A framework for inter-level explanations: Outlines for a new explanatory pluralism
Introduction
This article is about inter-level explanations. Because these terms can be used in a variety of ways, let me start by making some clarifications. First, I shall assume what I call a scientific disciplines account of levels. That is, with ‘level’ I mean a level of description, or a specific grain size, through which phenomena are studied and explained, not levels in nature.1 Thus, when I speak of an inter-level explanation, I mean that the explanandum and the explanans are couched in different scientific vocabularies. Obviously, on this account an intra-level explanation is an explanation where the explanandum and the explanans are couched in the same scientific vocabulary.
Second, I shall restrict myself to situations in which multiple explanations address the same explanandum, with at least one of these being inter-level with respect to that explanandum. Third, although formulated like this, the issue is neutral with regards to whether the inter-level explanation is at a higher or a lower level than the explanandum, I will mainly focus on the latter type of situation. Thus, to illustrate these three points, an example of the type of situation I am concerned with is when a psychological state like ‘being depressed’, is simultaneously addressed by an explanation couched in psychological terms (e.g. feelings of neglect during childhood) and a neurophysiological explanation (e.g. reduced serotonin levels in the prefrontal cortex). From now on, I shall use the phrase ‘multiple inter-level explanations’ as shorthand expression for situations like this. Finally, this paper deals with causal explanations. I do not wish to suggest that non-causal (e.g. mathematical) explanations do not exist, but this article is not about them.
In any case, when we have multiple inter-level explanations, which one should we go for?2 Do we prefer the explanation that addresses the target phenomenon at its own level, or do we prefer a lower-level explanation? Of course, if we maintain that one of these levels is somehow privileged over the other, we end up with positions such as reductionism or eliminativism. If, on the other hand, we believe that there is no privileged status for any level, then we might opt for some form of methodological dualism. Alternatively, we might say that these relations apply only locally. That is, though there might be cases of reduction or elimination, these relations as such do not represent essential ingredients of scientific progress. This latter position has become known as explanatory pluralism (McCauley, 1996, McCauley, 2007, McCauley and Bechtel, 2001).
Of course, the history of science testifies to successful local reductions and eliminations, and these are not ruled out by explanatory pluralism. The point is that as the grand, sweeping theories such as classic reductionism and methodological dualism fell out of favor, philosophers increasingly realized that explanations of different grain sizes coexist and exert mutual influence on each other. Rather than on philosophical ideology, the choice of grain size depends on contextual or pragmatic factors.
It would be unfortunate however, if this is where the theorizing stops, i.e. if the claim that pragmatic factors determine grain size is viewed as a terminus for the philosophical debate about inter-level explanations.3 In my view, this picture is at best incomplete. While it is true that pragmatic considerations play an important role in selecting or emphasizing a particular explanatory level, there are also other, non-pragmatic ontological factors at work.4 However, their influence is not deterministic: although it puts constraints on our choices and preferences, within those constraints, there is ample room for pragmatic considerations to come into play. Moreover, the fact that pragmatic factors play this role does not mean that anything goes: as we shall see, it is possible to draw up guidelines to help us understand just what pragmatic factors are at work, and how they lead us to emphasize one explanatory level over the other. Thus, although the explanatory pluralist is right in stating that reduction, elimination etc. apply only locally, it does not follow that they apply randomly. To make sense of the choices scientists make when confronted with multiple inter-level explanations, we need to understand the subtle interplay of pragmatic and ontological factors that influence these choices.
In this article, I offer a framework for classifying both pragmatic and non-pragmatic factors influencing our preferences for particular levels when providing or pursuing causal explanations. I will argue that although this framework stays true to the basic tenets of explanatory pluralism, it represents a step forward, in that it identifies situations in which ontological factors constrain the influence our pragmatic interests have, and offers guidelines that help us to understand the choices made by scientists. Although the result will be a general conceptual framework, rather than a finished product, it does clearly point to ways to augment and refine it with subsequent research. As such, the framework presented at the end of this article has a programmatic character.
Here is an outline of the paper. First, I will briefly present what I take to be the central claims of explanatory pluralism (Section 2). Next, I will introduce a distinction between situations in which multiple inter-level explanations cite one underlying causal process, and situations in which they cite multiple, genuinely distinct causal processes (Section 3). While the debate between explanatory pluralism and reductionism (of various sorts) seems to mostly focus on the former type of situation, in this article, I will concern myself with the latter type. Combining the previous material with a distinction between relevant causal factors and productive causes, I draw up a preliminary taxonomy of types of relations that can obtain between multiple inter-level explanations (Section 4). I then introduce some pragmatic factors that influence our choices to emphasize one explanatory level over another (Section 5) and draw on these factors to construct three guidelines that help us to make such choices, and understand the ones made by scientists (Section 6). I will illustrate how these guidelines work in practice by considering a case study, namely multiple inter-level explanations of the Korsakoff syndrome (Section 7).
Next (Section 8), I shall present my conceptual framework. I will contrast it with standard explanatory pluralism, and argue that although it stays true to the general spirit of explanatory pluralism, it does suggest a more nuanced picture of the ways multiple inter-level explanations can relate. I will end by considering some ways the framework might be expanded in the future.
