Over email I was having an exchange with someone… thought I would put down some of my thoughts here.
I don’t really think it is very much established that reading fiction affects beliefs, at least in the social science sense of “affects”… in the philosophical sense, everything affects everything, so that is trite… in the common-sense sense, introspection clearly suggests that when immersed in fiction and in the warm glow shortly after, there is an effect. But in the social science sense, that reading fiction affects the reader any more than any other social experience (watching TV, engaging in conversation, going on a walk) where these are randomly assigned, well…. not sure. The correct framing of the question, to me, is whether random assignment (of this activity, for some significant period of time or changing the opportunity cost in some significant way) generates differences in outcomes (at time frames bigger than 15 min, which is what psych seems to focus on, just because to get published you have to have significant findings…;-) large enough to lead one to think “wow this is probably bigger than going for a walk”… well… to be honest… I’m skeptical… even though as you see from my website I am a huge reader and promoter of reading (I run a non-profit that supports community libraries in Africa). I think a lot of reading fiction is self-selection and so endogenous to beliefs….
There are lots of psych studies like this Harry Potter one… they have several problems:
a) they usually report effects “shortly” after treatment (does anyone doubt that very recently reading an account of the Holocaust, or making that reading salient, would change how you responded to questions in a survey?)
b) they often condition on “the reading affected you” (so then really just self-section-salience effect)…. if we give you $1000, and you are very smart and clever and care about getting rich, then giving you $1000 is a good way to raise people’s income
c) psych is notorious for selecting positive p-val studies for dissemination…. (the lure of media promotion must be irresistible)
My point (I think!) is that we do not need evidence that “reading a book can affect a person” because that is similar to saying “experiences can affect people” … so the real issue is the magnitude and the counterfactual…. and I think the magnitude of the short-term effect is pretty meaningless…. because any experience changes you in a short term so there are millions of counterfactuals (instead of reading the book you were asked to take a nap, or to meditate…)… so I guess my inclination is to only take seriously experiments that offer “prolonged exposure” and that measure outcomes “at a remove” from the exposure. To my knowledge, in terms of reading fiction, I do not think anyone has ever done such an experiment.
An ideal experiment, then? 500 young adult are randomly given Autobiography of Malcolm X (contains a strong message of self-autonomy and self-identity construction, perhaps!) and a control group not given it. 10 years later, a previously defined (and registered) outcome measure is collected…. and verification that in meantime no special salience to the book (which could generate an interaction effect).