Home » Uncategorized » You know nothing about causality, Jon Snow! (Psych favorites, post 2)

You know nothing about causality, Jon Snow! (Psych favorites, post 2)

(2nd in my series of most important things to know about or from psychology)

Psychologists study why people do what they do.  They know almost nothing about why people do what they do.

Today I’ll write about tive problems in inferring cause.  1) We mistake correlation for cause.  2) We forget that actions can have more consequences than the ones we study.  3) We study linear causes when people are embedded in systems. 4) It’s really hard to study a lot of possible causes. 5) We reason backwards from effect to cause.  This piece ran long, so I numbered them if you want to skip ahead.

1) The amount of ice cream consumed on a given day is correlated with the number of homicides on that day. So eating ice cream makes people homicidal, right?  You of course know better.  Killing people makes people crave ice cream!  We had it backwards.  Or perhaps a third way of looking at this is better.  Hot days prompt both ice cream consumption and homicides.  You know correlation is not cause, so hopefully you followed this example.  You know that we need to do experiments (manipulating one variable and holding the others constant) to determine if something is a cause.  If you want to know if eating ice cream causes homicides have some people eat ice cream and others not and see if innocent bystanders are more likely to die at the hands of the former.  I’m no longer IRB Chair, so I don’t even have to review your study for its ethics!

But do you remember to question whether those headlines you read in the news about cause are based on correlations rather than experiments?  You should.  For instance, one British study found that 40% of press releases exaggerated the meaning of the studies they described, with confusion of correlation and cause one common problem.  http://www.nature.com/news/study-points-to-press-releases-as-sources-of-hype-1.16551  I just went to the Washington Post health/science site and found a piece from only 12 hours with this: “How much money people have to fork over when they go to the doctor can make a big difference in how satisfied they are with their health plan, a recent study suggests.”  Those with more out of pocket costs were less satisfied, and this reads like they know cause and effect, right? But did the costs actually “make a big difference”?  Who knows?  It’s not an experiment.  It could be that something makes people choose plans with more out of pocket costs and separately makes them less satisfied with their care.  (Income?)  We can’t know from the study as described.

With the election coming up you will see lots of statements from politicians about how what they propose will cause particular outcomes. Reflect on how many of these are backed by experiments.  Without that why should you buy into their predictions?

2) Actions have multiple effects.  If I buy a donut I not only end up with a donut but also with less cash in my wallet.  But science is often conducted by measuring only one outcome, for instance, counting the donuts I have after the purchase but not the shortage of cash.  Consider research on treatments for depression.  You will find experiments (studying a bunch of different treatments!) with reasonably good control groups.  You will see that the treatment affects depression. What you won’t see is an examination of all the other variables that might also be affected by the treatment.  For instance, might drug treatments for depression both reduce depression symptoms and also create other unwanted effects?

This strikes me as particularly problematic when the unwanted effects are difficult to measure.  One of the prime examples of this is the misguided (to my mind) emphasis on high stakes testing.  Folks with a very reasonable concern (how do we know kids are learning?) introduce testing so they can see if learning actually does increase.  But what other effects might high stakes testing have?  Might it lead to teaching to the test so that other things (like learning wonder, cooperation, and survival tactics for boredom) fall by the wayside?  Might some treatments for psychological disorder cause short term improvement at long term cost?  Might some forms of bucking up support for our favorite political candidate help win an election but also undermine the long term survival of democracy?  Might sparing people from challenge today leave them helpless a year from now?

Next time you read an article reporting that an experiment shows that X causes Y ask yourself what the Z’s might be that X also causes but that might not have been studied.

3) Have you ever left home, changed yourself, then returned home to find yourself returning to your old ways?  If so, this hints a the powers of systems, of the ecologies in which we live.  Psychology prizes (for many good reasons) experiments that involve manipulating X and looking at the consequences.  It is the way we get at cause.  But we only study people for so long, then they return to their homes.  So we might via experiment find a great way to get someone to give up alcohol off at our secluded treatment center. But when they return home their sobriety has a consequence-friends desert them!  And this serves to bring people back to the drink.  People live in a web of causes that can tend to hold things in balance.  Push on one factor and other factors will subsequently push back to return things to the status quo ante.  Lasting change might be better served by understanding this.

How much has our emphasis on the experiment caused a picture of the human that misses the centrality of our complete ecology?

4) If we can only know about cause through experiments then the set of causes of human behavior that we will know will be those causes that are easily manipulated.  If you want to know if going to college actually has an effect on something you need to randomly assign some people to go to college and others not, then look for the difference. This is insanely hard so I really have no clue if college actually causes good outcomes.  (Sorry if I just cost a lot of colleges tuition money!)  And I live my life understanding that I do not actually know if I am helping or hurting my students.   I can’t do the experiment.

Perhaps a treatment for X that takes 2 years will be much more effective than a treatment that only takes 2 months.  But you could only find out if you could do a study that actually delivers 2 years of treatment. Good luck with that! It will often be enormously expensive and logistically excruciating a study that takes that long.

How much does this limitation of science shape our understanding of what can cause good outcomes?  You have before you the rest of your life. You could imagine embarking on an experiment with it to see what happens.  How could I tell you that the life path you choose helps or hurts?  The experiment can’t be done.

5) Often the question we have is why something happened. The science can often only tell you what some causes of a given outcome are.  We can say that blowing up a thermonuclear device within fifty feet of you causes bad things for your health.  But just because your health is bad does not mean that you just had a thermonuclear device go off nearby.  We can speculate about the things that caused us to be where we are now, but it will mostly just be speculation.

People want to know why relationships ended (or thrived!), why jobs are lost (or found!), why their children are learning (or slumping).  While we know some factors that affect relationships, jobs, and children, it is awfully difficult to know about what in the history of a particular person has been an important cause.

6) A concluding thought.

All of this (and more!) says that most of what I would like to know about why people do what they do I cannot know.  Paraphrasing Gleitman (I think), psychology consists of small islands of coherence in a vast sea of chaos.

And in this I come to one of the main stories of psychology.  I have before me a choice.  I can pretend to know things I do not.  Or I can despair from knowing how little I know.  Or I can understand the limits of what I know and then act, knowing the act is at least to some degree a leap of faith.  What do we think of “faith” in modern America?  My sense is that we value certainty, instead.  Might understanding the limits of what we can know about people, and noticing that our actions then are leaps of faith rather than acts born of certainty help us find new appreciation for that word “faith?” And new tolerance of ourselves and others when our actions do not bring us to the verdant pastures we sought?

And for the field of psychology–can we neither claim more than we know nor despair in that which we do not know, but rather do the next study, in the faith that our collective enterprise will turn those small islands of coherence into somewhat larger ones?

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: