Inference to the best explanation

21 Dec, 2018 at 15:41 | Posted in Theory of Science & Methodology | 15 Comments


In a time when scientific relativism is expanding, it is important to keep up the claim for not reducing science to a pure discursive level. We have to maintain the Enlightenment tradition in which the main task of science is studying the structure of reality.

Science is made possible by the fact that there are structures that are durable and independent of our knowledge or beliefs about them. There exists a reality beyond our theories and concepts of it. Contrary to positivism, yours truly would as a critical realist argue that the main task of science is not to detect event-regularities between observed facts, but rather to identify and explain the underlying structure and forces that produce the observed events.

Given that what we are looking for is to be able to explain what is going on in the world we live in, it would — instead of building models based on logic-axiomatic, topic-neutral, context-insensitive and non-ampliative deductive reasoning, as in mainstream economic theory — be so much more fruitful and relevant to apply inference to the best explanation.

15 Comments

  1. I find ITTBE makes me uncomfortable precisely because it seems to sidestep the concerns of critical realism as it focuses on trying to categorize inference as a type of argument identified by its procedural form.
    .
    We have a well-established and elaborated idea that the procedural form of a purely logical argument can establish its logical validity. It is an idea that traces back to Euclid and his geometry, to Aristotle and his syllogism, to the unnamed progenitors of arithmetic and algebra. It is a powerful idea and it proves its power when procedural form is transferred to mechanical processes in systems of notation and machines ranging in complexity from the abacus to the digital computer, which amplify the human ability to reason.
    .
    It does not follow, however, that investigating the factual state of the world can or should be reduced to a procedural form strictly analogous to a purely logical argument.
    .
    What is accomplished by analytically reducing the processes of inference to a forced analogy with deductive reasoning and placing them side-by-side on some teacher’s blackboard?
    .
    I think it is confusing students and obscuring what matters in developing knowledge of the world as it is.
    .
    In the video, Walter Sinnott-Armstrong contrasts “explanation” and ITTBE with regard to their form and formal order, labelling “premise” and “conclusion”. He points out in passing that ITTBE does not constitute a logically valid argument: the conclusion does not follow from the premises in either explanation or ITTBE, whatever their order. Which kind of begs the question, why teach students to analyze the formal order of premise and conclusion?
    .
    The analogy with logical analysis or deductive reasoning does NOT hold. Inferential reasoning derives no logical validity and no power from its form and ordering as premises and conclusion — at least, none whatsoever from a formal ordering dictated by such a forced analogy with deductive reasoning. So, what’s the point of positing such a formal order as defining inferential reasoning?
    .
    Walter Sinnott-Armstrong allows in an aside that ITTBE is “defeasible” reasoning, which in this case is just a fancy way of saying formal order gives a superficial plausibility to a confidently presented inferential argument, but, again, without any real power.
    .
    As a pragmatist, I would be inclined to insist that we gain knowledge of the world by testing the world thru manipulation aimed at control. When we can control a process well enough by strategic manipulation of some mechanism or system to get a predictable outcome, we are justified in believing we understand what is going on. We have an operational model that works, and “truth” about the world is what “works” (in the important sense of confirming the power of our understanding of deep structures and processes). On the most basic level of developing facts, measurement is methodical(ly-controlled) observation. (And, yes, I am just fine with a geologist imagineering the processes involved and confirming by observation and measurement; I am not so literal-minded as to insist that the geologist personally build a mountain range in real time. It does help my confidence if the geologist’s reasoning can find oil or mineral deposits.)
    .
    We might reduce that model, that understanding of systemic process to “an explanation”, but “explanations” per se are mere by-products of understanding deep structures and processes. Moreover, an “explanation” is likely to be somewhat ad hoc, applying to a particular case, and in any case, a reduced form application of the base of knowledge (and only in a very limited and locally contingent sense, knowledge per se).
    .
    If I am standing in my living room and a drop of water lands on my head, and I can fix that problem by calling a roofer, than the best explanation is a hole in the roof and if I have to call a plumber, than the best explanation is, perhaps, a burst pipe or overflowing bathtub. I test the world by forming or building an operational model and use that model of the mechanisms involved to try and manipulate an observable outcome. If I have a tentative explanation, I may seek to find out if my explanatory interpretation can be extended to other observable facts; if my theory of the water dropping onto my head is a leak in the roof, I might check to see if it is or has recently rained or if I hear water running in the upstairs bathroom.
    .
    Inference in this guise is not usefully understood as an analogue with deductive reasoning. Formal procedure is less important than identifying the “substance” of particular mechanisms and probing for additional and mechanically relevant information. Deductive reasoning may be a component tool of inference (!!!), particularly in the counterfactual manipulation of a modelled mechanism (maybe what Sherlock was referring to), but what matters centrally is the specific content of the identified mechanism and its observed consistency with interpretative extension or experiment.
    .
    Knowing the law of gravity is not going to help understand the Table of the Elements and neither is going to do much to explicate evolution by natural selection or analysis of waiting times for people queued in a bank lobby. Substance over form. If there is a useful discipline to be applied, it will be in the form of something like Koch’s Postulates, criteria for determining when a hypothesized mechanism has been identified and confirmed to exist and operate in fact. By the standard of Koch’s Postulates, ITTBE seems childish and misdirective.
    .
    As a final note, I will venture that what goes wrong with statistical inference procedures and reasoning is too often a reluctance to embrace manipulation as inherent to setting up and framing a situation where probability applies. But, that’s another comment.

