Explanation In The Social Sciences Research Paper

1. Explanation-Seeking Questions

Explanations can be understood as answers to certain questions, the so-called explanation seeking questions. In the social sciences, a very important type of explanation seeking questions refers to actions of a certain kind performed by a certain person. An explanation of individual actions is not only of interest in itself, but it is also required for a proper under- standing of many macrosocial phenomena that are construed as the aggregate result of individual actions of large numbers of persons. As Little (1991) puts it, all social processes, causal influences, systemic interactions, etc., are ultimately embodied in the actions of individual actors within a specific social and natural environment. This may be one reason why most publications about explanation in the social sciences primarily, or even exclusively, deal with the explanation of human actions.

Suppose a person p performed an action of kind a. There are at least the following eight questions relevant to the explanation of this singular action a of person p (cf. Tuomela 1977):

(Q1) Why did p do a?

(Q2) What was the reason of p for doing a?

(Q3) What was the motive (emotion, feeling, sensation, personality characteristic) which led p to do a?

(Q4) Which of the wants (desires, decisions, duties, obligations, norms) of p brought about his or her doing a?

(Q5) Which of the beliefs (thoughts, considerations, knowledge) of p brought about his or her doing a?

(Q6) What was the purpose (intention, aim, end, goal) of p for doing a?

(Q7) How is it possible that p did a?

(Q8) What does p’s doing a mean (express)?

Answers to these questions are not independent from each other. Q1 could be answered by referring to causes and reasons for action a, whereas Q2 seems to accept only the latter. Q4 and Q5 which already mention certain kinds of reasons in the formulation of the question are included in Q2. Thus, an answer to Q4 and Q5 can be transformed into an answer to Q2. An answer to Q3 can be interpreted as an answer to Q1 or Q2. Tuomela assumes that answers to questions Q1–Q5 are statable as:

(A1) p did a because of c.

Here c gives the causes or reasons for p’s doing a. All answers to Q6 can be paraphrased in the form

(A2) p did a in order to do g.

Here g is p’s goal action for the purpose of which a was performed. An illustration of A2 would be: p saved money in order to enjoy life after retirement.

Tuomela claims that answers of the type A2 can be further reduced to answers of the type A1 and that the expression ‘because of ’ in A1 signifies a causal relation between a and c. Answers to how possible questions of type Q7 are incomplete variants of A1. Explanations of individual actions by means of sociological variables such as minority status, social class, or social position of an agent are good examples of answers to questions of type Q7.

Answers to questions of type Q8 are of a totally different nature. They belong to category E4, i.e., the explanation of the meaning of an entity, in that they reveal an additional understanding of action a. The so- called speech acts, that are linguistic actions, may illustrate this type: if action a of person p consists of the utterance ‘I will do that!,’ an appropriate answer to the question ‘What does p’s saying ‘‘I will do that!’’ express?’ might be ‘This utterance of p is a promise, not only an announcement or a prediction.’

Answers of type A1 seem to be fundamental in the explanation of human actions. Differentiating between types of psychological explanations allows a more comprehensive treatment of component c of A1.

2. Types Of Psychological Explanations

Bunge and Ardila (1987) suggest the following classification of psychological explanations: (T1) tautological explanations, (T2) teleological explanations, (T3) mentalist explanations, (T4) metaphorical explanations, (T5) genetic explanations, (T6) developmental explanations, (T7) environmental explanations, (T8) evolutionary explanations, (T9) physiological explanations, and (T10) mixed explanations. Each type shall be briefly explained (in the sense of E4) and illustrated by two simple examples.

T1: Tautological explanations refer to basic capabilities or mental faculties: person p is able to anticipate (or anticipates) events because of p’s forethought capability. Person p is able to think about (or thinks about) herself because of p’s self-reflective capability.

T2: Teleological explanations refer to goals or purposes (cf. component g in A2): person p sold his shares in order to maximize his profit. Person p studied medicine in order to become a doctor.

T3: Mentalist explanations refer to mental events: person p developed a neurosis because p suffered from an intrapsychic conflict. Person p works so hard because p is highly motivated.

T4: Metaphorical explanations refer to analogies with physical or social processes or with animals or machines: person p processes information like a parallel processing computer does. Person p’s drivings follow the same principles as a steam boiler.

T5: Genetic explanations refer to the genetic endowment of a person: person p has a high intelligence because her or his parents also had a high intelligence. Person p suffers from Alzheimer’s disease because p carries two apolipoproteine E type 4 alleles.

T6: Developmental explanations refer to stages or levels of biological or psychological development or to events in the persons’s past: person p does not understand the lever laws because p is not yet in the stage of formal operations. Person p suffers from agoraphobia because p has experienced a severe trauma in early childhood.

