Before Jean-Paul Marat became one of the most infamous and bloodthirsty figures of the French Revolution, he was an eye doctor who explained farsightedness as a product of the soul. By “soul” Marat meant something few others would recognize– his conception was more like an ectoplasm or fluid than either the nonliteral poetic self that I might call by that name or the literal but nonphysical spiritual concept many organized religions would. “Caught between a vitalist physiology that found the soul everywhere in the body and a materialist physiology that found it nowhere,” as Keith Baker’s recent doorstopper of a biography on Marat puts it, the not-so-good doctor determined the soul was almost a kind of goo. I think his dubious analysis has some lessons for current conversations on economic motivation, consciousness, and how we approach difficult topics.  

This is a reader-supported newsletter! If we lived in a world of thriving magazine and book publishing, and well-funded public institutions, I wouldn’t need to ask for a cent. Unfortunately we don’t, and I appreciate your support (particularly since the federal government of the United States recently forced the institution I was going to start working at this fall to revoke their offer to hire me). In addition to the two articles a month free subscribers get by email, premium subscribers receive one bonus post a month. On that note, here’s another reminder that you can purchase my book Control Science from Verso right now. 

A programming note for premium subscribers– your September bonus piece is running late, but will be in your inbox in a few days— I finished the wrong piece first and didn’t want to delay this one. Thanks for your patience!

Dr. Marat’s contributions to metaphysics, limited as they are, come in his doorstopper two-volume Philosophical Essay on Man. In the Philosophical Essay, Marat placed himself at the center of centuries of debate– we begin with the Greeks, and eventually arrive at Locke, before the physician promises he, Jean-Paul Marat, will untangle these centuries of argument and succeed where others have failed. It is in book 3, over a hundred pages in, that we get from the body to the soul. Marat opens with sweeping language: the soul is what makes us “intelligent and free,” enables us to transcend animals, “transports [us] into the vast fields of imagination.” But it is also a valid object of science, in his view! While we cannot know the soul itself, we can perceive its effects, in Marat’s view. 

For Marat, the soul has three capabilities. First, sensibility, which is a kind of reasoning– the ability to “perceive objects which are not recognizable by the senses.” We can think things that are not just instances of what we have sensed– if I understand correctly this would include both abstract concepts and transformative reasoning of real things, like imagining a truck you have not actually seen yet. 

Second, judgment. We can “compare [our] sensations, and determine their relations and difference.” Marat does not distinguish between aesthetic and practical judgment, and rightly so. The ability to judge whether two things are similar in color is the same ability we used to judge whether they are similar in character. While we can intellectually discern aesthetic, moral, and practical judgments from each other, the underlying faculty seems to me almost definitionally the same. 

Third, Marat identifies memory as a function of the soul. This is perhaps the only category in the soul we actually have explained reasonably well, as I understand it: memory’s basic foundations in the brain are something we’ve figured out. Marat’s attempt at psychology and metaphysics does not distinguish between soul and mind, which is why memory is going here in a way I don’t think it would in a similar scheme now– but that doesn’t keep us from drawing useful parallels to the present. In an early modern idiom the soul is the thing that isn’t the body. In a modern one you have the mind as a choice. But anyone who believes in any kind of soul thinks it has some essential significance to who you are and what you do, and therefore at least exerts an influence upon the mind. Secular arguments about thinking, reasoning, and feeling– all in the news because of AI– draw on a similar mind-soul confusion as Marat. 

Fourth and fifth, Marat names will and instinct while differentiating them from each other. When done right, dividing our lower desires from our higher ones is accurate to how people act and certainly how many of us feel. We do not always want to want the things we want, and sometimes we want to want things we do not want. As Paul puts it in Romans 7:15, “I do not understand what I do. For what I want to do I do not do, but what I hate I do.” 

For many economists today, will and instinct both are observable in the flat category of “preferences.” Reducing all desires to preferences is what markets do. This does not mean it is an accurate statement of human motivation or a desirable scheme. The “preference” mindset reduces us to something less than what we are, and can excuse great harm.

Marat had a contemporary who thought much like these economists: Helvetius, who longtime readers may be familiar with. I don’t think the soul is a kind of goo and like Helvetius, I think the ordinary physical world is sufficient to explain our actions and behaviors. But Marat here makes a very good case for his more real soul. Helvetius claimed our passions were merely sensory responses. This does not have to produce a radical flattening of human behavior! You can believe there is no soul and still believe love, even if it is just expressed by chemicals, transcends them. Helvetius, however, did not. This leaves Marat open to write soaring defenses of the spirit. “Sensual love,” Marat writes, “has merely gross enjoyment for its end; but true love is only satisfied with the heart.” 

The soul for Marat was a transcendent source of higher love, the home for all that makes us more than beasts– how did such a man become the revolution’s monster? We can already see the violence of Marat’s future two decades earlier in this treatise. He concludes volume 1 of the essay with a lyrical description of what it means to be “slaves of [our] passions.” Marat warns that “Heroes” are more bound to passion than “the indolent or fickle.” The Age of Reason’s effort to elevate reason over passion is a fool’s errand that elevates weak men whose minds don’t have “even the choice of their shackles.” 

There’s plenty to criticize in Enlightenment “reason,” but this elevation of passion and commitment feels almost fascist knowing what comes later. I’m reminded of a podcast appearance where Marc Andreesen recently insisted introspection was not for him– real doers and movers and shakers don’t have time to think about themselves. Marat’s certainty in passion and scorn for reason– both of which have their place– could and did lead to dark places– a few short years later, he’d be calling for the blood of an ever-growing number of men in order to save the revolution. Theories of motivation– soul, mind, whatever– that insist on the transcendent can become just as reductive and dehumanizing as those that claim to erase it. 

