Notes on Causality and Explanation in the Social Sciences

Peter Taylor, July 1994

A. A base-line account of causality and explanation in the social sciences
Drawn from R. W. Miller, `Fact and method in the social sciences', in R. Boyd, P. Gasper, and J. D. Trout (ed.), The philosophy of science (Cambridge, MA: MIT Press, 1991), 743-762, which is reprinted with revisions from D. Sabia and J. Wallulis (ed.), Changing social science (Albany: SUNY Press, 1983), 73-101.

1. An adequate explanation is a true description of underlying causal factors sufficient to bring about the phenomenon in question.

2. Causal factors are things necessary, given the circumstances at hand, to the phenomenon, that is, something else would have happened without the factor.

3. Scholarly (or other) communities follow rules about the kinds and number of causal factors to be distinguished as against consigned to the background "circumstances at hand." These rules are subject to theoretical and empirical debate, and can be rejected in favor of another set of rules if practice guided by them is an inferior source of explanations for other phenomena (p. 757).

4. The hermeneutic criteria for explanation, namely, that explanation enhances the reader's "capacity to interpret the words, acts, and symbols of others in the interest of mutual understanding, and self-reflection" (p. 751), should not be applied.

5. Singular explanations are acceptable: A causal factor in one situation need not have the same effect in all other situations or even in any other situation. In other words causal factors do not have to be construable in terms of implicit covering laws or statistical regularities.

B. A series of modifications to the base-line, which capture the sense of causality and explanation used in Taylor, P. J. (1995). "Building on construction: An exploration of heterogeneous constructionism, using an analogy from psychology and a sketch from socio-economic modeling." Perspectives on Science 3(1): 66-98.
1' An explanation is an account by someone (the explainer) claiming (perhaps implicitly) to identify causal factors sufficient to bring about the phenomenon in question.

2' Causal factors are things that the explainer wants readers to see as necessary, that is, to see that something else would have happened if the factor were changed.

3' 

To distinguish among:      is to distinguish among things that:                        
a focal set of causal      could have been different and would have made a             
factors;                   difference if different;                                    
background conditions;     things upon which the focal causes are contingent, but      
and                        that are not considered changeable; and                     
incidental conditions;     things that occurred, were changeable, but that if          
                           changed would have made no or negligible difference.

1'' A sufficient composite of causes is a set of conditions each of which is seen by the readers as necessary, and additions to which either a) are not considered changeable or b) make no or a negligible difference.
Herein lies the first of several ambiguities to be noted, each providing some freedom of movement for the explainer:
Ambiguity 1. Background conditions are causes and may be included in the sufficient composite. Given 1''a) the explainer could add more background conditions or drop some out without changing the sufficiency.
Ambiguity 2. A second ambiguity is raised by the comparative method. This method identifies a contrast among the background conditions for two different situations ("background" because in each situation the condition is constant) and then claims that the difference is the cause of the difference, all other things being sufficiently similar. Or, taking situation A as the base-line, the way situation B's background conditions depart from situation A's is the cause of the outcome in situation B, even though B's background conditions are not considered changeable. So, in considering background conditions as causes a distinction is being opened up among
a) "if the factor were changed" (in 2');
b) whether it was actually changeable; and
c) whether anyone at the time and place considers/considered it changeable.
What can we do we with this ambiguity? We can accept an explanantion at the purely hypothetical level a, but we can also take it as an invitation to go back in time and identify the conditions that could have been changed (level b) that led to (i.e., caused) the background condition. And, in doing so, we have to consider if it matters to us as explainers or readers whether or not anyone considers/considered it changeable (level c).

3'' Distinguishing between background and focal conditions is a particularly active site for theorizing; here explainers attempt to define their audience around conditions that those readers are prepared to consider changeable.

4' One way of preparing readers to consider a condition changeable is if they could imagine themselves in the situation attempting to change it (corresponding to level c of Ambiguity 2), that is, if the condition marks a "site of intervention." In short, a form of hermeneutic criteria is (contra Miller) potentially applicable. Similarly, this "imaginative projection" can also inform readers' assessments of conditions being necessary; composites being sufficient; and focal causes, background conditions and incidentals being distinguished.
Ambiguity 3. 4' is implicitly an explanation of what can influence acceptance of an explanation in the social sciences, so if I want to use 4' I should support it and enlist supporters (see 1''' & 6 below).
Ambiguity 4. What are the alternatives to the imaginative projection? If it is not necessary, but only sufficient, should we use it or not?