T7: Environmental explanations refer to external conditions and factors: person p shows a high response rate because p is working under a variable ratio reinforcement schedule. The agoraphobic symptoms of p are weakened because p is massively exposed to the threatening stimuli.

T8: Evolutionary explanations refer to the survival value of a behavior or behavior tendency, its selective advantages or disadvantages: male persons show sexual jealousy in order to ensure paternity in putative offspring by preventing encroachment by competing males, and this behavior tendency has evolved because of the selective advantage of a greater reproduction rate of these persons (cf. Buss 1997). Persons have a high pain threshold under duress because of its survival value.

T9: Physiological explanations refer to physiological, especially neurophysiological and endocrinological, processes and mechanisms: introverts are more arousable than extraverts because introverts have a lower threshold for activation in the ascending reticular activating system than extraverts. The depressive person p experienced an elevation of her or his mood because p took a cyclic antidepressant which increased the chemical neurotransmitter serotonin.

T10: Mixed explanations are combinations of two or more of the above mentioned types of explanation. Mostly, explanations are not pure cases of one type, but combinations of at least two types, i.e., they are mixed explanations. Cases of one type usually provide only a partial answer to the respective explanation seeking question. In answers of the type A1 to explanation seeking questions and in most of the given examples of the different types of psychological explanations, the relation term ‘because’ occurs. It links that which is to be explained to that which explains. A characterization of this relation is provided by the so-called models of explanation.

3. Models Of Explanation

Explanation is one of the main subjects of the philosophy of science. So far, many different models of explanation have been constructed, most of them incompatible with one another (cf. Salmon 1989). Each model attempts to answer at least two questions: (a) What is (the structure of) an explanation? and (b) What is a good (proper, appropriate, adequate) explanation? Some answers to these questions will be briefly outlined.

3.1 The Deductive-Nomological Model Of Explanation

According to this classical view, an explanation is an argument which shows that the phenomenon to be explained can be inferred from certain other facts by means of specified general laws. This type of argument may be schematized as a deductive inference of the following form (cf. Hempel 1965):

E as the description of the phenomenon to be explained is called the explanandum (sentence). The statements of antecedent conditions, which make assertions about particular facts, and the general laws together form the explanans, i.e., that which explains. Explanations of this kind are called explanations by deductive subsumption under general laws, or deducti e-nomological explanations. (The root of the term ‘nomological’ is the Greek word ‘nomos’ for law.)

D-NE gives an answer to the question ‘What is (the structure of) an explanation?’ A comparison between this structure and the one underlying the examples of the different types of psychological explanations reveals that most of the latter are stated in an elliptical form, viz., ‘E because of c’ (cf. A1). The component c refers only to a subset of the antecedent conditions C₁, C₂,…, C_k that are required for a proper explanation of the explanandum E, and the general laws, if there are any at all, are totally omitted. Thus, explanations of the form ‘E because of c’ do not count as proper explanations.

To give an answer to the second question, i.e., ‘What is a good explanation?,’ requires the formulation of conditions of adequacy. In the case of the model of deductive-nomological explanation, there are four such conditions (cf. Hempel 1965):

(R1) The explanandum must be a logical consequence of the explanans, i.e., the explanandum must be logically deducible from the information contained in the explanans.

(R2) The explanans must contain general assumptions or hypotheses, and these must actually be required for the derivation of the explanandum.

(R3) The explanans must have empirical content, i.e., it must be capable, at least in principle, of test by experiment or observation.

(R4) The sentences constituting the explanans must be well confirmed.

It is not easy to find examples of deductive-nomological explanations in the social and behavioral sciences that display the proper structure and satisfy the conditions R1–R4. One example may be: the phenomenon to be explained is Peter’s frequent aggressive behavior in the presence of his mother. The explanation is given with reference to the theory of operant behavior. The explanandum statement E can be formulated as ‘Peter shows a high frequency of aggressive behavior if his mother is present.’ The principle of positive reinforcement ‘If a positive reinforcer is presented contingent on the occurence of a behavior, then the frequency of this behavior will increase’ can be the law statement L. At least two statements of antecedent conditions are required, viz., A₁ ‘Contingent on Peter’s aggressive behavior, his mother regularly shows attentive behavior,’ and A₂ ‘His mother’s attentive behavior is a positive reinforcer for Peter.’

