Responses to Comments on Psychology No Big Idea

Yes, psychology will be a real science some day

A while ago, I published an essay called Why Psychology Hasn’t Had a Big New Idea in Decades. This was inspired by my experience teaching a class about psychological paradigms at Hampshire College, and by thinking about this problem for the majority of my adult life. 

My basic argument was that psychology isn’t making progress because it doesn’t have a paradigm, and we can’t make any real progress until we have one. I spend the body of the essay describing what we know about how sciences get their first paradigm, and I close it out by listing some directions that psychology might pursue next, mostly as inspired by discussions I had with my students.

Despite it being so dang long, I got a number of thought-provoking comments on this post, and I wanted to respond to a few of them here. Here goes.

---

Colin M. Fisher writes: 

I’m with you and Adam Mastroianni that psychology has only proto-paradigms. But I think the comparison to chemistry takes us awry. Even if we have a stronger paradigm (and I think its stronger than it was), we still won’t be like chemistry. Because people are changed by the awareness of their own psychology. And those situations are human creations. So we’ll always be shooting at a moving target, paradigm or no. And that will forever keep us from being like chemistry.

 I really don’t get this argument. 

Compare: The chemicals we use are changed by our awareness of chemistry. As we learn more, we make new chemicals that are human creations. So we’ll always be shooting at a moving target, which will forever keep chemistry from becoming like chemistry. Huh? 

My best guess about the disconnect here is that Colin might see psychology as the study of specific people and specific situations. 

And yeah, if you think of it this way, it’s easy to see why it might feel endless. Specific people and specific situations really are moving targets. How do people behave at the grocery store? How do people behave at dinner? How do people behave at the movies? Look out, there are a lot of different situations. Just when you think you’re getting a handle on things, someone goes and invents a new situation, “brunch”, and you have to start a whole new line of research. Then the pandemic hits and you have to learn how people behave in a “zoom room”. When you see the job of psychology as enumerating these million million things, it can seem kind of hopeless!

But I see psychology instead as the study of the nature of people and situations, and the laws that govern them. I think it’s possible to ask questions like: what do dinner and brunch have in common? Why do people talk at dinner but not usually at the movies? Why do people mostly not talk at the movies, but sometimes they do? Where do the exceptions come from? When “zoom room” is invented as a new situation, we don’t have to approach it as something totally new. It clearly has some things in common with other situations. 

Taking Colin’s concerns as stated, if people are changed by the awareness of their own psychology, we can still ask how they are changed by the awareness of their own psychology. If situations are human creations, what kinds of situations can exist, how do situations differ from each other? Why do humans create some situations and not others? Are there exotic situations that we could be creating, but that no one has yet discovered? Can I try them?

“Paradigm” is a famously hard concept to define, but one gloss is that it can be read as, “way to talk about [whatever]”. I think most of us agree that we have better ways to talk about psychology today than we did 100 years ago, when professionals still used terms like “neurasthenia” and “hysteria”. Colin seems to agree: he says, “it’s stronger than it was”. 

I think there are going to be even better ways to talk about psychology in the future. And I think chemistry is one of the right models to look at, because of how huge their subject matter seemed at first, and how much success they’ve had at breaking it down into manageable parts. When psychology gets a strong enough paradigm, we will be like chemistry.

The fact that psychology seems so impossibly complex and nuanced makes this a good comparison, not a bad one — chemistry also seemed impossibly complex and nuanced, until we found a paradigm that let us break down the confusion into individual parts.

Chemistry started with the goal of understanding every possible kind of matter in the universe. This seems like a pretty tough ask — even if you limit yourself to just, say, kinds of mud, there’s a lot of variation. How are you supposed to find a way to describe the difference between every different kind of mud?

But chemists eventually found that every material thing (not just mud) is made of combinations of only about 90 different kinds of things: the 90-ish naturally occurring elements. Then we learned that all the elements are actually made up of only three different things — matter is just different combinations of protons, electrons, and neutrons.

So, what pieces are the mind made of? Maybe it’s more than 3 things, maybe it’s more than 90 things, but chemistry has a pretty clear message for us. When you stare at something for long enough, and refuse to give up on the question, “what pieces is it made of?”, eventually you get somewhere. You can start with “every kind of material body in the universe” and then discover that it’s all made up of just 90 or so things. And then you can discover that those 90 or so things are made up of just three things in different combinations. The mind is complicated, but I’m not sure it’s as vast a problem as “the nature of every substance in the universe”. This makes me optimistic.

