Josh Haas's Web Log

I have the most ridiculous 2 am thoughts.  Like, why am I thinking about this right now?  It’s totally random.  Anyway, here goes:

So there’s an ongoing debate in the sciences about “reductionism”.  If I’m a reductionist, I believe that psychology and nutrition are just applied biology, that biology is just applied chemistry, that chemistry is just applied physics (and is physics just applied math?)  Note that you can remove the word “just” and make the reductionist position sound less condescending, depending on your taste.  If I’m not a reductionist, I believe that there are facts / laws about psychology + nutrition that can’t be paraphrased / explained in terms of biology.  These facts “supervene” on the underlying physical facts, or perhaps they “emerge” from a sufficiently complex system, whatever that means.

It’s an unsolved debate right now.  On the reductionist side of things, you have chemistry, which (as I understand it — chemists please jump in) reduces to physics fairly nicely.  For instance, the interactions of the periodic table can be explained in terms of the mixture of protons, electrons, and neutrons in an atom.  As you get further and further out from chemistry, though, our ability to deductively reason about things gets weaker and weaker.  I think (oh god I’m sure I’m getting this totally wrong) that RNA folding into proteins is currently partially reduced to chemistry: we can explain why any given protein structure is stable by reference to chemical bonds, but on the other hand, we can’t reliably calculate in the other direction and figure out what combination of RNA will fold into any given molecular shape (though we’re working on it).

If you zoom out to psychology and nutrition (which I’m picking on because a) they are generally labeled as being “science” vs something like sociology or economics where people feel obliged to qualify it as being a “social science”, and b) because I find them interesting), we really know stunningly little, and our ability to make meaningful predictions at the level of the whole human being is pretty much, despite all the research that has been done, garbage.  That’s a strong statement, but let’s look at the state of the art.

In psychology, for a large part of the 20th century, the establishment took Freud’s theory that the mind could be subdivided into the id, the ego, and the superego seriously.  In terms of scientific rigor, I’d put that as roughly comparable to Hippocrates’ theory that the human body consists of yellow bile, black bile, blood, and phlegm.  Over the last 40 or 50 years, things have gotten a bit better, and we know now things such as: if I fry brain region X, you lose ability to do Y, and if I feed you drug Z, this will stimulate brain region Q.  There’s also been a plausible effort done mapping out a lot of the more automatic circuity of the brain: we can tell a reasonable story, for instance, about how incoming light becomes representations of objects becomes emotional reactions, or text becomes sound becomes words, and back that story up with descriptions of the way neural networks grow and interact.  But when you move past the stuff that happens more mechanically, and start trying to understand personalities, creativity, intelligence, etc., it’s basically a giant mystery.

There’s a fair amount of literature on those topics, but my strong sense is (and I want to write a longer, better researched entry backing this assertion up because it’s definitely my opinion) that that research is essentially psuedo-science, in a very particular way.  It sounds like science, because the researchers form hypotheses, do proper controlled experiments, and find statistically significant results, but if you look at the theories that those experiments are testing, they’re somewhere in between “begging the question” and “making shit up”.  For instance, (citation needed) there was a study that found that of a group of nuns, the ones who were more optimistic tended to live longer.  Okay, nice, so a journalist reports “psychologists prove optimism is good for your health”.  But although that makes a nice headline, we haven’t actually learned anything, because “optimism” is just a word for people who got a certain score on a survey that some researcher made up (in fact, a lot of psychological research consists of building such surveys to provide definitions for various terms).   We don’t actually know if “optimism” is a meaningful description of the way people’s minds work, or whether “optimism” is just the symptom of something else, like a runny nose could mean a flu or it could mean allergies.    We don’t have the power to deduce anything about optimism in general.

Likewise nutrition is in the same basic state.  Michael Pollan in In Defense of Food does a really good job laying out exactly how little we actually know about nutrition.  Quiz: is margarine or butter healthier for you?  Answer: not sure, but I’d bet on butter.  And yet margarine is successfully, to-this-day, marketed to people who would prefer butter but feel obliged, for health reasons, to pass it by.    The big problem with nutrition is similar to the big problem with psychology: you can do a study that demonstrates a correlation (feed rats more of a particular nutrient, see how many develop cancer), but we don’t have a theoretical framework powerful enough to draw conclusions about the operation of the entire system.  Quiz 2: suppose I show you that feeding rats antioxidants reduces cancer risk, and suppose I show you that feeding rats fiber reduces cancer risk.  What happens if you feed rats both antioxidants and fiber?  Answer: trick question — not enough information!  What if — and this kind of complicated interaction happens all the time in cell biology — the antioxidants trigger cellular processes that, combined with the fiber, release a third chemical that actually increases cancer risk?  We can’t answer that question without performing another experiment.  So when you step back and try to answer the big questions that people care about, like “is vegetarianism healthy?  is veganism?  how about low-carb diets?  how do I lose weight?”  you get a lot of conflicting, confusing information, which is why people keep on successfully writing and marketing diet books.