3'''. Given that the "situation" (see 4') is specified in terms of a number of background conditions and focal causes, any analysis of the effect of a cause (including imaginative projection) operates jointly, not piecewise.
The larger the set of focal causes and number of background conditions, the more permutations and the more difficult the analysis.
The analysis can be simplified in many ways:
a) conditions are separated into background, focal and incidental (see 3');
b) background conditions are taken for granted, and simply subsumed into the analysis of the effect of the focal causes (cf. Ambiguity 2);
c) background conditions are subsumed by converting from determinations to propensities, that is, assessing the effects of the focal causes in terms of likely effects;
d) incidentals are ignored;
e) background conditions are combined into synthetic variables (e.g., the "environment");
f) the number of focal causes are minimized by shifting some or most to the background or incidental categories; and
g) the focal causes that remain are of similar kinds (e.g., relating to the level of family situation and not spanning across the economy, government, etc.).
Ambiguity 5. Simplification of complexity in these ways employs heuristics, which include heuristics about whether the audience will accept the result as a sufficient composite or find grounds to revise it (see 1'' -- to whom are incidentals negligible? -- & 3'').

1''' Explanations are strong to the extent that the explainer enlists an audience to consider the focal set of causes and accompanying background conditions to be a sufficient composite of necessary conditions.
Ambiguity 6. This strength is audience-specific and provisional. Revision of explanations is driven by:
a) competition from other composites;
b) changes in the audience and its boundaries (perhaps brought about by the explainer); and
c) counter-heuristics (e.g., shifting a background condition into the focal set).

4'' To the extent that imaginative projection operates (see 4'), an explanation is likely to be superceded if the detail of a competing composite allows the reader to imagine more intimately how the agents were acting in the given situation.

5' Singular explanations are acceptable, but the proposed causes are not singular. Instead, they are construable as heuristics about this situation being sufficiently similar to others that the audience will see how the cause could have its proposed effect. In other words, even if regularities are not necessary for explanation, a sense of "reliability" of a cause is.

1'''' The use of heuristics (see 3''' & 5'), audience specificity (see Ambiguity 6), and other Ambiguities makes explanations themselves complex phenomena. In explaining the strength of an explanation, therefore, it is difficult to reduce the strength (see 1''') to the explanation's truth (contra Miller's 1). However, heuristics can be employed to achieve this simplification (see 3''') and the result accepted by certain audiences (see 1'''). For other audiences, acceptance of the use of the truth heuristic is a social phenomenon needing explanation.

Against this background the particular explanatory style proposed in Taylor 1995 can be characterized as follows:
--It emphasizes the "sites of intervention" sense of changeability (see 4') and uses imaginative projection, but does so heuristically, that is, Ambiguities 3 & 4 are acknowledged; and it uses:
3'''' Counter-simplification heuristics that attempt to
a) keep background conditions from fading out of sight, in other words, highlight the openness of the explanation for revision (see Ambiguity 6); and
b) maintain a high number and heterogeneity of causes within the focal set.

Ambiguity 7. Three contrasts should be noted between the production of the accounts of severe depression and of the Kerang farm model:
a) the simplification heuristic of speaking in terms of propensities was employed (implicitly) for the first account but not for the second, i.e., the first case referred to a class of phenomena to be explained;
b) the linkage of causes, stressed in both cases, is presented as taking place as a process over time in the first account, but not in the second; and
c) the methodology behind the severe depression case involves reading empirical regularities as causes.

C. Extensions
6a. Given that heuristics about simplification can factor in the explainer's intended audience (see almost all the items in B), there is room for reflexivity, that is, for examining the ways the explainer produces the appearances of necessity and sufficiency.
b) Reflexivity is not a necessary component of explanation, but one that might be elicited by the audience or undertaken to preempt the audience's call for it.

7. The framework in B for social explanation also applies to natural scientific explanation, with the following modifications:
a) In natural science item 4' about what appears changeable is grounded in repeatable interventions;
b) Conversely, a condition can appear unchangeable if control can be achieved so that the focal causes can be changed while the background controlled. In this sense the causes of natural science are not causes in themselves, but causes in combination with controlled background conditions;
c) In natural science explanations tend not to become a condition involved in further explanations. For example, gravity operates independently of its self-understanding or of our representations of it. In social science there are degrees of independence, but zero dependence is not the norm.

To be continued:
situation-phenomenon
Natural science & hybrids
non-partitionability
"structure" to deal with different time scales of changeability, difficulty of the change
explanation vs. prediction
causal powers
empirical regularities as "reliables"
relate to ther views e.g., INUS
more on comparative method (see Ragin)