A central task in many of the sciences is to explain. The proper account of scientific explanation is still up for debate. What are the proper objects of explanation: facts, events, or both? Can we explain laws and necessary truths? Are mathematical explanations possible? If x is explained in terms of y, what must we know about y in order for the explanation to count as increasing our understanding of the phenomenon in question? Without being able to give an exhaustive account of scientific explanation, most of us have an intuition about what explanations are and what a request for an explanation looks likes. (Why does the pitch of a siren rise when it gets closer? How does a person contract malaria? Why did the average price of gasoline fall last month?)
It seems that an explanation is one kind of thing, given that all explanations share a name. But if we look to the natural and social sciences we find explanations that look quite different. What scientists call "explanations" differ with respect to the form and structure of the explanation and (possibly) with respect to the information given. An explanation can be a picture, an equation, a mechanical model (like an orrery), an argument, a story, merely citing a law, etc.
The popular approach to scientific explanation is to treat all successful explanation as either giving information about the relevant cause or causes of the phenomenon to be explained or giving information about something that is "entangled" with or stands in for the relevant cause (Salmon 1984; Lewis 1986; Strevens 2008; Skow 2014). Given that in many sciences there are explanations that refer explicitly to the function of a phenomenon and not its cause, we should ask: are functional explanations just another way of giving causal information, or are they noncausal?
To illustrate: Friedrich Engels explains the origin of the modern state by referring to its function in moderating class antagonism and protecting “the possessing class” from “the nonpossessing class.” Michel Foucault explains the origin of modern disciplinary practices by referring to their function in (in part) "making useful individuals." Roy Rappaport explains the religious behavior of the mid-twentieth century Tsembaga of New Guinea by referring to their ritual cycle’s function in maintaining the local biodiversity, limiting fighting, adjusting boarders and populations, and facilitating trade, epigamic and demographic displays, and the distribution of protein-rich pork.
The question then is whether explanation by citing the function the phenomenon has for a society and explanation by citing the relevant cause of the phenomenon are the same thing. Can the function (or purpose) that a social phenomenon serves cause that phenomenon to exist?
The crude answer is that functions are not causes because the function that a phenomenon has is an effect of that phenomenon: i.e., it arises after the phenomenon comes into existence. And something that exists later in time cannot cause something earlier in time. But this is false with respect to artifacts that are created to serve a purpose by human intention, like toothbrushes and telephones.
Do long-lasting and large-scale social phenomena (like religious rituals, disciplinary mechanisms of an entire civilization, and the modern state), stem from human intention? Given the large number of agents and institutions involved in the particular social phenomenon, the compounding effects of misperformance and intentional altering, and the time scale over which the phenomenon lasts (at least several generations), it is unlikely that an intentional agent or group of agents could purposefully create these social phenomena in their entirety.
Furthermore, there are social phenomena in which individuals intentionally engage, but that nevertheless serve a function different from what they believe. The Tsembaga say that “they perform [their] rituals in order to rearrange their relationships with the supernatural world." And yet, Rappaport leaves this belief out of his functional explanation of their ritual cycle. The actual functions of the ritual cycle are unintentional.
How can a social phenomenon exist in order to serve a specific function if there is no intentional agent behind it? Does this mean that social behaviors, policies, and institutions exist for no reason; that all functional explanations of social phenomena are false?
We must distinguish two things: the fact that something has a function in the society and the fact that something exists in order to serve some function in the society. It is possible that a social phenomenon is caused to exist, not by the function that it has, but by other events, and yet the phenomenon may still have a function in the society; and that knowing this function helps us understand the phenomenon. Modern disciplinary mechanisms and practices were slowly fashioned in response to particular events in history (plagues, wars, disasters, etc.), but for Foucault it remains the case that these modern practices have the function of “[increasing] the possible utility of individuals,” and that this function is informative.
The problem is that if the function is not given some distinguishing mark (say, by being the relevant cause of the phenomenon, by being what an intentional agent created the phenomenon to do), we cannot distinguish the function of a social phenomenon from a mere consequence. Is it that the telephone exists in order to facilitate long-distance communication? Or do telephones exist in order to sit in landfills? (Assuming that there are more telephones in landfills than those currently used.) What makes the one explanation informative and the other not? If the success of a functional explanation is not judged by the criteria of a causal explanation, we are left needing some way to distinguish a successful from a fallacious functional explanation. If there is no way to distinguish a function from a mere consequence, then we cannot judge whether a putative functional explanation of a social institution is informative.
The Norwegian philosopher Jon Elster in “Functional Explanation: In Social Science,” gives an alternative account of how functional explanations are causal. As we saw above, there are cases where a phenomenon is caused to exist in order to serve a function—where an intentional agent creates the phenomenon for some purpose. But barring God or a "World Spirit," it is unlikely that a large-scale social phenomenon is caused by an intentional agent or agents. A second way that a function can cause a phenomenon to exist is through what Elster calls a "feedback loop." A feedback loop is created when the effect of a cause contributes to the persistence of the cause. The actions of the gardener cause the carrots to grow and the carrots, when eaten, give energy back to the gardener to continue to grow more carrots.
To return to our examples, the ritual behavior of the Tsembaga functions to maintain the human and nonhuman environment. This maintained environment in turn causes the Tsembaga and their rituals to continue to exist. A functional explanation is successful when the function referred to is a cause of the continued existence of the phenomenon. A functional explanation is successful when we disclose the causal mechanism linking the function (the explanans) to the social phenomenon (the explanandum). Elster’s feedback-loop approach answers our leading question: functional explanations are causal explanations where the cause of a phenomenon is its function.
