Reductionism

Reductionism - looking at a complex situation and identifying a few main forces to explain what is happening - is an incredibly powerful tool for understanding the world around us. There are an almost infinite number of known and unknown factors that go into even the most mundane phenomena, and an even larger number of ways to describe them. Trying to take stock of everything all at once will drive you insane and leave you with a lot of “patterns” that say more about your thinking than the reality of the situation.

Reductionism also rhymes beautifully with the Scientific Method, as it is much harder to test the effects of many interacting factors on an outcome than it is to test the effects of one or two. There is a satisfying progressive nature to identifying and testing the most important factor, then the next, and then the next. At each step you have fewer anomalies to contend with, and you’re always standing on stable ground if you test the factors as you go. Even though, at any given time, you may only be able to account for a small fraction of variation in outcomes with the causes you propose, you can be relatively sure that those causes really do influence the outcome. The steady drumbeat of reductionism has been the marching music of science for hundreds of years. Many of the triumphs of science have been reductionist in nature, and most scientists see reductionism as a cornerstone of scientific practice. As a tool of investigation, it has worked very, very well.

To pick a controversial example, say we want to test whether Prozac treats anxiety. We collect a group of people with anxiety and measure their anxiousness with something like the Beck Anxiety Inventory. Then we split the groups into a test and control group, administer Prozac for a period of time to the test group and a placebo to the control group, and measure their anxiety again. If the test group has a lower anxiety measure than the control group, we could say that Prozac is an effective treatment for anxiety. The world is a messy place, and because of that, scientists accept some statistical wiggle-room when allowing that a certain finding is significant. When the test is over, if 95% of the test participants have anxiety levels like what we would expect if Prozac improved anxiety measures, then we can say it works. Because 95 is a good number, right?

There are many stories contained in the
               spaces between the points and the line.

But that other 5% might not be a random fluke or statistical noise; those participants could all have a specific genetic variant Y that acts to negate the effects of Prozac. In this case, the better headline would be “Prozac is an Effective Treatment for Anxiety in those without Genetic Variant Y,” but that’s not captured in the significant finding. Furthermore, and more importantly, answering the question of whether Prozac significantly treats anxiety says nothing about the effect of exercise, sleep, financial security, social position, or any number of other factors that also likely play a role. As the reductionist ball rolls forward, these other factors can be incorporated into a model of what plays into anxiety. Then we can see if people who don’t take the Prozac, get a good night’s sleep, or get adequate exercise are, on average, more anxious than people who do. We get better and better at explaining the variation we see in anxiety across people as we add more factors. In psychology, explaining 30% of that variation is considered a strong result. But why should we even bother with all this work if, in the end, 70% of what goes into anxiety is still completely unexplainable?

Well, one reason is that we might not even know that genetic variant Y exists, or what categories of social position are important. In a world of unknown unknowns (to borrow a phrase from Donald Rumsfeld), we measure what we can and try to find the most solid relationships possible. We accept that our explanations will not hold water in every circumstance, and we try to pursue further the circumstances where the explanation fails. This has to do with how we form questions in science; we ask “Does Prozac improve anxiety outcomes in most people?” because we don’t have the tools to ask “What is every factor that goes into this particular person’s anxiety?” Reductionist science focuses on measuring the outcome based on a few factors, and then replicating that across many trials to identify the effect of those factors on the outcome. If we want to know whether those factors actually have an effect in reality, there’s not much more we can do.

The unknown unknowns are blowing his mind

However, there is a disconnect between how this tool is properly used and how its findings have been communicated, especially at the points where science intersects with the human experience. Scientists endeavor to find the factor X that explains the most variance in trait Y in the population, but the message that too often comes across is that that factor X causes trait Y. To illustrate what I’m talking about, I googled around and cherry picked some examples. This is obviously not a fair survey, but I’m sure you’ve been bombarded with news stories like these, argued on the nature side as well as the nurture side.

Most of these articles couch their statements in statistical or quasi-statistical language - the gene is “linked to,” “associated with,” “implicated in,” or “may shape” certain behaviors - that are likely to be lost on the audience. That is partly a failure of education, in that many people don’t have a clear idea of what those terms mean, or how statistics operate in the context of reducing complex phenomena to statistical shifts in test outcomes. But the larger failure is in communication: it would help enormously if more of these articles had clear caveats - “not everyone with this gene will be anxious,” or “this gene plays a role in anxiety, in interaction with other genes as well as your environment.”

Reductionist explanations-in-progress are often presented as if they were the whole story, and the simplified relationships that get into our heads result in a lot of confusion and frustration about categories and causes. The world we actually live in is incredibly complicated, and the real reason why things are the way they are often has a lot to do with history, chance, and “emergent” phenomena that arise from the interaction of multiple parts. Remember, reductionist explanations-in-progress are good at identifying which causes are real, but at a given time they might only be able to account for a small percentage of the variation you see.

Often, we can’t explain why Bob has anxiety, even if we can show that social environment, exercise, genes, endocrine regulation and a host of other factors contribute to anxiety. Often, someone will attempt to explain (judge might be a better word) why Bob is anxious using just one or two of those factors. Often, this is frustrating, offensive, and wrong. The explanation is taking a complex phenomenon (Bob’s anxiety) and trying to reduce it to one or two inadequate and possibly irrelevant dimensions. And when it comes to explanations and judgments about people (or social phenomena, or a number of other very important topics) irrelevant and inadequate can quickly become disrespectful, confusing, and painful. It’s not cool to use incomplete evidence to explain away Bob’s experience.

Questions like “What should I do to be healthy?” or “Why is he anxious?” require us to synthesize all that information we’ve gotten from reductionism about factors that affect these things, and integrate it into a wider world view. It is our job as thinking people to take that information for what it’s worth, and recognize that a study showing that a genetic variant makes one more likely to become anxious does not necessarily mean that a given person’s anxiety is solely genetic in cause. We should treat reductionism and holism as complementary approaches to the truth, and different ways of thinking of the same problem. We need to change the way we think and communicate about these simplified but powerful takes on complex relationships.