And as a parting example, “situations” may be 100% the wrong way of framing these questions. In fact, I strongly suspect it’s the wrong approach, something we inherited from a failed attempt at a paradigm (one that I hope to write more about soon!). For someone like Colin, tossing out the concept of “situation” might be the start of the paradigm shift.

I got several more comments when my essay was generously re-posted by The Seeds of Science. Here I want to focus on one comment in particular, from Becoming Human, who said:

What a fascinating read! And it wasn't short, so getting to the bottom says something (about how compelling it was, and about my attention span these days).

I will poke things and say you are mixing apples, oranges, and aircraft carriers. Chemistry, Physics, and even Biochemistry are discrete sciences that operate at relatively low scale with discrete, relatively predictable units (let's not get into the whole quantum thing). They are paradigms because they yield to relatively mechanistic reduction.

Psychology (or thinking or mind) is not in that category. If it is anything more than an oddly stable phenomenon, it is a complex system (again, let us not quibble that complexity is indefinable). While we suspect that whatever the psyche appears to be, it is a manifestation of activity at more granular scales, there is no decisive proof. As such, it is bearishly difficult to break it down into subsystems.

In this way, psychology is in the family with biology, ecology, sociology, and economics (which you allude to above). These, and related fields like nutrition, are so complex that we can only evaluate them by looking at tendencies and correlations, not underlying mechanisms. Sometimes we find little machines, such as efficient market theory or particular diets. Still, they are almost always gross approximations that fall apart under meaningful scrutiny, leaving us with only broad generalizations like "people tend to die without food."

Some folks will insist that the psyche is an emergent property of materialist principles, but that conclusion is an article of faith, not science. Much of what you list above, like DNA as memory or embodied cognition, makes the "mind as computer" argument a little slippery, and should one keep sliding to panpsychism, all semblance to science goes in the sh*tter.

As such, it is probably more instructive to allow psychology to operate as a form of complex systems analysis and stop expecting it to become math or chemistry. This is not to say stop looking for better answers, but we should stop looking for paradigmatic approaches if that means coherent, self-contained systems for analysis.

I think what BH is saying is that the study of the natural world can be broken up into two categories: 

Some fields are “discrete sciences”. This includes chemistry, physics, and parts of biology, like genetics. These fields study “discrete, relatively predictable units” which are usually very small. Because these small discrete things obey clear rules, you can figure out those rules, and use the units to model other parts of the world in a relatively mechanistic way. In these fields, reductionism works.

Other fields are “complex systems”. This includes ecology, sociology, economics, and other parts of biology, like nutrition. These fields study real systems, but the systems they study are so complex that they can’t be easily broken down into simple laws. You can learn things about these systems, like "markets are mostly efficient”, or “iron is good for you”, but there are always exceptions. To use BH’s metaphor, some fields are aircraft carriers — impossible to reduce to simple principles.

Complex systems are inherently arbitrary, like studying bumper cars at the amusement park. You can absolutely learn more about bumper cars, you really can become an expert. There are some distinctions that are very useful — most bumper cars are powered either through an overhead system (OHS) which uses a conductive floor and ceiling, or through a floor pick-up (FPU) which uses alternating strips of metal separated by insulating spacers. But people can and do invent other ways of powering their bumper cars, because the system is arbitrary. There are laws of chemistry and physics, but there are no true laws of bumper cardom.

Complex Systems

I think this is a pretty good way of looking at things. When studying the natural world, some fields study the parts. Other fields study the complex systems. People mistake these two kinds of fields for each other all the time: both are scholarly, both study the natural world, and they’re both empirical, they involve taking measurements and running experiments. So they both look like “science”. But the two kinds of fields are different in really critical ways. Complex systems have immediate practical uses and the list of questions is never-ending, so there’s always more to do, they are a good investment in time and energy. But studying the parts leads to the greatest successes of the sciences, a profound understanding that cannot be matched.

So far so good. But here’s where I start to disagree.

The way we study psychology today mostly does look like BH’s “complex systems”. We search for tendencies and correlations, little machines, not underlying mechanisms. Our findings are almost always gross approximations and broad generalizations. We have some knowledge that is useful, but almost no knowledge that is profound. 