The merit of Elster’s feedback-loop account is twofold. It gives us a way of distinguishing functions from mere consequences; that is, it distinguishes explanatory from nonexplanatory consequences or effects. The effect of a phenomenon that causes it to continue to exist through a feedback loop is its function. Second, it gives us a causal account of functional explanations without appealing to an intentional agent.
For all its advantages, Elster’s account runs into at least three problems. First, there are consequences of phenomena that cause the phenomena to persist through a feedback loop but nevertheless fail to qualify as functions. Imagine that there is a small stone holding up a large stone in a running stream. If the small stone wasn’t holding up the large stone, it would be washed away. So the fact that it holds up the large stone explains why the small stone is there. This is a case where a consequence of a state of affairs causes that state of affairs to continue to persist, and so an Elster-type functional explanation is appropriate; nevertheless, the case seems to be giving us inconsequential information about the small stone, not explaining why it is located where it is.
Second, the functional explanation of a phenomenon will always be incomplete. We need to add to our explanation the proximate causes of the phenomenon in the first place. A lesson learned from the crude response above is that on the traditional theory of causation, an effect cannot precede its cause. A feedback-loop analysis of functional explanation can only explain why a social phenomenon continues to exist, or persists; it cannot explain why or how the phenomenon first came into existence. Peter Godfrey-Smith sees this distinction in the explanation of a biological trait. There is an evolutionary explanation of the trait and a functional explanation. This bifurcation is significant because a straightforward causal explanation of why a phenomenon occurs can be a complete explanation, a functional explanation can never be a complete explanation. A functional explanation can give information about what causes the phenomenon to persist, but not about what causes it to come into existence.
Third, if we stay close to what the social scientists and philosophers say, they rarely talk about causes. Is it that they assume that the functions they cite are causes of the social phenomenon? Or is it that the information grasped when you understand a functional explanation of a social phenomenon is different from causal information? Among others, the Verstehen tradition of Max Weber and Clifford Geertz and the hermeneutic approach to the social sciences of Charles Taylor argue that causal information is not sufficient for understanding social phenomena. Perhaps understanding the positive and negative contributions of a social behavior or institution to a society yields scientific understanding regardless of whether the contribution or function cited is a cause of the phenomenon. But without relying on causation, where do we draw the line between informative functions and inconsequential effects?
References
Lewis, David. "Causal explanation." Philosophical Papers Volume II (1986): 214–240.
Salmon, Wesley C. Scientific Explanation and the Causal Structure of the World. Princeton, N.J: Princeton University Press, 1984.
Skow, Bradford. "Are There Non-Causal Explanations (of Particular Events)?" The British Journal for the Philosophy of Science 65.3 (2014): 445–467.
Strevens, Michael. Depth: An Account of Scientific Explanation. Cambridge, Mass: Harvard University Press, 2008.
Thomas Morrison has an M.A. in philosophy from the University of Chicago. His main area of research is the philosophy of science, including the social sciences and psychology. He currently lives in Lawrence, KS, where he continues to write philosophy and work on his beef stew recipe.
Very interesting and entertaining essay.
A bit surprised the behavioral psychology wasn’t mentioned as part of the functional social sciences as it’s part of the core of determining cause and function for applied work. Looking at the clinical work, a functional analysis as conducted in applied psychology uses “function” as the consequence of a behavior that maintains it, whereas the cause is seen as the antecedent or what triggered a behavior. Or in probably more familiar terms, its’ the basics of reinforcement. Antecedent events (among other things) causes a behavior to occur, the consequence selects the behavior and changes the frequency of future behavior. This seems very similar to the feedback loop of Elster’s.
I also agree that the functional explanation in some sense is incomplete but I think that ultimately comes down to what the goal of your analysis is as there can be many different levels of proximate cause. In psychology, we can think of proximate causes as the event the immediately preceded the behavior but we can also think of the physiological proximal cause of hormones triggered, neurons fired, etc.
For example, a functional explanation for an applied psychologists looking to change a child’s behavior can be complete by knowing the proximate cause in terms of what happened before the behavior and the function by what followed the behavior. However, the proximate cause of physiological responses is not valuable in this context (but could be depending on the case).
Without a goal criteria for our inquiry, we could chase our explanatory account endlessly without perceivable stopping point that we can consider giving us a complete account.
“Without relying on causation, where do we draw the line between informative functions and inconsequential effects?”
I love the scientific method for attempts to explain things:
1) Define a question 2) Gather information and resources (observe)m 3)Form an explanatory hypothesis 4)Test the hypothesis by performing an experiment that controls for variables and collecting data in a reproducible manner 5)Analyze the data 6)Interpret the data and draw conclusions that serve as a starting point for new hypothesis 7) Publish results 8)Retest (frequently done by other scientists) perhaps a metanalysis of similar experiments.
A social scientist never “proves” causally anything by an experiment, but only that the results are less than .05/.01 probability of being false (null hypothesis testing).
Even the new results of the pentaquark are based on a 9 sigma error probability.
We are thus left with testing which provides probabilities, not causation. The causation is left to speculation/interpretation of the data. True, many still confuse probability with causation, whether scientists or not.
Great exposition and evaluation. Would love a write-up on functional explanation in physical science. I have some thoughts on the topic especially around Dewey and Poincare/structural realism. Would be cool to collaborate. Hit me up on Twitter if you’re interested.