So, back to reductionism.  Because psychology and nutrition have not been succesfully reduced to biology, it’s an open question about whether or not they can be.  The question is important, because whether or not you believe the answer is yes determines to a certain extent how you set about doing research.  For instance, if you’re a reductionist, you probably want to know if “optimism” can be mapped to something in the brain, such as a particular configuration of neurons, whereas if you’re a non-reductionist, you doubt there’s a clear mapping and so treat it as a basic element of research (or come up with your own theory instead — I kind of like “yellow bile” as a basic element of human psychology).

I’m going to go out on a limb and say the answer is “none of the above”.  I think it’s very likely that concepts that we care about, such as “optimism”, “healthy”, “happiness”, etc., don’t map neatly to underlying physical phenomenon.  The reason I don’t think we’re going to be able to find neat mappings is because those concepts are in part descriptions of reality but in part value judgments.  One person’s definition of healthy might be longevity.  Another person’s definition of healthy might be “able to benchpress 300 lbs”.  Most people are going to have a mix of those things, with millions of little factors thrown in, such as “how do I feel?” “how do I look?” “how energetic am I when I get up in the morning?”  You can try to nail down and quantify pieces of that, but you’re going to lose the gestalt of healthiness.  To me, I think “healthy” and “happy” are like “beauty” — you can debate whether or not something’s beautiful, and there is some element of an objective character to the discussion because people will agree or disagree with each other based on physical evidence, but at the end of the day you can’t prove anything.

So, I don’t think you can “reduce” the kind of questions that nutritionists and psychologists are interested in.  But neither would I say that they are scientific facts in their own right (yellow bile!)  Rather than thinking of psychology and nutrition as sciences, I think a better paradigm is to think of them as a branch of design / engineering.   People think of them as science because the human body and mind already exist and we want to learn about them, but if you think about it, it’s just a coincidence that we have a starting point that already exists… we can still imagine redesigning them from scratch (in theory… in practice, not any time soon…) and, more to the point, we can modify them.

So why do I think psychology and nutrition should really be a branch of engineering? Consider the similarities:

-There is more than one right answer in engineering.  Is an iPad better or worse than a Lenovo laptop?  Neither; or rather, it depends on what you’re using them for, and even then it is still subjective.  Likewise, there are plenty of examples of people with radically different diets living long, healthy lives, and people with radically different psychological makeups and life experiences being happy.  I’ve read so many flame wars between proponents of one diet vs another, or one set of beliefs vs another, that would just go “poof” if both sides recognized that there is more than one right answer.

-Certain things don’t work.  If you ignore the laws of physics when architecting a house, it will fall down.  A laptop produced today runs much much faster than a laptop produced ten years ago. Likewise, certain diets will kill you, as will certain psychological attitudes & beliefs.  Engineering is subjective, but it’s not relativistic.

-Engineered systems are complex, and must be understood as a series of interactions and tradeoffs.  The human body and the human mind are considerably more complex than anything engineers have ever architected, so assuming that there are simple laws for describing them seems naive, when you look at the amount of judgment and detail that goes into, say, designing a car engine.

-There is a design / aesthetic component.  When evaluating the tradeoffs of different solutions, people are going to weight them differently based on their individual values, and that’s totally legitimate.

-You can apply the scientific method to do benchmarks on a design, but a benchmark only tells you so much.  Examples of benchmarks in engineering:  How many pounds of pressure can a given structure hold?  How fast can a database system perform a join of two large datasets?  Both questions can be answered via the scientific method, but neither will tell you the answer to “is that a good design for your structure / database?”  For me, this is the most important consequence of thinking in terms of “engineering” vs “science”.  Much of the research being done in psychology and nutrition seem to be along the lines of benchmarks.  “What is the correlation between nutrient X and effect Y?”  “How do people who journal regularly rate their happiness on a scale of 1 to 10?”   You would never design a car through those kinds of tests; at best, they’re useful for evaluating your designs after the fact against a certain narrow set of criteria.  But that’s really what we want out of those sciences: the ability to design diets, make smart life decisions, be healthy and happy.  So why are we treating them like a science?  If you think in terms of engineering, then research becomes about building prototypes, sharing best practices, critiquing design decisions, debating aesthetics.  You try things, you see what works, when something doesn’t work you ask “why”.   That seems more productive to me than putting a rat or a psych undergrad into yet another controlled study…