The Difficulties of Breaking Through Disciplinary Boundaries
For this week, I’ve spent a lot of time thinking about the best way to build a bridge between two fields - each with their own separate methodologies, terminologies, interests, and philosophies. As an early project of the post doc, I’m putting effort into trying to write a review paper of sorts. The idea is that I’ll highlight what psychologists can gain from NLP and what NLP researchers can gain from psychology. What this means on a more nuts-and-boltsy level is that I’m trying to think of variables which meet two criteria:
They are psychologically interesting or meaningful
They are apparent in raw text.
With an aim toward filling a third criteria as well:
- Focus on variables relevant for the values-affirmation intervention.
From a distance, this seemed like a simple enough task. Get a bunch of variables (especially ones relevant to values affirmation) -> eliminate those which are too abstract to be present in text. Et Voila! Done!
However, my ignorance was quickly exposed. In particular, I do not know enough about text, language, and NLP to know what to look for regarding point 2. For instance, believing that the use of first person pronouns vs 3rd person pronouns might be psychologically interesting and meaningful (as well as important for values-affirmation) is straightforward. Also, this could be easily spotted in text! Heck, we can even go a bit more complicated than that. We might be interested in how an individual makes attributions. This has a pretty straightforward text representation:
did some event because reason.
So, if we’re looking for attributions, we can try to determine the identity of the actor (coreference resolution), look for use of the word because and then look at the reason. Depending on what we find in the text for reason, we may realize that the pronouns used there indicate internal/external attributions (e.g. because I vs. because he vs. because it). This is a bit fancier, but still straightforward.
But, things get a little trickier with just a slight bit of psychological abstraction. What if we were interested in whether a writer had a fixed vs growth mindset? How would this be represented in their text? I suppose to find a growth mindset we might look for some combination of language indicating self-confidence, struggling against problems, and eager undertaking of tasks (I’m not sure this is true, but let’s say for the sake of argument that it is). What exactly would that language look like? We need to be very precise when we talk about this because we’ll have to quantify it at some point. Another example is relying on cognitive heuristics and the associated biases they produce. Maybe we have reason to think anchoring and adjustment is playing a role in the affirmation. How would we find that?
In some ways this is a great set of rules - it forces us to be explicit and precise when we define our variables. This is one of the hallmarks of a mature science. And psychology has a history of desperately wanting to prove to everyone that it, too, belongs on the big-kids playground. Unfortunately, many well-established psychological concepts do not have simple, precise definitions. Or rather, the operational definition can vary pretty widely depending on the context in which it is being measured.
I’ve got no solutions at the moment, only a whole bunch of question. Here’s to hoping that once I figure out the answers to these questions, they can be replaced by questions of a more sophisticated variety.