But it doesn’t have to be this way, because every study of complex systems has to be matched by a field that studies the parts of that system. Machines aren’t always simple, but they are always made of something. Words are made of letters, letters of strokes. Bodies are made of cells. The electrical grid is a complex system, maybe too complicated to be studied by reduction. But it still has parts that can be studied in isolation, it’s all current and wires. Systems can be non-reducible in function, but still reducible in structure.

So take psychology: yeah, there are an infinite number of possible thoughts. Thoughts are complex. But what are thoughts made out of? They must be made of some kind of parts, what parts are they made of? There are an infinite number of possible memories, but the memories are stored on some substrate using some compression algorithm. What can we learn about those parts? There are an infinite number of different personalities, but the personalities must differ in some ways, and the differences will ultimately be expressed in differences in their parts, whatever pieces personalities are made of.

BH says, “some folks will insist that the psyche is an emergent property of materialist principles, but that conclusion is an article of faith, not science.” And yeah, maybe this is an article of faith, but I also literally can’t imagine how it could be some other way. In fact, I would say that the faith that the world can be understood as an emergent property of various parts is the core article of faith in science. If things can’t be understood as a system, then what are we doing here? 

BH says, “While we suspect that whatever the psyche appears to be, it is a manifestation of activity at more granular scales, there is no decisive proof.” I never claimed any proof, but again, it’s hard for me to imagine how it could be any other way.

If you look at a Jeep and say, I think it can drive forwards and backwards because of the interactions of the parts inside, you would be right. If you look at a person sick with tuberculosis and say, I think they are sick with these symptoms because of the interactions of the parts inside, you would be right again. 

Maybe these examples are too easy, because we already know that Jeeps and humans are made of granular parts, but I think the logic holds even for a system with unknown parts. If an alien computer crashed to Earth, capable of amazing functions we didn’t understand, I still think it would be reasonable to think that it accomplishes those functions through the interactions of some parts inside. I’m not sure what the opposing position would even look like: “it does those things through a homogenous process that can’t be broken down at all”? Honestly this position seems incoherent. It would have to entirely reject the idea of internal causal mechanisms, which is hard to even think about.

So when people tell me that the mind isn’t the result of activity at a smaller scale, I don’t really know what they mean. Literally what else could it be? Everything else works that way, why not the mind?

Even if you can only explain some things in the aggregate, even if some problems can’t be understood by reducing them to the action of smaller parts, the smaller parts are still there. You may not be able to predict the action of an airplane by studying the screws that hold it together, but the airplane still has screws. 

If you see loose screws on your plane, should you worry?

This is the difference between the naturalist and the biologist: There are basically an infinite number of different bugs; the naturalist’s job is never done. But the biologist can discover what bugs are made of (DNA, proteins, goo, etc.) and build a system that will encompass all bugs, even the ones that haven’t been discovered, even the ones that don’t exist yet. You still need the naturalist to go out and catalogue all the things that can be built out of DNA and proteins. But that doesn’t mean you can’t do biology. There are parts underneath. 

So yeah, I think psychology has discrete, relatively predictable units that will yield to mechanistic reduction. We don’t know what the units are yet, but it’s gotta be made of something. Eventually we’ll find out what that is. I think it’s what we should be looking for now!

Final comment today is less of a comment, more of an essay, one from Jared Peterson, called Rethinking the Edges of the Mind. Jared told me (personal communication) that this piece was directly a response to my question of "why psychology hasn't had a big new idea in decades".

For starters, Jared offers a great perspective on how paradigms can be different from each other:

We need both: lenses to reveal, and boundaries to isolate. The boundaries prevent distraction and dilution; the lenses help us detect detail and structure. Together, they allow science to temporarily ignore the complexity of the whole in order to understand a part. Because if a psychoneuroendocrinologist had to account for bioengineering and evolution every time they ran a study, they’d never get anything done.

The two questions I wish to ask in this essay are the following: has psychology placed its boundaries in the right place? And does it have the right lens?

This definitely lines up with the list of paradigm ideas in my essay. Some of them are clearly about redrawing the boundaries of the field, like when I ask if psychology needs to include plants. Others are clearly about lenses, like when I suggest that we could focus more on the details of memory storage.