    • The problem with pragmatism and inference-by-control is that you might be on a movie stage, say, and the water comes from a hose held by a person. In other words, you can be able to control something and think you know why, but really you have no clue.
      .
      Epicycle theory predicted observed orbits accurately; but we now think epicycle theory wrongly describes the solar system. At the time however a helio-centric theory proposed by Aristarchus was dismissed. How do you know your theory of drop origin, or gravity, etc., is really what is going on?

      • If the source of the water drop landing on your forehead is a hose held by a stagehand, then repairing the roof or fixing the plumbing is not going stop water from landing. Empirical science, in my view, is the search for mechanism(s).
        .
        A lot of description and cataloguing might go into it, but the goal is to find mechanisms and when we discover a mechanism, it allows us to build powerful explanations and more coherent descriptions and interpretations of what we observe. We test the universe’s mechanisms as we discover them.
        .
        I do not “know” what the origin of the drop on my forehead may be until I investigate and find the mechanism that delivered the water, whether that was rain thru a hole in the roof or a hose in the hand of a stagehand or . . . .

        • You might investigate and find a leak and fix it but really a stagehand made the leak, and might do so again, arbitrarily … You might have a model that fixes the problem but is still unaware of the bigger “picture”(pun intended) …
          .
          “I do not “know” what the origin of the drop on my forehead may be until I investigate”
          .
          I argue that investigation itself relies on assumptions and though it may produce some results, ultimately you are likely still in ignorance.
          .
          You said before (I paraphrase) that if you control something, you know how it works. But if you control a slave, are your assumptions about what goes through his head necessarily accurate?

  2. “In a time when scientific relativism is expanding…”
    .
    Is there an explanation for why scientific relativism is expanding?

  3. “…….the main task of science is not to detect event-regularities between observed facts, but rather to identify and explain the underlying structure and forces that produce the observed events.”
    .
    Ideally it would be preferable to identify underlying structures.
    .
    However, sometimes science has to make do with what is available (event-regularities between observed facts), yielding what are called “effective theories”, i.e. theories which can be used to predict events in the future without understanding underlying structures.
    .
    Where has science actually produced an understanding of an underlying structure which has been durable. Take the most fundamental of phenomena, gravity. Scientific explanations begin with Newton’s Laws, which were then supplanted by General Relativity which in turn has been supplanted by Quantum Gravity.
    .
    Where does it end?
    .
    When can we be certain that the explanation truly reflects an underlying structure?
    .
    How can we be certain that the explanation truly reflects an underlying structure?

    • “How can we be certain …?” Never!
      If you want 100 % certitude, well, then you have to settle for things that are not (empirical) science, e.g. logic or mathematics 🙂

      • ““How can we be certain …?” Never!”
        .
        So, then, what, in effect, is the difference between critical realism and scientific relativism?