In this case, the explanandum E seems to be deducible from the explanans, i.e., the conjunction of L, A₁, and A₂. Thus, R1 as the first condition of adequacy seems to be satisfied. But what about the conditions R2–R4? That depends on the status of L. If L would have empirical content and could be regarded as well confirmed, R2–R4 would be satisfied too. With regard to L, two problems arise: (a) According to its critics, L is no empirical law, but only a logical consequence of the definition of the concept of a positive reinforcer (the much discussed problem of circularity of the empirical ‘law of effect’); (b) even if this were not the case, L is not a well confirmed universal law because its validity depends upon conditions which are not explicitly mentioned in L. One important condition is the following: (C1) there exists no other behavior in the respective situation contingent on the occurrence of which a more effective positive reinforcer is presented. If there exists only one such other behavior in the case of Peter and his mother—and this is not excluded by the antecedent conditions A₁ and A₂—a quite different law statement would become relevant, viz., the ‘law’ of parallel reinforcement. If there exist even two or more of the behaviors excluded by C1 which seems to be the usual case in unrestricted natural environments, it is difficult to find an appropriate law statement at all.

This example is representative of the problems that conflict with any attempt to apply the model of deductive-nomological explanation to the social sciences. Universal deterministic laws as they are required by the model are hard to find in this domain. Nevertheless, Hempel (1965) who is the best known proponent of the model defended its significance for the social sciences and even for history. He considered any deviation from the standards set by the model as leading to insufficient and incomplete explanations. Nowadays, few social scientists are willing to share his view. But the idea of a nomological social science with strict laws and explanations of the deductivenomological type is still alive (cf. McIntyre 1996).

The trust in the possibility of a change in the direction of a nomological social science is based on a certain feature of scientific explanations, viz., that any explanation of an event is an explanation of that event under a certain description. The explanandum statement E in a deductive-nomological explanation does not represent the event to be explained as such, but the event under a certain description. It might be that the current situation in the social sciences is the consequence of an unfavorable choice of modes of description—that, at least, is the argument. Other choices which would lead to a redescription of the subject matter of the social sciences could, perhaps, reopen the way to a truly nomological social science. Unfortunately, McIntyre (1996) who is a confirmed supporter of this view has not produced any laws, offered any new social theories, or even given any redescriptions so far.

3.2 The Statistical-Relevance Model Of Probabilistic Explanation

It may be difficult to find genuine nomological laws in the social sciences, but it is easy to find statements which describe statistical regularities between events. ‘85 percent of the persons who have a panic disorder and who undergo an exposure therapy experience relief from their symptoms’ is an example. Concrete percentages or probabilities do not occur in every statistical statement. Instead, fuzzy terms like ‘nearly all,’ ‘most,’ ‘the majority,’ ‘more than half,’ etc., are used. Hempel (1965) had recognized this situation very early and broadened the scope of his theory of explanation by proposing his much discussed model of inductive-statistical explanation.

Referring to the above example of a statistical law, an explanandum E could be ‘Person p experiences relief from his or her symptoms of panic.’ Appropriate ante cent conditions would be A ‘Person p has a panic disorder’ and A ‘Person p undergoes an exposure therapy.’ The statistical law in conjunction with these two antecedent conditions constitutes the explanans in this case. The inductive-statistical explanation of E could be expressed in two ways: (a) the inductive probability of the explanandum given the explanans amounts to 0.85. (b) The explanans inductively supports the explanandum in the degree 0.85 (cf. Schurz 1995/96).

The objections against Hempel’s model of inductive-statistical explanation led Salmon (cf. 1989) to the formulation of his statistical-relevance model of probabilistic explanation. Whereas in Hempel’s model an explanation of an event is an inductive argument that confers upon the event to be explained a high inductive probability, an explanation of an event according to Salmon’s model is an assemblage of factors that are statistically relevant to the occurrence of the event to be explained, accompanied by an associated probability distribution.

Salmon and Salmon (1979) gave the following example to clarify this notion: Albert, an American teenager, is convicted of stealing a car. A sociologist who attempts to explain that act of delinquency will presumably look for certain factors that are known to be relevant to delinquent behavior, e.g., sex, age, marital status of parents, religious background, socio-economic status, residence (rural urban suburban), etc. Each of those factors, as well as many others, is assumed to be statistically relevant to delinquency in the sense that the probability of delinquency is different for boys than for girls, different for urban dwellers than for rural youngsters, different for children whose parents have remained married than for those from broken homes, etc. If any of these factors does not make some difference in the probability of delinquency, it will be discarded as irrelevant, and hence as lacking in explanatory value.