Jared is most interested in re-drawing boundaries, and he frames boundaries in terms of context leaks:

The leaking of context into psychology is positively diluvian. Everywhere you go, you hear a magic phrase; context matters. A phrase that is a tacit acknowledgement that, in our attempt to bracket out the rest of the world and put some dam boundaries around our subject of study, there are gaps large enough for the world to leak back in and disrupt our focus. It’s a phrase that means that what we have left outside the dam boundaries is flooding back in making it difficult to study the thing we are interested in. A phrase that means the dam boundaries are not holding.

Context leaks always happen, but they happen in psychology a lot. This is part of why our ongoing conversation around replication has been so messy. Every replication is different from the original study. It will be conducted in a different room, by a different team of researchers. Even if by some coincidence it is conducted in the same room by the exact same team, it will still take place in a different time, a different political situation, etc. Are those differences meaningful to the effect we’re studying? Nobody knows, and there’s no obvious way to come up with a system of those differences. We don’t know how to keep the context out.

Jared takes this leakage as a sign that our boundaries are wrong. If we drew them better, they wouldn’t leak:

This is one way to understand the difference between the hard and soft sciences. In the hard sciences, like physics and chemistry, the units are more easily isolated, and so context doesn’t leak in and overwhelm the model. You can actually do math because the units of analysis are independent enough to behave according to the basic rules of arithmetic.

This strikes me as basically the same idea as BH’s discrete sciences vs. complex systems.

I guess I’ll bite the bullet here and say that I think that psychology as it’s currently done is indeed a soft science. The units of analysis that psychologists use are so big and so messy that you can’t model them. The boundaries around psychological constructs, at least the ones we use today, are not clear. Right on. 

But this is why I keep coming back to alchemy. Alchemy was also a soft science! The units were not isolated, you could not do math if you tried. All you had were weird findings about how gold wouldn’t separate in a fire but would dissolve in Aqua Regis. But with a lot of effort, people took those weird findings, figured out the units, developed methods to isolate them, and eventually started doing math to the whole thing. They created a hard science, and I don’t see any reason we can’t do it again. At the very least we should aspire to it!

Of course, psychology isn’t quite as lost as alchemy. I think that is a tad too insulting. 

I know that alchemy is thrown around as a dirty word sometimes, but I don’t mean the comparison to be insulting. In fact, I find it uplifting. People can go from alchemy to chemistry in just a couple generations. All you need is some relatively minor shifts in perspective! That makes me feel very good about our odds of making psychology into a mature science some time before I die (in the 2440s, of course).

In mature sciences, claims come with relevant provisos. “Sugar dissolves in water” assumes you don’t freeze it instantly. “Vinegar and baking soda create a volcano effect” assumes you’re in earth-like conditions, and so it doesn’t count as a replication if you do it on Pluto. These provisos are an implicit part of the paradigm, left unstated simply because there are simply too many of them.

I hear this kind of reasoning a lot, but I think it’s wrong.

“Sugar dissolves in water” is not how a chemist really thinks. It’s a gloss, chemical understanding translated very approximately into common language. A chemist actually thinks something more like, “crystal lattices held together by intermolecular forces (mainly hydrogen bonds) undergo physical dissolution in polar solvents through intermolecular interactions. The molecules in the crystal lattice are solvated, disrupting the lattice without breaking covalent bonds.” That’s closer, but it’s still a rough gloss to keep things to just two sentences.

The point is, it’s not that chemists have 1001 provisos that tell them what’s in and what’s out. It’s that they have a system, and that system has specific constraints. A chemist has a model of the world in their head. The “assumes you don’t freeze it instantly” and “assumes you’re in earth-like conditions” aren’t provisos or exceptions. They’re left unstated because they are baked into the model.

In Settlers of Catan, someone might say, “you can’t build roads if you don’t have access to brick”. Someone else might object that you can in fact build a road without brick if you have enough cards that you can trade for a brick card — for example, if you have 3 sheep and a 3:1 harbor of the appropriate type. But this is all baked into the rules, and the original statement wasn’t made in ignorance of these interactions.

This is why generalization has to be handled on a case-by-case basis. A good behavioral scientist never assumes a finding generalizes because all our theories are sometimes theories.

People are loss averse…sometimes. People obey authority figures…sometimes. Jared is agreeable…sometimes.

