In Defense of Good Writing

Having recently seen multiple political scientists recently, and boldly (absurdly, preposterously), assert that AI can outperform social scientists at our research, I couldn’t resist writing this defense of what might be construed as old school social scientific work and good writing, which comes from good thinking and not simply polished syntax.

You know, the stuff that led us to spend our 20s pursuing doctorate degrees, despite knowing that our friends outside of academia were generally having more enjoyable experiences in their 20s and may very well have more lucrative career paths as well.

You know, the reading and production of thoughtful research that got us into this line of work. Perhaps you haven’t totally forgotten it?

I’ll tell you why I am in academia. I had wonderful professors at UC Berkeley – people who introduced me to constitutional law and statutory law, people with whom I debated tensions between individual liberty and national security, people from whom I learned research design, and in a different part of campus, people with whom I delved into great literature by authors like Jane Austen and Virginia Woolf. Like many political science majors, I had planned to go to law school, but I wanted to work first and graduating in the great recession, jobs were scarce. So, I worked on campus at UC Berkeley’s Goldman School of Public Policy in programs, development, and events. But one day, I ran into one of my favorite professors and expressed interest in research, and the rest was history.

I began a full-time research assistant position, hand-coding all of the significant legislation from 1887 to 2008 (why yes, I have read No Child Left Behind AND omnibus budget reconciliation acts, because I know how to have a good time on a Friday night), and realizing that as much as I loved studying law, what I really loved was unpacking through data collection how Congress was designing laws under different political conditions, and thinking about what that meant for public policy.

And the rest was history. Bye bye law school, hello grad school. And I felt fortunate to be able to begin my academic career at Columbia, where I built my methodological toolkit and poured over research, scouring for not just gaps in the literature, but big ideas in the field.

My academic career has been rife with ups and downs, twists and turns. There have been periods during which I’ve been far more engaged with teaching than with research, and more recently, vice versa. But while the ideas have waxed and waned, the constant has always been a love of writing.

At a department party at the end of the academic year that I finished my doctorate, the outgoing department chair said a few words to us. I don’t remember everything he said, but one thing stuck out: There are people who are prolific readers and writers, and one should be a writer.

Now, I love both dearly. Joseph Conrad once wrote, “I assure you to leave off reading was like tearing myself away from the shelter of an old and solid friendship.” As someone who reads constantly but rarely for pleasure, this resonates. There is so much simple joy in curling up on the sofa with a cat, a blanket, and a good book – whether a treatise on the American health care system, the poetry of T.S. Eliot, or a novel by Michael Chabon. What’s more, good writing is often learned by reading others’ good writing, and when I find that words aren’t coming to me as readily and the blank page feels as though it’s judging me, I crack open some Hemingway, and I find clarity in my argument and how I want to convey it.

I have always been, and always will be, a writer.

From writing an unpublished novel when I was in high school to taking advanced fiction classes at 92Y while in grad school to having a forthcoming book Coverage Denied: How Health Insurers Drive Inequality in the United States, it is in my blood.

The writing of Coverage Denied was furious. It was indignant. It was righteous anger. It was a speed that I hope I can get back again without succumbing to a manic episode. But every time I crafted a particularly good sentence or paragraph, or did a satisfying reorganization to promote clarity, or integrated relevant research that contextualized a problem’s ripple effects and made the passage “sing,” there was always satisfaction and joy.

Some social scientists have asserted that AI can do our research better than we can. The reality is that AI could not have written Coverage Denied.

Coverage Denied draws on interdisciplinary literature, administrative data, and my nationwide survey, but it also draws on the 111 semi-structured interviews that I conducted with patients, physicians, former health insurance executives, patient advocates, and health insurance lawyers. As patients and providers welcomed me into their lives, often sharing with me their vulnerabilities, I pieced together new facets of the American health care system’s complexities and failures. This rich analysis of patient and provider lives being upended by health insurance barriers simply cannot be outsourced. And this is true of most of the political science work that I admire and to which I aspire – whether the marriage of data and patient narratives in Fragmented Democracy, the conceptualization of the Divided Welfare State, and so much more.

AI has many valuable uses. It has helped me to find words to trim from an abstract that’s a smidge too long, to punch up a chapter title, to come up with some class activities to break up my lecture, to come up with a list of wrong answer choices for the multiple choice section of an exam, and to design a figure for a lecture slide so that I don’t have to fuss too much with the work that isn’t intellectually rigorous. But good writing comes from good thinking, and AI does not think creatively in new and innovative ways about health care delivery or the American welfare state or federalism. It synthesizes existing knowledge, rather than moving the field forward.

Of course, there are people in the field who are more focused on improving the precision with which we estimate more established relationships. Personally, I find a lot of those papers can be best characterized as mathematical masturbation rather than being conceptually innovative, but if they’re pushing the methodological field forward, more power to them. And when doing more squarely technical work, AI may look like a silver bullet. (I do wish to remind people that while there is a stereotype about academics knowing more and more about less and less, we don’t need to actively reinforce that perception.)

But to characterize AI as outperforming social scientists is not only a bald-faced lie, but it exposes one’s own undervaluing of creative thought and inability to discern (or disinterest in discerning) quality writing that pushes the reader to reconceptualize politics, policy, and law.

As nearly anyone who has graded a paper that was generated by ChatGPT can attest to, AI does not produce good writing. Not only does it hallucinate citations, but it fails to offer any originality or a clear point of view. It produces vague generalities without real insights. It offers a reminder that proper syntax and good writing are not the same, and as a professor I would certainly rather point out how an engaged student can refine their writing to improve their argument than grade another banal paper that is the product of intellectual laziness.

What’s more, this excessive reliance on AI leads me to the inescapable question: Why are you in this field?

Not everyone in the social sciences has an easy alternative pathway to YouGov or Meta, but the “AI will replace academics” crowd typically has the methods chops to have their pick of more lucrative industry jobs. I will never characterize a career as a calling – it is a profession, just as with law, medicine, and beyond – but it is profession that I love and to which we generally gravitate because we love to sit with ideas and converse with the literature and put pen to paper to push the field forward.

I dearly wish my students liked writing and learning the way I always have, but I expect fellow academics to, at least on balance. And we’re in the ideas-generating business. Why would we want to outsource the fun stuff?

I know exactly how I came up with the idea for my first solo paper: I was coding No Child Left Behind as a research assistant and came across a curious provision that seemed to be a workaround from the Supreme Court’s decision in US v. Lopez because it conditioned schools’ funding on guns-free school zones following the Court’s striking of the Guns-Free School Zones Act, which had been grounded in interstate commerce. So, I did a deep dive into Congress’s use of Title VI of the Civil Rights Act of 1964 to condition federal funds on regulatory compliance, potentially in domains where more direct enforcement mechanisms were foreclosed, and before I knew it, I had an article acceptance. And not only was that an obviously desirable outcome as a graduate student, but the process was fun!

Apart from being saddened at the thought of anyone wanting to outsource this enjoyable creative process, as opposed to using tools to finesse and tighten prose, I worry about the way that graduate students are thinking about academic careers when this framing in Substack posts, on social media, and beyond is becoming the model. Good teaching and mentorship inspired my pursuit of an academic career and reading good research inspires me to produce that myself, and I hope that that isn’t lost amid overreliance on these new bells and whistles that rob of us the intellectual rigor that is, or should be, our bread and butter.

I try – admittedly with imperfect success – to instill in my undergraduate and graduate students a love of learning and an appreciation of writing. And no matter what advances come our way (and of course, I dearly hope the AI bubble bursts as soon as possible), that’s not changing.

And with that, I am stepping off my soapbox. For now.