My dad emailed to say he didn’t quite follow last week’s post.
I did a gut check and decided I was OK with this. Part of what I’m trying to do with this blog is learn to communicate with a broader audience about science. But the desire to do that arose from an earlier, more primal impulse: the urge to call shenanigans when I see scientists misrepresent their fields to the public in self-serving ways. In this case, I was responding to an explicit claim that wasn’t true (neural network models learn the regular past tense of verbs just fine, thank you very much) and an implicit claim that hardly seems worth dwelling on any further here. I will explain it to my dad — who is, by the way, a very sharp and well-read guy, just not about machine learning — over Scrabble and cookies at the kitchen table when I go home for the holidays.
But my dad’s email did get me thinking about the implied audience for different kinds of scientific communication. In the seminar I’m teaching this semester, we shifted gears from talking about basic learning and memory systems to talking about emotion just in time for this paper on cues to emotional valence from body posture to be very topical. The results have been discussed all over the place. They are pretty interesting. Basically, extreme facial expressions don’t convey much information about valence. This is surprising because psychologists tend to assume that emotional information is conveyed mainly by facial expression, but it turns out that for extreme emotions, facial expression is ambiguous, and different postural contexts can turn the thrill of victory into the agony of defeat. It’s a bit like the old game of “orgasm or excellent marinara?” but easier to get past the ethics board. What’s really striking about this paper, though, is how it begins:
“Jennifer checks the numbers in her lottery ticket, when she realizes she hit the 10-million- dollar jackpot. Michael fumbles for his car keys while his 3-year-old son steps into the street and is hit by a passing car. In a split second, Jennifer and Michael experience the most intense emotions of their lives.”
So much human drama! Such vivid imagery! What is this doing in the pages of a scientific journal? Interestingly, though, it is not an outlier in this regard. Psychology papers in Science are apparently under great pressure to write as if the implied audience were someone you randomly sat down next to on the Subway. Here are the first few sentences from another two Science papers, both of which also report very high-quality, theoretically important work:
“In Aesop’s classic fable, the ant and the grasshopper are used to illustrate two familiar, but disparate, approaches to human intertemporal decision-making…Human decision makers seem to be torn between an impulse to act like the indulgent grasshopper and an awareness that the patient ant often gets ahead in the long run.” (McClure et al., 2004)
“Should you shout at your dog for soiling the carpet or praise him when he does his business in the yard? Most dog trainers will tell you that the answer is both. The proverbial ‘carrot-and-stick’ motivational approach refers to the use of a combination of positive and negative reinforcement: One can persuade a donkey to move either by dangling a carrot in front of it or by striking it with a stick.” (Frank et al., 2004)
Compare this with the papers before and after the emotions-from-posture paper. Even the titles require a relatively up-to-date knowledge of biology to parse (“Progenitor and Terminal Subsets of CD8+ T Cells Cooperate to Contain Chronic Viral Infection” and “A Mutation in EGF Repeat-8 of Notch Discriminates Between Serrate/Jagged and Delta Family Ligands,” respectively). Here’s the first sentence from the one about mutant fruit flies:
“The evolutionarily conserved Notch (N) signaling pathway affects numerous cell fate and differentiation events as well as proliferation and cell death.”
What’s that, dad? You don’t know what “cell fate” or “differentiation events” are? Or what a “signaling pathway” is? That’s not really my problem if I’m a developmental biologist. But if this had been a psychology paper, it might have started out something like “Do you ever wonder why your heart isn’t just an undifferentiated mass of smooth muscle dangling uselessly from the end of your aorta? The answer turns out to be related to a protein discovered in the wings of fruit flies in the early 20th Century…”
I am ambivalent about this. On the one hand, it would be nice if anyone who is generally scientifically literate could get through the first paragraph of a paper on immune function or climate change or superconductors without feeling completely lost. On the other hand it’s clear that psychologists are expected to talk about our work in a way that allows people to relate it to their own experiences. This is a problem, because the concepts and phenomena we deal with are often quite subtle and arcane. The simple cover story about why a problem is interesting can lull readers into imagining that the results of the research are actually at the same grain size as the metaphor used to communicate it. When this leads to a disconnect between the narrative framing of a result and its actual implications, which do you think will stick in most people’s minds? Which will end up in the press release?
For example, in the “carrot and stick” paper, it turns out that the learning paradigms were based entirely on reward. There was no punishment in the experiments. Instead, some stimuli were less likely to result in reward than others. The students in my seminar are always confused by this. “How is a 20% probability of reward a ‘stick?’ It doesn’t make any sense.” And I patiently explain that not getting a reward when one is expected is a potent learning signal, but no, they’re right, it’s not like being hit with a stick, not even at the level of striatal learning circuitry. It’s more like not getting a carrot, at which point I start to draw the predictive reward coding figure from this beautiful Schultz paper on the blackboard.
While I’m drawing, a discussion erupts among the students about how when you lead a donkey with a carrot, you never actually give the donkey the carrot, but actually it is the promise of the carrot that urges the beast along. Last year when this came up, I had just heard Bobcat Goldthwait interviewed on The Sound of Young America, where he made an arresting analogy about how at a dog track, dogs are destroyed after they catch the rabbit, and I had to stop myself from quoting him. (He goes on to say that in show business, when you catch the rabbit they just wait for you to destroy yourself.) These conversations are very engaging, difficult to reign in, and ultimately contribute little to understanding the role of dopamine in reinforcement learning.
Psychologists have always reached for analogies from everyday experience. We typically don’t even make up our own jargon. One of the most famous quotes from William James starts “Every one knows what attention is…” In his defense, at that point James was working with a psychology that was just an inchoate offshoot of philosophy. In our century, though, psychologists use the word “attention” in specific ways, many of which have only the remotest connection to the word’s usual meaning in English.
Perhaps the most serious problem with the “nice lady on the subway” as implied audience for scientific communication is that it contributes to an environment in which over-interpretation of results is essentially standard. By accepting the assumption that the concepts we’re working with should be familiar and accessible to everyone, we invite the misapprehension that our results can speak directly to the kinds of questions ordinary people have about how their minds work. We can do better than resignedly sighing and shrugging when people outside the field ask whether they would be better at freestyling if they could just shut down their prefrontal cortex, or if they’d be better at cuddling if they had more oxytocin. We can, instead, insist on describing our results in a way that makes it clear that their scope is limited. But as long as psychologists are rewarded for framing our work in relatable ways — and there are few tastier carrots than a publication in Science — papers will be written in a way that invites readers to look to them for simple answers that are not forthcoming.
Image credit: Screen grab of Jay-Z introducing himself to a woman on the subway who didn’t recognize him, but could tell he was famous from the entourage, cameras, and crush of excited onlookers. She turns out to be Ellen Grossman. Ms. Grossman’s artist statement is arguably a lot more interesting than Mr. Carter’s.
BTW, this really is quite a nice paper: Aviezer, H., Trope, Y., & Todorov, A. (2012). Body Cues, Not Facial Expressions, Discriminate Between Intense Positive and Negative Emotions Science, 338 (6111), 1225-1229 DOI: 10.1126/science.1224313