No psychological theory is as law-like as anything in chemistry because we don't know the provisos. Context leaks in, and we can’t cleanly isolate causes, or pin down universal mechanisms. Many of are most important findings are just glorified averages. (e.g., people tend to be Loss Averse)

And maybe part of the reason for this unsatisfactory situation is that we’re stuck using categories as confused as earth, water, air, and fire—studying the phases of the elements of reality, but not the elements themselves. If we are not ever studying the right type of things, of course the models will fall apart. Of course we can't define the provisos, consistently replicate findings, or predict how things will generalize.

I like this metaphor. Concepts like “agreeableness” definitely seem more like the phases of psychology than the elements.

It’s easy to see when someone is agreeable, just like it’s easy to see when something is a liquid or a gas. But it’s hard to see what little bits are behind the agreeableness, just like it was hard to find the elements. 

And all this makes me wonder what would it take to redraw our boundaries? To move beyond endlessly repeating "context matters" without ever saying exactly what that context is? To build models that can replicate not by ignoring context, but by learning to incorporate it?

Maybe psychology’s leakiness is inevitable. Maybe, like Kevin Munger suggests, we need to abandon replication as the cornerstone of the social sciences. After all, if physicists struggle with modeling three bodies in space, why should psychologists—who study something far more multi-causal—expect to ever have anything less ambiguous “people tend to be loss averse”?

And yet, I can't shake the sense that psychology systematically underfits the complexity of human behavior by treating context as noise rather than structure. That our models leak not just because human behavior is messy, but because we drew our dam boundaries in the wrong damn place. Because if it changes behavior, then shouldn’t it be the very content that we study?

One of the most hopeful things I know is that in high-dimensional spaces, local minima are very rare. You can always find some kind of improvement just by moving around. So while it may be hard to find the right boundaries, and there may be some discontinuity, I think there’s a lot of promise in just trying to re-draw our boundaries in as many ways as we can, and seeing what happens. I offered a few suggestions in my original piece, but we could easily come up with more.

Abandoning replication, on the other hand, is the worst idea I’ve ever heard. If you think a question is important enough to bother running a study about it, you should also want to know if other people find the same thing when they run a similar study. And more than that, no one else should have to take your word for what you found in your study — they should run their own version of the study and make up their mind for themselves.

Perhaps the reason cognition and behavior don’t appear in Wimsatt’s chart is that they are not the result of compositional layering. Wimsatt’s chart shows levels that emerge by combining entities of the same kind: atoms into molecules, molecules into cells, cells into organisms, etc. Each level nests within the next, preserving ontological similarity.

But cognition isn’t built that way. It doesn’t arise from the stacking of like onto like. Instead, it emerges from the interaction between different kinds of things, namely, the organism and its environment.

Cognition, in this view, isn’t a “higher level” built atop the others. It’s a relational phenomenon—something that exists between levels rather than above them. That may be why it resists being plotted on the chart: it breaks the compositional metaphor that underlies it.

I strongly disagree. I think cognition does work by combining entities of the same kind, we just don’t know what those entities are yet.  

This line of argument makes about as much sense to me as saying, “auto engineering is a relational phenomenon, it emerges from the interaction between different kinds of things, namely, the truck and the road.” This is true in some sense — obviously the truck was designed with the road in mind, obviously it responds to and is shaped by the road. But it’s also true that the truck is built by combining entities of the same kind. You really can study the nature of the truck!

The fact that organisms interact with their environment doesn’t mean that you can’t study biology as a hard science. You can invent cell theory, disprove spontaneous generation, discover the base pairs, etc. Biology is made of entities of the same kind, even though it’s also relational.

To zoom in on one section in particular:

…shows levels that emerge by combining entities of the same kind: atoms into molecules, molecules into cells, cells into organisms, etc. Each level nests within the next, preserving ontological similarity.

People enjoy this pattern a lot, how it seems like you can explain biology in terms of chemistry, and chemistry in terms of physics. But I think this pattern, while cute, has been misleading. Scientific ontologies don’t have to nest inside each other.

I think psychology may be more like evolution, an information science. Evolution was discovered in the context of biology, but the paradigm doesn’t just apply to DNA, it applies to everything that suffers variation and selection. That’s why we can, at least potentially, apply evolutionary thinking to things like memes, or software.

Computers can be built out of anything, as far as we can tell: silicon, gears, hydraulics, you name it. The arrangement and the function matter more than the materials you use. Cognition is also probably less about what the pieces are, and more about how pieces are arranged. You can make logic gates out of swarms of crabs — so you can probably make the pieces of the mind out of different substrates, if you put your mind to it.