AI and Tocqueville at Harvard
David Shaywitz
I found myself this week back on my undergraduate campus, giving a seminar about “AI, Health, and Agency” for, of all things, Harvard’s Program in Constitutional Government, invited by Professor (emeritus) Harvey C. Mansfield, who at 94, remains sharper than most of us are or likely ever were.
He is not only one of the university’s most distinguished scholars, but also, famously, one of its only (openly) conservative faculty members — and also one of the most demanding.
Back when I was an undergraduate (in the Days of Giants), he was known as “Harvey C-Minus Mansfield” because of his tough grading. Even after grade inflation became effectively mandatory, he reportedly would provide students with two scores – their official Harvard grade along with what he thought they actually deserved.
Professor Harvey C. Mansfield and David A. Shaywitz
For years I knew him only by reputation, and was intrigued by his extraordinary scholarship and independent thinking. At some point, I reached out with a question relating to a perspective on history that I had attributed to a Harvard political science professor but couldn’t locate a specific citation. It didn’t ring a bell for him, but he pointed me towards other, far more useful references, including Tocqueville, one of his areas of great expertise. An occasional correspondence ensued, and more recently, the invitation to speak.
As I prepared for the talk, I landed on a question I wanted to ask him, related both to agency and to Tocqueville; it’s his thoughtful response – related not to agency but to AI – that I want to share with readers.
First, some context.
Tocqueville
“Democracy in America,” write Mansfield and the late Delba Winthrop in their Editors’ Introduction to their definitive translation, “is at once the best book ever written on democracy and the best book ever written on America.”
Tocqueville, accompanied by his friend Gustave de Beaumont, had journeyed from France to America, and together, the two men spent nine months from 1831-1832 visiting as much of the country as they could, eager to see, as Mansfield and Winthrop explain, “what a great republic was like.”
Tocqueville would publish his observations in two volumes: the first, in 1835, was “a sensational success,” Mansfield has written, but the second, in 1840, was “received without much enthusiasm.”
Tocqueville’s “Democracy in America”
Nevertheless, Volume Two includes one of his most memorable essays (Chapter 20) contrasting what he describes as “aristocratic” historians and “democratic” historians, establishing a tension that resonates to this day.
Aristocratic historians, Tocqueville suggests, overindex on the influence of individual people, taking what we might now term a “Great Man” view of history.
Such historians, he says, “ordinarily make all events depend on the particular wills and humors of certain men.”
When aristocratic historians “cast their eyes on the theater of the world,” Tocqueville writes, “they perceive first of all a very few principal actors who guide the whole play.”
Focusing in this way, “gives them an exaggerated idea of the influence that a man can exert and naturally disposes them to believe that one must always go back to the particular action of an individual to explain the movements of the crowd.”
Democratic historians, on the other hand, “attribute almost no influence to the individual over the destiny of the species or to citizens over the fate of people,” Tocqueville asserts.
These historians tend to see “the actors much less and the acts much more.”
Tocqueville worries that in adopting this approach, democratic historians “not only deny to a few citizens the power to act on the destiny of the people, they also take away from peoples themselves the ability to modify their own fate, and they subject them either to an inflexible providence or a sort of blind fatality.”
Mansfield suggests that the explanations offered by democratic historians “imply that man is not a master of himself, he is not the master of events, thus not free.”
Democratic historians, Mansfield continues, “seem determined to show that progress is not a goal, achieved consciously, voluntarily by human beings,” and cites the observation by Tocqueville (also in Chapter 20) that “Historians of antiquity instruct on how to command, those of our day teach hardly anything other than how to obey.”
Astutely, Mansfield observes that this formulation, while leveling most citizens, elevates the historians themselves: they “appear great…seeming to dispose of the great causes they describe and complacently looking down on the rest of humanity unaware of the forces driving them forward.”
This air of righteous condescension seems easily recognizable within much of contemporary academia.
Projecting forward
It doesn’t take much imagination to see a remarkably similar tension to that described by Tocqueville expressed today, particularly between approaches (typically from the right) that focus on elevating individual achievement and those (more often from the left) emphasizing the need to reduce disparities.
One example: increasingly divergent views of agency. Martin Seligman’s just-published book, Agency: The Psychological History of Human Progress (expanding on this 2021 publication) argues that agency is the hidden engine of civilizational advancement. Not surprisingly, Silicon Valley, in particular, has latched onto the ideal of agency, spurring critics like Sophie Haigney, who wrote a New York Times op-ed entitled “All the Worst People Seem to Want to be High Agency.”
Many academic leaders who were influenced by Seligman nevertheless are increasingly emphasizing the role of the social environment in enabling agency.