To explain Albert’s delinquent behavior, the population of American teenagers to which Albert belongs is partitioned into a series of subclasses in terms of such relevant factors as those already mentioned. For each subclass in the resulting partition, the probability for delinquent behavior on the part of members of that subclass will be determined. Then, Albert will be assigned to one such subclass in terms of his actual attributes, e.g., 17-year-old male, suburban dweller, middle-class undivorced parents, etc., and the probability will be cited for that subclass. If the number of relevant factors which enter into the partition is large, the probability obtained in a given case may be quite low. For Salmon, that is not crucial. What matters is that the relevant factors be taken into account in partitioning, and that the probabilities associated with the various subclasses be accurate.

This model is confronted with two problems. (a) The requirement of a homogeneous relevant partition of the reference class (in the example: the population of American teenagers) is not easy to meet. In most cases in which explanations of single events are looked for in the social sciences, a comparable probability distribution will not be available. (b) The fact that an event counts as appropriately explained even if the probability value associated with the respective subclass of the partition may be very low comes into conflict with deeply rooted preconceptions about proper explanations. With regard to Tuomela’s explanation seeking questions, the best that could be achieved by this kind of explanatory efforts is an answer to a question of type Q7, i.e., an answer to a how possible question.

These problems of the statistical-relevance model of explanation led others to challenge its claim to be a model for the explanation of single events. Schurz (1995 96) comes to the conclusion that the genuine explananda of statistical laws are not single events but sample frequencies. In his discussion of the explanatory value of causal modeling techniques in the social sciences, Woodward (1989) assumes that the statistical information in the delinquency example may contribute to the explanation of facts about population level parameters but not to the explanation of why any particular boy became a delinquent.

Even if this critique of Salmon’s model were to be accepted, it would still be of value to the social sciences, since it provides a standard for a deeper statistical-causal understanding of social phenomena as, for example, juvenile delinquency, unemployment, or terrorism in certain reference classes. This standard could also be applied to the increasing use of meta-analytic methods for explanatory purposes in the social sciences when several explanatory variables are involved at the same time (cf. Cook et al. 1994).

3.3 The Model Of Aleatory Explanations

A more recent model of causal explanations of specific events is the model of aleatory explanation (see Table 1) introduced by Humphreys (cf. 1989). He agrees with Salmon that causal explanations are possible within the realm of chancy, or aleatory, phenomena. But in his model, in contrast to Salmon’s, no probability value is assigned to an explanation. The demands on the knowledge base are much lower than in Salmon’s model, e.g., no complex probability distribution on the basis of a homogeneous relevant partition of a reference class has to be available.

What is required for a proper explanation of a specific event Y in S at t are two lists of causes of Y, i.e., a list of contributing and a list of counteracting causes. In Humphreys’ model, for something to be a cause it must invariantly produce its effect. But causes in this model are probabilistic causes, and they produce changes in the value of the chance of the effect. A contributing cause of Y produces an increase, a counteracting cause of Y a decrease in the value of the chance of Y.

This model seems to be applicable in various contexts within the social sciences (cf. Henderson 1993). It takes account of the fact that social phenomena are usually the result of multiple causal influences. It does not presuppose the existence of complete lists of all influences which, positively or negatively, affect a given outcome. Aleatory explanations are conjunctive. They can be improved by including additional probabilistic causes which may come up in further research.

The model has already been applied to explanations within a theory of social interaction in small groups (cf. Westmeyer 1996). But its importance goes well beyond that particular context. Take, for example, the interactionistic theory-frame of Magnusson (1980) under which many theories of social interaction can be subsumed. One fundamental assumption of this theory is that changes in the probabilities for behaviors of a person in a certain situation are a function of changes in the factors within this situation. Such determining within-situation factors are the physical situation, behaviors of the person concerned, and behaviors of the interaction partners.

In terms of Humphreys’ model, ‘Y’ would refer to a change in the probability of a behavior of the person concerned; ‘S’ would refer to the setting in which the intrasituational dynamic interaction between the person and the interaction partners takes place; ‘t’ would refer to a certain point or period in time at which the change in the probability of the behavior occurs; ‘Φ’ would refer to the set of all those determining withinsituation factors, the changes of which facilitate (contribute to) the occurrence of Y; and ‘Ψ’ would refer to the set of all those determining within-situation factors, the changes of which inhibit (act counter to) the occurrence of Y. Contributing and counteracting causes could be identified by methods of event sequence or time series analysis.

3.4 The Embedding Theory Of Explanation

Most of the models and notions of explanation dealt with so far are strongly connected to the so-called statement view of scientific theories that assumes theories to be sets of laws or theoretical assumptions. A very different picture is given by the structuralist view which is a variant of the model-theoretic approach to the understanding of scientific theories. Here, theories are reconstructed as complex set theoretical structures. The structuralist program has already demonstrated its fruitfulness in the analysis of various theories from the social sciences (cf. Stegmuller et al. 1982, Westmeyer 1992).

