Back to the task at hand. Posts about the purpose of this blog have started with explaining the title: Cause, Models and Inference. Cause and Models are dealt with sufficiently for now. Inference remains. The last post got us started with an introduction. This post is about inductive inference, also called induction.
Induction: The process of moving from specific observations (acceptance of some propositions), to general conclusions about the way things are (acceptance of others).
Stated from an empiricist perspective and related to causal association:
Induction: The process of moving from specific constant conjunction observations to general conclusions about causal associations.
In the past post I noted that this involves a lot of assumptions related to methods, statistics (probability), prior knowledge and consistency. Also that induction is part of what we use to learn through empirical observations to help formulate as causal models.
There is a subtle point to make first. A point about induction and the move from empiricism (foundation for EBP) to critical realism as a foundation for a KBP. Induction as defined above leads to conclusions about causal associations. But a conclusion ABOUT causal associations is all empiricism can provide. For knowledge OF causal associations we need critical realism (a simplistic way to think of it is we do more with our observations than what sensory experience allows, we integrate them in a mental (rationale) process). So a study will provide a conclusion ABOUT a causal association, a systematic review will provide a conclusion ABOUT a causal association with the added benefit of replication and consistency; but reasoning provides knowledge OF a causal association. In future posts (system development) I will explain how this influences the hierarchy of evidence.
As a restatement: the most empirical observations can get us is an idea or impression of cause which is what I mean by saying we have conclusions ABOUT causal associations. We come to know causal associations through a rationale process, a mental (model) building process based on not only empirical observations. This is where critical realism provides knowledge based practice with something that evidence based practice does not have, and this influences our hierarchy of evidence.
Induction provides the raw materials for weaving a tapestry of knowledge about the way things are, the causal network, and what I call our “set of general premises.” As a set of general premises we can list if we want, our knowledge. Our set of general premises is part of our knowledge, but our knowledge is bigger than our set of general premises. We have a lot of them (here are 2):
If dropped without anything to support it an object falls (acceleration of 9.8 m/s^2) (gravity)
If there is a pressure gradient, fluid will flow from higher pressure to lower pressure (fluid is broadly defined here - things in gaseous and liquid states)
These general premises are part of our set about the way things are. We use them in reasoning about the future state of things - how things will be if I drop my egg; or if I squeeze my osprey hydration bladder and open the valve. The empirical observations that have occurred over time and incorporated to form mental models that I have a justified belief about (knowledge). I can use these mental models without further empirical evidence to make a lot of decisions, and once accepted as true I need not dwell on the evidence for that knowledge. Now, should I be presented with a new observation, a new system, a new approach that challenges the knowledge, or that creates a new situation or context for the knowledge then further observations are needed. But some things just do not need to be studied - such as this paper in the Hazardous Journeys section of BMJ: “Parachute use to prevent death and major trauma related to gravitational challenge: systematic review of randomized controlled trials.”
Assumptions related to methods, statistics (probability), prior knowledge and consistency are the typical material covered in a research methods course, reliability and validity of measures, sampling, control, randomization, statistical inference. My initial plan was to cover these all with this post - that is too much. I had started and about 2500 words into it and not being close to done, I stopped, cut it up and will deliver it over time as appropriate.
These topics (those of research methods, statistical inference) are all important for determining whether we believe the conclusions of a study are true (and no, this is NOT what the “p-value” tells us - the p-value DOES NOT tell us whether the conclusions of a study are true!!!; in case you missed that, the p-value DOES NOT tell you whether the conclusions of a study are true!!!!). This is all part of induction (and abduction - more on that later). There are issues to be dealt with here that are common to EBP and KBP - that of whether randomization is necessary; concerns about the lack of studies with actual random samples (most use convenient samples); how much control is control; whether blinding is necessary; frequentist vs. Bayesian statistical inference; etc). As I said, these are all relevant to both EBP and KBP so before I get into them on the blog, I want to make sure it is clear why they are relevant to both EBP and KBP; and what about the use of inductive conclusions is different from an EBP and a KBP approach.
The last thing to say is that induction is currently the only form of inference discussed explicitly in an evidence based practice. This is a major flaw of the EBP system of thought. Practice - as you will see - involves a dynamic interplay of induction, deduction and abduction. Yes, what we learn through induction can inform our deduction and abduction, but that process needs to be explicitly articulated and until recognized explicitly that this is happening as part of the transition from evidence to knowledge to practice then it cannot be explicitly discussed.
For those still reading along, thanks for you perseverance!