1. All the evidence we have as scientific psychologists is publically observable behavioral evidence.
2. The safest epistemic strategy is to limit as much as possible going “beyond” the evidence, an inevitably risky gambit.
3. “Subjectivity”, “mental states”, “cognition”, “representations”, “feelings”, “consciousness”, “awareness”, “experience”, etc. are not publically observable i.e. if you open up someone else’s skull you will not see cognitions or mental states, you will see a pulsating hunk of flesh.
3. If we value epistemic safety above all, we should never leap beyond behavioral evidence to talk about unobservable mental states, unless such talk is self-consciously understood to be an abbreviated paraphrase of a long conjunction of behavior reports. Therefore,
4. The safest epistemological stance in psychology is behaviorism.
But wait! Don’t scientists in other fields go “beyond” raw data by talking about “unobservable” theoretical entities like atoms and black holes? If it’s epistemically warranted for physicists to appeal to “unobservable” theoretical entities like atoms in order to explain the experimental data, then it should also be okay for psychologists to appeal to “unobservable” theoretical entities like “episodic memory” or “engrams” in order to explain the behavioral data.
Two things can be said in defense of behaviorism.
First, it’s an open question in the philosophy of science whether physicists are in fact epistemically warranted to go “beyond” the data. According to physicist-philosopher-of-science Duhem, theories are only supposed to be tidy and convenient summaries or compressed descriptions of experimental findings, not statements literally describing an unobserved metaphysical reality. On this view, we do not use theoretical entities and equations to explain the data but rather use equations and theories to help us cope with the large and unwieldy collection of facts gathered by experimenters. Duhem argues that if humans didn’t have such finite memories, scientists would not find it necessary to tidily represent messy experimental findings in terms of neat equations and law-like statements.
Consider this: If a theory about domain X is true, then all possible experimental findings relevant to domain X deductively follow from the theory and thus have the same truth-value as a long conjunction of descriptive reports of real scientific experiments. But once you have all the experimental findings on your head, what’s the need for the theory? The need is purely practical, a result of human finitude and our desire for convenience, simplicity, and genuine understanding.
Second, even if we grant physical scientists an epistemic license to go “beyond” the data and talk about theoretical entities, this practice only works well when there are widespread conventions in place for operationalizing theoretical terms (i.e. translating theory into real experimental operations) as well as standards for conducting and verifying results of experimental procedures (measurement verification procedures). It’s not clear to me that cognitive science has reached any widespread consensus on any of these issues.
Compared to “mature” sciences like thermometry with widespread industry standards, there seems to be little if any widespread consensus in the “mind sciences” about theoretical terminology let alone operational criteria for testing theoretical claims or even nailing down what exactly it is we are supposed to be studying in the first place.
Thus, the true problem with psychology is not that it talks about “unobservable” entities or employs theoretical jargon but rather there is no widespread consensus on how to define our concepts and operationalize our methods for getting access to the unobservable phenomena.
The problem facing psychology is two-fold: (1) a lack of consensus on how to pick out the phenomena due to a lack of theoretical consensus in understanding the ostensive definition of a psychological concept and, (2) a lack of consensus on how to interpret the evidence once we have collected it.
Case in point: recent developments in the “science” of consciousness. First, there is there little to no consensus on where to even look for consciousness to begin the process of measuring it and studying it as a natural phenomena. Can any theorist answer this simple question: should we look for consciousness in insects?
Some theorists think if we looked for it in insects, we will find it because on their definition “consciousness” is not that fancy of a phenomena (e.g. enactivists and neo-panpsychists would both predict a consciousness-meter would register a small amount of consciousness in insects). According to other theorists, if you looked for consciousness in insects you will not find it because on their preferred definition “consciousness” is fancy and thus probably found only in “higher” animals like mammals. Who is right? No appeal to empirical facts will help in this debate because the problem is fundamentally about how to interpret the evidence given all we can go on as psychologists is behavior, which is of course neutral between rival theories of consciousness.
Some might object that I have picked an easy target and that the science of consciousness is a bad example of how psychology in general is done because it is the newest and most immature of the psychological sciences. But in my humble opinion, the science of consciousness is on no worse footing than most other subfields and niches of psychology, which are continually making progress “towards” various grand theories. However, insofar as another subfield of psychology is on firmer ground than consciousness studies, it will be because they have imitated the physical sciences by operationalizing their theoretical concepts in terms that can be directly measured by physical instruments. That is, a subfield of psychology is on firmer epistemic ground insofar as it sticks closer to physical, behavioral evidence, which is all any psychologist has to go on in the end. This is close enough to behaviorism for me.
I have much more to say on this topic, but I promised to be quick and dirty. Remaining questions include: how should we define the observable vs unobservable distinction?