Section snippets
Explanatory pluralism
In contrast to traditional reductionism or eliminativism, explanatory pluralism acknowledges the multiplicity of explanation. Explanations at different levels can exhibit anything from reduction to mutual co-evolution, from elimination to integration. Rather than strict ontological commitments, the identities postulated between the entities of lower- and higher-level explanations are at best heuristic and hypothetical in character, open to revision or abandonment as the need arises (McCauley &
How many causes?
As I have said in the introduction, I am concerned with cases of multiple inter-level explanations. However, this phraseology hides an ambiguity that, although by no means controversial, is not usually noted in the context of comparing explanatory levels.
Sometimes, two explanations of different grain sizes will refer to same underlying causal process. This is the situation we are naturally inclined to consider in the context of positions such as reductionism and methodological dualism. For
Relevant causal factors versus productive causes
One may explain something by referring to causally relevant factors, or to a productive cause of the explanandum. Suppose we say that the window shattered because it was hit by a baseball. In such a scenario, the baseball's hardness may be a causally relevant factor in the sense that had the baseball's hardness been different, the window would not have shattered. Similarly, if we say the baseball being white was not a causally relevant factor, we mean that if it were another color, the window
Two pragmatic factors explicated
In this section, I will explicate two pragmatic factors. In Section 6, I will draw upon these to formulate some heuristic guidelines that help us to decide on which explanatory level we should focus when confronted with two inter-level explanations that describe two distinct causal factors. The first pragmatic factor concerns the epistemic interests underlying our explanations, the second proximity of causes.
Combining epistemic interests and proximity of causes into heuristic guidelines
How do epistemic interests and proximity of causes bear upon the issue of grain size? One reason someone might choose to emphasize the most proximate cause is that it allows for more reliable interventions. Again, this was already recognized by Mill:
Thus, in our former example, the law that the contact of an object causes a change in the state of the nerve, is more general than the law that contact with an object causes sensation, since, for aught we know, the change in the nerve may equally
Applying the guidelines: the case of Korsakoff syndrome
Korsakoff syndrome is a memory disorder, the symptoms of which include anterograde amnesia (the impairment of the function to create new memories), retrograde amnesia (the impairment of the function to recall events prior to the onset of the disease), and confabulation (invention of false memories). A strongly related disorder is Wernicke's encephalopathy, a potentially lethal disease which has the following ‘classic’ symptoms: oculomotor abnormalities, ataxia, and global confusion (Wernicke,
A framework for understanding inter-level explanation
We are now in a position to explicate a framework for studying inter-level explanations. When confronted with a choice between two inter-level explanations, or between emphasizing one or another part of a compound explanation, the framework presented in Fig. 1 may help one decide.
As one can see, the second tier distinguishes between cases of two inter-level explanations citing one or two causes, the third between relevant causal factors and productive causes. The domain of inter-explanatory
Conclusion
Admittedly, the framework presented in this article is very general, and as I have said in the introduction, its character should be interpreted as programmatic. Ultimately, its strength (how much of an improvement it is over standard explanatory pluralism) depends on the degree to which it can be further developed. In particular, rules to decide between two inter-level explanations citing one cause (both causally relevant factors and productive causes) would be a welcome addition. Furthermore,
Acknowledgments
The research for this paper was supported by the Research Fund Flanders (FWO) through project nr. G.0031.09.
References (33)
Thiamine administration in alcohol-dependent patients
Alcohol & Alcoholism
(2005)Reducing psychology while maintaining its autonomy via mechanistic explanations
- et al.
Multiple realizability revisited: Linking cognitive and neural states
Philosophy of Science
(1999) - et al.
Discovering complexity: Decomposition and localization as strategies in scientific research
(1993) Psychoneural reduction: The new wave
(1998)Philosophy and neuroscience: A ruthlessly reductive account
(2003)Explaining the brain
(2007)Explaining technical change. A case study in the philosophy of science
(1983)Mechanisms, causes, and the layered model of the world
Philosophy and Phenomenological Research
(2010)How to build a theory in cognitive science
(1996)
Wernicke's encephalopathy: A more common disease than realised
Journal of Neurology, Neurosurgery, and Psychiatry
An international perspective on the prevalence of the Wernicke-Korsakoff syndrome
Metabolic Brain Disease
Clinical signs in the Wernicke-Korsakoff complex: A retrospective analysis of 131 cases diagnosed at necropsy
Journal of Neurology, Neurosurgery, and Psychiatry
Program explanation: A general perspective
Analysis
Science, truth, and democracy
Is there a significant distinction between cognitive and social values?
Cited by (4)
Social intelligence: How to integrate research? A mechanistic perspective
2019, AI and SocietyExplanatory pluralism
2016, Explanatory PluralismExplanatory pluralism and the (dis)unity of science: The Argument from Incompatible Counterfactual Consequences
2016, Frontiers in PsychiatrySocial intelligence: How to integrate research? A mechanistic perspective
2014, CEUR Workshop Proceedings