Creating Explanations
Sources of explanations in practice can be very diverse, once you take on the role of a researcher:
- You can create an explanation yourself: you notice an unresolved problem in the world (indicated by surprise, interest, "wow," etc.); you see that there is no satisfactory explanation that corresponds to it; you build a cause-and-effect model; you confirm or refute it; and work with its parts.
- You can get an explanation from other people, and then first you update the epistemic status of these people (that is, understand what they think about the world), and only then, if you trust them, incorporate some part of the world model -- if the explanation fits into it based on assumptions, and so on.
In any case, the practice of creating explanations consists of the following steps:
- Problem statement
- Listing hypotheses
- Pre-experimental argumentation
- Listing observations/setting up the experiment
Other people may provide you with not only explanations but also any other descriptions: individual observations (which will prompt you to reconsider a piece of the world picture), predictions that will make you think about which explanation this prediction can follow from (which, in turn, may shake your own explanation), and so on.
Incorrect Understanding of the Explanation Creation Process
There is a fairly common misconception: we observe the world; we generalize the observations; we produce explanations from all of this. It actually doesn’t work that way -- it is impossible to derive explanations from observations because there is no object of observation. Until there is an object of study and a problem statement, an identification of something unexplained in the world, we will not observe it to create explanations. The attempt to explain occurs earlier than observation or experiment (data collection) and starts with a guess.
Model Process
We will look at the model process of creating an explanation, and then we will search for its components everywhere -- and when not finding them, we will know what we are missing.
- Problem Statement
A problem (research question) arises that we want to solve. Or a question we want to answer. Often this happens like this: we think about something, assuming it works in a certain way, and suddenly find out the opposite. This is the moment of surprise. We may also notice -- "catch" our attention -- a problem without being especially surprised.
After that, you intuitively pick out important objects from the background to solve it, or even just their contours. Fine prognostic ability is involved -- at every moment, our brain predicts what will happen in a fraction of a second; what we may need in a fraction of a second.
- Listing Hypotheses
When the research problem [task] is recognized and important objects for its solution are identified, a hypothesis of explanation, or guess, begins to emerge.
All past explanations you knew converge here.
Here, you ask those around you how they think X works, and critically analyze their responses, comparing them with the observations that led you to engage with this problem in the first place.
- Pre-experimental Argumentation
Next, the most interesting part starts.
Explanations can be epistemically informally assessed, i.e., before obtaining data, without direct verification, without an experiment. For the last hundred years, humanity has preached the doctrine of holy data: "let's gather a lot of data and then derive some theories from them!" It is costly.
Before collecting data, you need to have at least two competing explanations so that the data help you differentiate between them.
Explanations are never created from scratch; they always defeat the previous ones, fill the gaps in them, new explanations replace old ones[1].
Essentially, you will need to demonstrate how the new explanation is better than the previous one, what problems of the old explanation it solves -- that's it. If you think the area is completely new and you know nothing about the area in which the explanation is being created, it is almost always not the case there -- you must have had some intuitive notions and general considerations, and you could probably roughly explain phenomena in that area.
Further in the process of pre-experimental argumentation, you will need to convince others of the validity of your explanation.
- Listing Observations
And now you start observing (or even experimenting and observing) to find out if the new explanation is indeed better than the old one or if it just seemed that way. It is important to try to refute your hypothesis, not to confirm it, to avoid falling victim to confirmation bias -- the tendency of human attention to ignore evidence that contradicts its view.
In short, the process of creating explanations:
The chain goes like this: you live -- something happens that makes you notice a problem -- you automatically begin to pick out objects related to it from the background -- you build a hypothesis based on these objects and what happened -- and only then do you begin to observe.
During the execution of this process, you may encounter internal problems -- then the same process will occur with them inside the already existing larger process. You will return to it later. Or not return -- anything can happen.
This all happens very quickly and not very consciously. You hardly keep a diary of investigating processes in your brain. But it would be useful -- at least in professional activities.
In every work, there is an element of surprise -- hence, a component of unsolved research problems with a mismatch between the world picture and reality.
If you often experience things like this: "I did not expect him to do that," or like this: "I could not anticipate that it would turn out this way," then your world picture does not align with reality in that place. This, in turn, means that your explanatory model in that place is poor -- it did not give you correct predictions. And this can be confidently marked as a problem and passed through the cycle described above.
More on this is written by D. Deutsch in the book “The Structure of Reality” ↩︎