As a developmental step to knowledge based practice, the blog has two posts on cause, one on models and there will now be five posts on inference. This first inference post will broadly define inference as part of a knowledge based practice including the inclusiveness of the concept of cause (and thus causal inference) in all of the inferential modes; then there will be a posts on modes of inference:
Abductive inference (retroductive, hypothesis, inference to the best explanation)
The development of this component, inference, will be the **most obvious **distinguishing feature of a knowledge based practice as compared to an evidence based practice. This is simply because an evidence based practice does not address this aspect, evidence based practice was not developed as a system of thought regarding practice. EBP did not go back to build from it’s epistemological foundations (it simple built on the prevalent, highly empirical foundation (borderline logical positivism)) and it did not follow all the way through to practical reasoning. Inference is not only how empirical observations become knowledge, but it is how knowledge is implemented in practice. EBP addresses the first part of this conjunction, but not the second.
Above I emphasize “most obvious” because that is true. However, I believe the most substantive distinguishing feature is the epistemology foundation of critical realism over empiricism. It is from that foundation that the dynamic inference of practice follows. There are other distinguishing features that follow from critical realism to be developed once I fully explain the critical realist foundation (recall from my introductory posts that I will do this after articulating the purpose of the blog through its title component (cause, models and inference).
“Inference: The process of moving from (possibly provisional) acceptance of some propositions, to acceptance of others. The goal of classical logic and of classical epistemology is the codify types of inference, and to provide principles for separating good from bad inferences.” ODP
The logician Lewis Carroll (yes the same Lewis Carroll that wrote Alice in Wonderland) “raised a Zeno like problem of how a proof ever gets started.” Basically, the problem is that a proof requires not only axioms but rules of inference as well. (“Axiom: a proposition laid down as one from which we may begin; an assertion taken as fundamental, at least for the purposes of the branch of enquiry in hand.” ODP, and high school geometry). Rules of inference allow movement from the axioms. Lewis did show that it is important to distinguish the categories of axioms and rules of inference, but also that there may be choice as to whether a proposition is an axiom or a rule of inference. (More on this at another time)
All I want to point out now about “rules of inference” is that the terms - inductive, deductive, abductive - are not rules of inference. These are modes (or types) of inference and they share the same rules of inference which is based on a previously established and accepted set that dates back to Aristotle.
From the definition of inference: “The process of moving from (possibly provisional) acceptance of some propositions, to acceptance of others.”
Inductive: The process of moving from specific observations (acceptance of some propositions), to general conclusions about the way things are (acceptance of others).
Deductive: The process of moving from (possibly provisional) acceptance of some general and some specific propositions, to acceptance of specific propositions.
At this point I must add the consideration of “causal inference.” A causal inference is inferred (inductively) from a set of constant conjunctions as a causal association and therefore infers effect from cause. If cause, therefore effect. Otherwise known as an implication, cause implies effect. But if you have taken logic, you may have learned about “material implication” which establishes “implication” as a rule of inference that does not require cause. Here, I am not talking about material implication, I am talking about implication, that is the implication will all sorts of causal implications (pun intended).
In practice the great majority of our inferences are causal. To reform inductive and deductive:
Inductive: The process of moving from specific constant conjunction observations , to general conclusions about causal associations. (Note, this involves lots of assumptions, methodological, statistical, prior knowledge, consistency - we will hash these out in the next post on inductive inference). (This is what we hope to learn through empirical observations to help formulate as causal models)
Deductive: The process of moving from causal relations and some specific cause, to acceptance (or expectation) of a specific effect. (This is what we do when we are deciding on a treatment; strength training (ST) makes muscles stronger (MS); ST, therefore I expect MS)
The reason I had to introduce the causal component for abduction is that abduction looks like deduction with one very simple (yet profound) reversal. There is still a general premise about a causal association, but there is a proposition about the effect, which then infers the cause.
Abductive: The process of moving from causal relations and some specific effect, to acceptance (or expectation) of a specific cause. (This is what we do when we attempt to determine the cause of a patient’s problem. Since inflammation causes pain; and since the patient has pain (effect), I may infer that they have inflammation (cause). The subtle reversal is full of profound implications though - as we will unpack in the post on abductive inference.
I promised myself to keep these posts under 1000 words, so time to wrap up.
In this post I have defined for our purposes three modes of inference, inductive, deductive and abductive. All a critical part of a knowledge based practice, making the connection between empirical evidence, our causal models and our diagnostic and treatment decisions.
As Gries and Schneider point out: “Logic is the glue that binds together methods of reasoning, in all domains.”
Now at 980 words.