Can fiction improve you?

Was Gwern wrong? Maybe...

If you’re interested in this, you might be interested in the book I am working on.

---

Gwern against fiction?

In his recent interview with Dwarkesh, Gwern said this,

You could definitely spend the rest of your life reading fiction and not benefit whatsoever from it other than having memorized a lot of trivia about things that people made up. 

I tend to be pretty cynical about the benefits of fiction. Most fiction is not written to make you better in any way. It's written just to entertain you, or to exist and to fill up time.

Gwern isn’t denying that fiction is fun, but he does think that it can’t be improving. The two are not incompatible, of course. You might still value fiction despite it being a “waste of time”.¹

This raises two interesting questions. Why might we think fiction isn’t improving? And if it isn’t, could it be damaging?

Fiction and Theory of Mind

In 2013, a study claimed that reading a short story improved your ability to recognise other people’s emotions. This is called Theory of Mind, i.e. reading fiction improves your ability to think about how someone else might be thinking or feeling. This is a claim you often see being made by literary people. But the effect size is small, and we don’t know if this is caused by the fiction reading or not.²

If reading fiction is beneficial in ways we can measure with psychological tests, then it is beneficial to an extent you wouldn’t notice very much. This meta-study shows a very small effect size. In this study, 214 students were given a whole book to read and a much bigger set of tests was administered. Some read literary fiction, some science fiction, some non fiction. They once again found very small effect sizes.

The basic methodology of these studies is that after reading short stories or non-fiction, study participants were shown pictures of people to see if they could read “the mind in the eyes”. There is also an “author recognition test” which asks people to identify which author’s names they know on a list which includes some fakes. This is supposed to be able to test whether long-term fiction reading is beneficial, and again, those people who know a lot of authors do get a small advantage on the Theory of Mind test.³

It is likely that the benefits that are showing up in these studies are just the result of other factors, like random effects due to the variation between readings, or the people with better Theory of Mind being more drawn to fiction. A study that was designed to prevent these confusions concluded:

we are not confident that reading a short text of any kind can reliably improve theory of mind. Any immediate effect of reading on theory-of-mind abilities is likely to be fragile and depend not only on the individual reader and text, but also the relationship between the two.

“If you have a strong theory of mind already, you are more likely to read fiction” is probably the conclusion we can draw from all of this.

So when Gwern says that fiction can’t improve you, these are the topics he is sceptical about. The science here just doesn’t hold up very well.⁴ Here’s what he wrote about this,

Justifying fiction on practical benefits is a powerful justification — if it’s true. But these claims of benefit are a little questionable, being largely correlational. More generally, the question is one of “transfer of learning”: the typical result in psychology is that if you spend time learning or training on something, you will improve substantially on that, improve a little or moderately on things which resemble that closely (“near transfer”), and improve hardly at all on anything else (“far transfer”). So no matter how much of a mental workout you get playing chess, it won’t “transfer” to, say, learning English vocabulary. Anything which might cause far transfer would be unusual and exciting (such as dual n-back), but the task is tantamount to increasing IQ in normal healthy people - a holy grail which remains out of reach half a century later. (Even years of schooling, spending hours a day on a variety of subjects, fails to increase childrens’ IQ more than a few points at best.)

Is fiction bad for you?

In an essay arguing that we should probably tax the production of new fiction (because there is too much being produced and we ought to read the best fiction but instead get dragged into the new), Gwern worries that fiction is in fact harmful. It’s not just that studies showing improved empathy are unreliable, but that reading fiction teaches you to believe falsities.

Gwern says fiction can be “unfairly persuasive”, bypassing rational thought. He cites various findings to build a picture of fiction as subtly persuading us to false beliefs,

Priming & contamination is very bad news for anyone who wants to think that they are not affected by the fiction they read. People seem to believe what they dream, the more engrossing a fiction the more you blindly favor the protagonist and believe the story realistic, people who watch TV believe4³ poetic justice actually happens outside stories and the world is more dangerous than it is and looks like TV (which may explain why it’s easier to brainwash people with TV into supporting the death penalty rather than gay marriage), perhaps because viewers can emotionally treat characters on screen as real44 ^and replacements for real relationships with negative consequences like reductions in family size or pregnancy rates

When we read ⁠thorough research⁠ on narratives that lead to changes in behavior and reduction in stereotyping, we should not forget that this makes fiction a double-edged sword.

Everyone tells kids that what they see on TV isn’t real — of course! But what makes you so sure that you don’t believe what fictions you read?

But this isn’t the same as a robust finding showing that people who read a lot of Shakespeare are more credulous. The fertility study is about women of low socioeconomic status watching soap-operas and the effect was to reduce fertility not to prevent it. Something about those particular shows was the likely cause. The study authors write, “In societies where literacy is relatively low and newspaper circulation limited, television plays a crucial role in circulating ideas.” The study about pregnancy rates was about reality TV, not fiction. This is hardly an indictment of fiction in general.

Fiction may well be, as he says, a double edged sword. What isn’t? How far do we take this? Was the telling of fairy tales and folk stories a negative utility on human culture? Was the adaptation of those modes into Shakespearean drama a further worsening? Is it possible to create a human society that doesn’t reason by analogy and metaphor, which doesn’t fundamentally mingle fact and fiction? Should we ban all gossip and tax all rumours? A lot of clever people believed in the Soviet project and it was fiction that shook some of them out of their mistake. It’s not a simple situation.

