How neuroscience comics add KA-POW! to the field: Q&A with Kanaka Rajan

Kanaka Rajan says it took her twice as long as it should have to complete her Ph.D. in computational neuroscience—because she had to decipher “inscrutable physics papers. It turns out, they weren’t that hard. They were just written badly.”

Now associate professor of neurobiology at Harvard Medical School, Rajan tries to make her own scientific papers more comprehensible—through comic strips.

Rajan, who herself sketches regularly, began collaborating with professional artists in 2018 to make narrative illustrations to explain the findings in some of her papers. The comics, she found, helped make computational neuroscience accessible to students who have little scientific training.

Over the past five years, the practice of including comics in scientific papers has taken off, Rajan and several illustrators say. Neuroscientist and artist Matteo Farinella, who co-wrote and illustrated the neuroscience-focused graphic novel “Neurocomic” more than 10 years ago, says that today’s demand for artistic collaborations is the highest he has seen.

Creating comics also has some personal benefits, Rajan adds: Her papers that feature comic illustrations have an “exponentially higher” number of citations than her other papers. “People have written to us. More people have downloaded the GitHub where the code lives. Students have written or have talked to me. The h-index is better! The comic disseminates the work in a deeper way.”

The Transmitter spoke with Rajan about how comics can aid scientific research and how other neuroscientists can try the approach in their own work.

This interview has been edited for length and clarity.

The Transmitter: How did you become interested in incorporating comics into science? 

Kanaka Rajan: I had two points of entry: One is that I sketch a lot. I think everything is interesting once you draw it; it helps your brain process things.

The other point was realizing that scientific papers are kind of an awful medium. They’re incomprehensible. It’s kind of a secret form of gatekeeping. As if we keep our science special sounding, then it will attract only the luminous geniuses. That thinking has led to several problems in the field that diversity, equity and inclusion efforts alone will not be able to solve. The pipeline is leaky everywhere.

So one of the concrete things is to democratize access to the information in my papers. The content in it is not 200 years of physics you have to learn. It’s high school linear algebra for the most part. It’s programming, which pretty much every high school student can do. A dedicated high school student or an undergrad can ramp up to a research problem very fast.

TT: Why are comics effective at lowering some of those barriers to entry in neuroscience? 

KR: In making comics, we’re not diluting the content. We’re just removing jargon from it. We’re democratizing access to it, and we’re turning technical details into schematics and graphics. That’s it.

We’re not stupidifying the work. I’m making it visually engaging without jargon. We’re saying, “Come play with us; the sandbox is open.” We’re not saying, “We’re luminous geniuses and we have special sand.”

TT: Do you have an example of when working with a comic illustrator helped make your work more accessible? 

KR: One year, I was asked to lead a workshop at the Computational and Systems Neuroscience (COSYNE) meeting. I was meant to teach them foundations of why it is that neural network models are such a big deal in neuroscience. And when I was trying to explain that on my own, without a professional artist, I tried to use the metaphor of a lightbulb. I said, “When we first built models of the brain, we were building models that were like lightbulbs, but the problem with those models is that they didn’t do anything. When they tried to do something, they would blow up.”

Then, once the field started using nonlinear recurrent neural networks (RNNs), we needed a new metaphor. We had this lightbulb that blew up, and we turned it into a lava lamp, because now you have interesting dynamics within the model. But in a lava lamp, these interesting dynamics are all chaotic: Every time I turned it on, it was a different pattern.

So the thesis of my talk was that, if you add inputs to the model, you can quench the chaos somewhat—and then you can get reliable responses.

There are lots of ways in which the lightbulb analogy is imperfect, though. For one thing, it’s just this one object. And everyone knows that when the switch is off, the brain doesn’t shut down. It’s contrived at best.

Now let me show you what happens when a professional takes over: Here comes Jordan Collver, who is an illustrator and science communicator. And he turns that story into a comic that contains the exact same content but with a completely different metaphor. He said forget the lightbulbs; get to the meat of the problem: What does it mean to train a network? He was like, “I don’t have a metaphor to connect lightbulbs with training of a model. But I do for Lego blocks.”

Everyone knows Lego blocks. They come in different colors. There are various shapes of them already. You can mix and match them. The metaphor was just so beautiful and adaptable, flexible and universal, that it basically left the lightbulb in the dust.

TT: Have you seen a difference in how people respond to the Lego explanation rather than the lightbulb?

KR: People have seen this and written me emails saying that now they understand what it means. Now they can read articles about artificial intelligence machine learning, and they go, “Oh, my God, now I know what they’re talking about.”

Another example is that I recently gave a webinar and talked about how, in the RNN models I build, I think about inter-area communication a lot: What information is passed within and across brain areas to make your brain work. We think about these flow fields and flow of activations between active neurons bidirectionally, and which type of neuron is projecting to which other type of neuron and so forth, all the time.

After the webinar, this physicist wrote to me and said, “You know, there are techniques in physics, such as persistent homology with weights, that you can use to extract these types of directional flow quantities. Have you considered those?”

I was like, “No, I didn’t know that existed. But thank you!” Now I have a whole other set of techniques to read.

TT: What are some of the other benefits you see? 

KR: Well, the person who is funding the research is the taxpayer, and they don’t know what the paper is about as it’s written. Given that public trust in science is at an all-time low, I find this important.

More selfishly, people have to reach funders. They have to reach program officers. They want to reach what I will call the earnest public—scientifically literate readers of The New Yorker, for instance. They want to reach them, but they can’t with jargon alone.

TT: What would you tell other neuroscientists who want to get started working on collaborations with illustrators?

KR: I would say message artists you like and say, “Look, I have an idea for you. I have this paper. I would like to turn that into a comic strip.” People usually respond. X, honestly, even with its problems, is great for this.

Another thing to keep in mind is that you can write the costs of these things into grants now. No one writes papers now where they’ve slapped together their Python figures; they have professional artists make multi-panel figures using Illustrator. People do that already, but there is this whole other venue for representing your work visually.