    • I would distinguish an analytic theory — a way of thinking thru the functioning of a system the logic of which is thought thru a priori — from an explanation, which may apply a theory to construct a narrative and operational model for a particular case with measured or observed values for parameters and variables.
      .
      The different theoretical paradigms you mentioned — Newton’s System, Einstein’s Relativity, and Quantum mechanics — are not “explanations” per se.
      .
      Critics of Newton observed that he did not try to explain why gravity worked and he agreed that limiting the scope of his theory was a key to its apparent general success in building useful explanations or interpretation of a few aspects of seemingly disparate phenomena, including the motion of planets and falling objects closer to home. Newton limited himself to applying an analysis of how mass, inertia, energy and momentum worked together to produce the observable regularity of the law of gravity and an explanation for the elliptic orbits of the planets that identified a force of gravity as the key factor. But, it is worth noting that he was proposing an analytic theory of general principles of physics and not incidentally an extension of mathematics in order to express his law of gravity. An explanation of the elliptic orbits of the planets around the sun was a specific application of the theory and the math, utilizing someone else’s work of observation and measurement.
      .
      If our understanding of the Deep structures and processes of the Universe advances, of course, we are going to be able to construct better explanations and interpretation of additional phenomena, ones that make use of more precise observation and measurement. Newton, interested as he was in optics, still had no access to precise measurement of the speed of light; he may only barely have grasped that light must have a speed, let alone that its speed might constitute a fundamental constant and limit with implications for how to think about relativity.
      .
      It seems churlish at best to complain about the progress of knowledge. It is not as if Einstein repealed the law of gravity. Gravity remains a fundamental force, even as physicists struggle to use quantum understanding to explain fundamental forces.
      .
      It seems to me that ITTBE, by outlining inference by analogy to the formal procedure of deductive reasoning, is confusing the epistemology and drawing attention to the “uncertainty” of inference without sufficient attention to the testing and methodical observation that go into constructing an explanation and convincing us that that explanation “works” well enough to be considered “best” in the particular circumstances where the explanation applies.

      Any attempt at scientific “inference” is going to employ as an essential tool, an analytic theory. That theory is going to be built using deductive reasoning or similar. Newton’s calculus or Einstein’s non-Euclidean geometry was as certain in its logical validity as its formal operations made it. No explanation is possible without a theory. Some may imagine that a clever explanation could bootstrap a theory, though I doubt it works that way, but I think the difficulty of explaining might expose the limits of a theory. That a theory has limits is to be expected, no?

      • I would agree, I have been terminologically imprecise. However, in defining “explanation” the way you have, you are giving new life to a term that isn’t necessary, to say the least. I would say you used the word more usefully/normally in the last paragraph. I do get your distinction between analytical theory and operational theory (this is akin the the term I used, “effective theory”).
        .
        I used the word explanation because I would say the three theories I nominated have, in varying degrees, characteristics of both analytical and operational theories.
        .
        I am not sure if you are saying I was complaining about the progress of knowledge. In any event, I was not complaining, merely making the observation that scientific theories are wanting in durability.
        .
        Certainly theories have limits to their “truthiness”. All theories are useful until invalidated by observation or supplanted by another theory.
        .
        Personally, I don’t have much trouble with ITTBE even if based on scanty observation, as long as all available and relevant evidence is considered. And it is OK that new evidence may skew the explanation in another direction.

  4. “How can we be certain that the explanation truly reflects an underlying structure?”
    .
    Perhaps this need to ask this question explains why science is reduced to relativism.
    .
    The exemplar historical development of theories of gravity reinforce the notion, perhaps, that we can never “know” that an explanation truly reflects an underlying structure.

  5. “Science is made possible by the fact that there are structures that are durable and independent of our knowledge or beliefs about them. There exists a reality beyond our theories and concepts of it.”
    .
    Isn’t that just your story?
    .
    [P.S. I recognize that video from the MOOC it’s from, which I took. Pretty good class.]

    • The MOOC also presented the Problem of Infinite Regress, which makes everything scientific just stories again …

  6. I do not understand your point about “durability”. In my estimation, Newton’s discovery of gravity endures, just as Darwin’s discovery of natural selection endures. That general relativity expanded the conception of how gravity works as a force did not overthrow Newton, any more than genetics overthrew Darwin.

    • Newton’s law of gravity survives as “effective theory” within its scope, that is it can characterize certain phenomena at a sufficient level of accuracy for practical purposes but not all, for instance, classically, the precession of the perihelion of Mercury. General relativity is able to do this. And not being a physicist, I can only presume that quantum gravity can explain phenomena which Newton’s laws/general relativity cannot.

      I understand that not even Darwin’s theory is universally accepted in all it’s aspects.

      My point about durability is related to Lars point about the need to detect/define/understand/explain underlying structures. Science is ever in a state of flux and iteration. This being so, how can we know with certainty that the latest theory is the ultimate and these evasive underlying structures have been captured.


Sorry, the comment form is closed at this time.

Blog at WordPress.com.
Entries and comments feeds.