Here’s an interesting point to illustrate how difficult it is to generalise about fiction.

It’s worth noting that things like the Terman sample find that in real life, being Extroverted and low on Agreeableness correlate with success; the latter, at least, is more characteristic of antagonists than protagonists.

But Shakespeare’s successful heroines are disagreeable! They might not always be openly unpleasant like Lady Macbeth, but characters like Rosalind and Helena pursue their own ends and are prepared to speak out against men and to pull tricks on them. Sometimes quite significant tricks. How would we rate Cordelia in King Lear? She’s very nice, but she’s also openly disagreeable from the start. Isabella is very charming, but we’d hardly call her agreeable. Her silence at the end is a remarkable moment of disagreeableness! Simply being disagreeable isn’t the whole trick to success: knowing when and how to be disagreeable matters an awful lot. That’s one of the main features of King Lear. One of the most enjoyable things about All’s Well that Ends Well is just how intelligently Helena walks that line. Studies abstract, literature helps us think about how those findings actually work in the mess of life.

This debate is so difficult to make any progress on because it pits two very different modes of thinking against each other. Robin Hanson says,

What evidence suggests that readers of such stories have gained such insight, instead of merely gaining an intuition telling them of such insights? 

This gets to the heart of the matter. Intuition is valuable! And hard to measure! J.S. Mill understood that. He wrote to Carlyle in the 1830s:

… the means which are good for rendering the truth impressive to those who know it, are not the same and are often absolutely incompatible with those which render it intelligible to those who know it not.

Sometimes thinking about reality involves thinking speculatively, imaginatively, unrealistically. Tesla read Goethe and Darwin read Milton. This was not insignificant in the later development of the induction motor or the writing of On the Origin of Species.

Mill believed that rendering the truth intelligible to those who know it not was “the proper office of the logician or I might say the metaphysician, in truth he must be both. The artist’s is the highest part, for by him alone is real knowledge of such truths conveyed: but it is possible to convince him who never could know the intuitive truths, that they are not inconsistent with anything he does know; that they are even very probable, and that he may have faith in them when higher natures than his own affirm that they are truths.”

What are these truths? In his essay What Is Poetry? Mill said poetry was “the delineation of the deeper and more secret workings of human emotion… the truth of poetry is to pain the soul truly.” Fiction was supposed to give a “true picture of life”. Gwern and Hanson dispute that, but it is the general reaction and approbation of readers of Shakespeare and Eliot and has been for many generations now. At some point we have to accept that there are limits on what empirical research can tell us. Personality psychology is robust but it is also narrow. Plot is not the only way fiction delivers its meaning.

The great irony of all of this is that Gwern himself was deeply inspired by Borges.

I was blown away by the fact that you could be so creative, with all this polymathic knowledge and erudition, and write these wonderful, entertaining, provocative short stories and essays. I thought to myself, “If I could be like any writer, any writer at all, I would not mind being Borges.”

I will end on a point of strong agreement with Gwern.

Dwarkesh Patel 

But it sounds like your own ideas have benefited a lot from the sci-fi that you read.

Gwern 

Yeah, but it’s extremely little sci-fi. Easily 99% of the sci-fi I read was completely useless to me. I could have easily cut it down to 20 novels or short stories which actually were good enough and insightful enough to actually change my view.

That’s why we have a canon. That’s why serious readers pay more attention to the best works. And that’s why fiction’s uses are so hard to discern. Poetry offers us more ways of seeing into ourselves than logic ever can, but they must be used together, discerningly. Perhaps, after all, we have progressed no further than what Mill wrote to Carlyle,

Now one thing not useless to do would be to exemplify this difference by enlarging in my logical fashion upon the difference itself: to make those who are not poets, understand that poetry is higher than Logic, and that the union of the two is Philosophy.

---

If you’re interested in this, you might be interested in the book I am working on.

Claudie Blakley and Judi Dench in All’s Well

1. I posted that on Notes and some people seemed cross that he was denying the idea that fiction was valuable as a “useless” pleasure, which of course is not precluded by what he says here. Gwern has read a lot of fiction, is a major Borges fan, and has a whole essay on his site about reading fiction, in which he quotes David Hume.

The good or ill accidents of life are very little at our disposal; but we are pretty much masters what books we shall read, what diversions we shall partake of, and what company we shall keep…When a man is possessed of that talent, he is more happy by what pleases his taste, than by what gratifies his appetites, and receives more enjoyment from a poem or a piece of reasoning than the most expensive luxury can afford.

So he seems to be quite happy with the idea that reading for fun is a good thing to do.

2. And the studies often have small samples: Black and Barnes 2015 was done with 91 undergraduates.

3. I find this to be a slightly silly test. This study contends that “Readers of literary fiction must draw on more flexible interpretive resources to infer the feelings and thoughts of characters.” Is it not absurd to test that large idea by giving 86 people a random extract to read and then ask them to “identify facially expressed emotions”? Note that the study does not tell us whether the fiction provided includes physical descriptions associated with the emotions that were being presented in the facial recognition test...

4. As for the old canard that literature improves critical thinking, if you ask Consensus about that you get a load of studies with either zero citations, or with one or two. Why this even need studying is beyond me. If you want to learn critical thinking, study logical subjects like philosophy and economics. Not poetry.