One example: University of Pennsylvania scholar Angela Duckworth’s latest book, “Situated,” focuses on what she calls “situational agency,” emphasizing the contributions of your environment, and the importance of shaping your surroundings.
Another example: in a recent correspondence, a positive psychology expert from Harvard’s School of Public Health asked “How do we ensure the social conditions are in place to enable people to have sufficient agency? It’s not a viable choice for everyone, is it?”
Mansfield, Tocqueville, and AI
While the contrasting views of agency seemed to me an echo of Tocqueville’s two types of historians, Mansfield offered what struck me as an even more important question: is the impact of AI likely to be more “democratic” in effect or more “aristocratic,” each in the Tocqueville sense.
In other words, will AI affect everyone similarly, potentially helping everyone but resulting in a compression of differences, or will it provide an opportunity for the peaks of human performance and achievement to extend even higher – potentially at the risk of exacerbating differences?
My initial response was to emphasize both the opportunities to elevate everyone as well as the particular potency of technology when wielded by people who are already domain experts – not least because such experts are better able to guide it, to recognize where it’s falling short, and redirect it skillfully.
Mansfield wasn’t having it.
“Do you really think that if Melville had this technology, he’d write a better version of Moby Dick?” he asked rhetorically.
I didn’t think so either -– but, as Ethan Mollick has repeatedly observed, as recently as his latest blog post, “AI is still quite a bad long-form writer” –- it’s a lagging area of expertise.
In responding to Mansfield, I highlighted that in the hands of expert software developers, AI seemed to make the best ones even more productive and effective, and to enable users to contemplate more ambitious projects than they might have otherwise considered possible.
But hand-waving explanations aside, I think Mansfield’s asking a really important question (and perhaps two): what’s the likely pattern of benefit for AI — widely distributed or spiky –- and (more provocatively, perhaps) what pattern would we prefer to see?
I actually confronted a version of this question when I was tasked a few years ago with developing the AI strategy for a large pharma’s R&D organization. After an engaging and remarkably encouraging “listening tour” across the organization, I proposed a strategy focused on concentrating future AI resources around a handful of deeply committed individuals and teams who had already taken it upon themselves to pursue particularly promising approaches and had started to make real progress.
This strategy -– essentially, make the peaks higher — was explicitly rejected by both the head of the R&D Digital and Data group and by the company’s head of R&D, who both emphasized their preference for an approach that would “elevate everyone 5%,” rather than supercharge (gratuitously, in their minds) those who already “got it.”
In my view, the key to organizational adoption was actually demonstrating tangible, palpable success in some domain, which not only would organically motivate others in the organization (people gravitate towards conspicuously useful tools and technologies), but would also separate this from yet another perfunctory corporate innovation initiative. I lost the argument, but not my conviction in the approach.
More AI Questions, More Tocqueville
Back to the seminar: in a highly engaged discussion following the talk, attendees raised a number of thoughtful questions (albeit not related to Tocqueville). Two in particular stood out:
- An ethics professor who teaches MIT undergraduates shared that her students are very concerned about AI taking jobs, and asked how to counsel them.
- A student (I assume) pointed to the recent solution of a raft of difficult math proofs by AI, and asked how we think about a world where people (even if in the loop) can’t evaluate the quality or accuracy of what the technology is delivering.
What I took away from these questions (and many others) was deep appreciation for just how dynamic, exciting and unsettled moment we are all now in with respect to AI. Everyone seems to sense that big changes are ahead, but just what these changes will be, and how things will eventually settle out is completely unknown –- no matter what tech VCs confidently assert, and what playbooks management consultants inevitably provide.
You might say that we are like Tocqueville, as he reflects in his final chapter of Democracy in America on the dizzying dynamics of an evolving country in the throes of tumultuous change, trying desperately, eagerly, hopefully to determine its identity, and define its future.
“The new society that I sought to depict and that I wish to judge,” Tocqueville writes, “is only being born. Time has not yet fixed its form; the great revolution that created it still endures, and in what is happening in our day it is almost impossible to discern what will pass away with the revolution itself and what will remain after it.”
He continues, “The world that is arising is still half entangled in the debris of the world that is falling, and in the midst of the immense confusion that human affairs presents, no one can say what will remain standing of aged institutions and old mores and what of them will in the end disappear.”
Contemplating all he has observed, Tocqueville writes “I feel full of fears and full of hopes. I see great perils that it is possible to ward off; great evils that one can avoid or restrain, and become more and more firm in the belief that to be honest and prosperous, it is still enough for democratic nations to wish it.”
A contemporary, somewhat optimistic observer, surveying the current state of AI, could hardly have offered a more apt description.
