us better equipped when we try to formulate a y. Another might be to say that we are here dealing with different levels of description. petter.nafstad@uit.no 14