From this point of view, explaining a phenomenon means to explain it within the context of a certain theory. This requires to show (a) that the phenomenon can be construed as an element of the set of intended applications of the theory, and (b) that the theory can be successfully applied to that element. In short, what has to be done is to embed the phenomenon into the theory. This leads Bartelborth (1996) who gives the most detailed account of this idea to speak of the embedding theory of explanation. For a concrete example of this type of explanation in the social sciences see Westmeyer (1996). He shows that it is possible to combine the embedding of a phenomenon into a certain theory with the application of Humphreys’ model of aleatory explanation.

The embedding theory is the most promising variant of the unification approach (cf. Kitcher 1989) to the problem of explanation. ‘Explaining as unifying’ is a notion which increasingly attracts the attention of philosophers of science. It paves the way for an integrative treatment of explanation and understanding (cf. Schurz and Lambert 1994) which is especially of interest to those social scientists who are used to construe explaining and understanding as basically different scientific activities. It is in accordance with a coherence theory of truth and fully compatible with basic epistemological tenets of the more recent constructionist and constructivist approaches to the social sciences. These characteristic features of the embedding theory give rise to the expectation that this theory will play a leading role in future discussions of the issue of explanation in the social sciences.

Bibliography:

1. Bartelborth T 1996 Begrundungsstrategien [Strategies of Justification]. Akademie Verlag, Berlin
2. Bunge M, Ardila R 1987 Philosophy of Psychology. SpringerVerlag, New York
3. Buss D M 1997 Evolutionary foundations of personality. In: Hogan R, Johnson J, Briggs S (eds.) Handbook of Personality Psychology. Academic Press, San Diego, CA
4. Cook T D, Cooper H, Cordray D S, Hartmann H, Hedges L V, Light R J, Louis T A, Mosteller F 1994 Meta-analysis for Explanation: A Casebook. Russell Sage Foundation, New York
5. Hempel C G 1965 Aspects of Scientific Explanation. Free Press, Albany, New York
6. Henderson D K 1993 Interpretation and Explanation in the Human Sciences. State University of New York Press, New York
7. Humphreys P W 1989 Scientific explanation: The causes, some of the causes, and nothing but the causes. In: Kitcher P, Salmon W C (eds.) Scientific Explanation. University of Minnesota Press, Minneapolis, MN
8. Kitcher P 1989 Explanatory unification and the causal structure of the world. In: Kitcher P, Salmon W C (eds.) Scientific Explanation. University of Minnesota Press, Minneapolis, MN
9. Kitcher P, Salmon W C (eds.) 1989 Scientific Explanation. University of Minnesota Press, Minneapolis, MN
10. Little D 1991 Varieties of Social Explanation. Westview Press, Boulder, CO
11. Magnusson D 1980 Personality in an interactional paradigm of research. Zeitschrift fur Differentielle und Diagnostische Psychologie 1: 17–34
12. McIntyre L C 1996 Laws and Explanation in the Social Sciences. Westview Press, Boulder, CO
13. Salmon W C 1989 Four decades of scientific explanation. In: Kitcher P, Salmon W C (eds.) Scientific Explanation. University of Minnesota Press, Minneapolis, MN
14. Salmon W C 1998 Causality and Explanation. Oxford University Press, New York
15. Salmon M H, Salmon W C 1979 Alternative models of scientific explanation. American Anthropologist 81: 61–74
16. Schurz G 1995/96 Scientific explanation: A critical survey. Foundations of Science 3: 429–65
17. Schurz G, Lambert K 1994 Outline of a theory of scientific understanding. Synthese 101: 65–120
18. Stegmuller W 1983 Erklarung–Begrundung–Kausalitat [Explanation–Justification–Causality], 2nd edn. Springer-Verlag, Berlin
19. Stegmuller W, Balzer W, Spohn W (eds.) 1982 Philosophy of Economics. Springer-Verlag, New York
20. Tuomela R 1977 Human Action and its Explanation. D. Reidel, Boston, MA
21. Westmeyer H (ed.) 1992 The Structuralist Program in Psychology: Foundations and Applications. Hogrefe & Huber Publishers, Seattle, WA
22. Westmeyer H 1996 A concept of explanation for social interaction models. In: Hegselmann R, Mueller U, Troitzsch K G (eds.) Modelling and Simulation in the Social Sciences from the Philosophy of Science Point of View. Kluwer, Dordrecht
23. Woodward J 1989 The causal mechanical model of explanation. In: Kitcher P, Salmon W C (eds.) Scientific Explanation. University of Minnesota Press, Minneapolis, MN