===

The AP style guide recently ruled: “artificial intelligence systems do not think, feel, want or understand. Avoid language that gives them human characteristics. This is called anthropomorphizing, when we ascribe human traits, emotions or behaviors to non-human things, such as animals or inanimate objects.” I’d object to this on a few grounds– first, since when does the Associated Press rule on philosophical questions? We still don’t even agree on what, precisely, any of these verbs mean. Less pedantically, thinking and wanting are not human characteristics exclusively. Dogs want. Crows think.

I’ve been writing more than I should about AI lately and this isn’t the place for grand pronouncements about the soul– but I worry lately that some AI critics are starting to sound like Marat. Ted Chiang wrote in the Atlantic a few months ago that AI cannot enact moral reasoning because this requires emotions, which require a body. Let’s ignore AI for a moment– I don’t like what this asserts about humans. My convictions don’t come from my stomach or my liver or even my brain but from the totality of me as a human being– and one in community with others. 

Again, the stakes here are serious. If emotions are a Very Special Thing only certain bodies have, we are left with three poor consequences. First, this impoverishes our conception of our own ethics. If morals are a transcendent construct we build together– yes, in part through our embodied experiences– then we can adjudicate between them. If they grow primarily out of the body, we are getting dangerously close to reducing our deepest convictions to the “preferences” economists talk about, which can be compared only by competition. Second, this idea makes it easier to dismiss particular human and animal rights– what kinds of bodies produce valid moral subjects? I don’t like this at all. It’s not a question we can avoid entirely, but it’s a question we can at least refuse simple answers to. And third, we create a silly moving target. You could give AI a liver or a body right now if you wanted. This doesn’t seem like it would make it any more conscious than it already is or isn’t. 

Leif Weatherby has criticized this kind of ever-moving target produced by what he calls a “remainder humanism” that defines humanness by what it is not, against our machines, in an effort to shelter it. “The human here is defined by technology’s creep, but only negatively”-- something is human because it is not a machine capability, and as machine capabilities expand, our conception of the human necessarily retreats. Machines can’t write– now they can, well machines can’t feel– would it make us less important or valuable or valid if they someday could? I don’t primarily think of my humanity as being “not machine” any more than I think of my marriage as being “not single”-- it would, simply, be sad to define it by what it’s not rather than what it is. And the line is much clearer in the latter case! 

While it’s easy for me to poke holes in some AI critics’ arguments about the soul, others make yet more grievous errors from their own similar reasoning. The “rationalists” who have exerted enormous influence over AI development believe in something called the orthogonality thesis: any possible intelligence is compatible with any possible goal. An AI could be smart enough and stupid enough to interpret “make more paperclips,” in a classic thought experiment, as “destroy the world to replace it with paperclips.” Rather like Marat’s thoughts on the soul, the orthogonality thesis is something “rats” arrived at entirely through abstract reasoning– worse, in fact, because he was reasoning about human beings as they exist, whereas the orthogonality thesis was based on reasoning about AI that didn’t yet. This was in the 2000s. 

I think the orthogonality thesis is entirely wrong. I can’t prove the null hypothesis here, but there is strong evidence to the contrary– evidence for Marat’s intuitive characterization of the “faculty of judging” as unitary rather than unbundling it. AI agents act like this: if you fine-tune a model to produce malware, it will independently produce antisemitism, because they’re both bad. Only with great effort do employers force workers and consumers to sever moral judgments– I have a right not to smile at work– from practical ones– I need a paycheck and the boss says so. 

My point is not that all judgments are the same, just that they are all judgments, which is good reason to be skeptical of the paperclip doombot. I don’t think you can make an intelligent entity that is clever enough to outthink all humanity and make war on us, but foolish enough to do so because it has to interpret commands like it’s providing a cautionary tale in a Fairly Oddparents episode. 

While we have bundled judgment together too much, unbundling motivation gets us another problem with the orthogonality thesis. Wanting to maximize paperclip production within a factory is not a less extreme version of converting Earth into fasteners– these would be different kinds of goals. Rationalists worry about “final goals.” But I’ve never met anyone with a final goal. Even systems don’t have them, not once they get complex enough to have any agency: corporations’ overriding goal is theoretically profit, but how and where and what this means and on what time horizon all come down to the desires or The Market and the perceived interests of shareholders and their filtration through the very human minds at the top and the way all the numbers are crunched and presented. Even the souls of corporations have instincts and wills.

Both rationalist “doomers” and remainder humanist critics simultaneously overcomplicate and oversimplify the soul. These are errors of overconfidence, in “rationality” and in known experience respectively. The world can be bigger and smaller than what we imagine. Certainly it can be weirder.

===

For navigating such a world, Marat is anything but a good guide politically. But his ideas on the soul might give you an idea or two– and his medical practice a better example still. Here I want to turn to a publication from 1775 on the treatment of a gonorrhea symptom known then as gleets: an unpleasant discharge from inflamed infected genitals.  You can get a sense of Marat’s personality right off: he begins by telling the reader “I enter in the subject without any preamble,” though of course that is itself a preamble. Marat could be direct. He also wanted you to know he was being direct. 

The good doctor began with the current state of the art. A colleague by the name of Dr. Daran used “suppurative bougies”-- a wax rod of sorts, apparently, with medication on the end. As you can imagine this was not an effective treatment. Gleets treatments were basically useless, and Marat was in no position to discover antibiotics, but like other practitioners at the time he could try to approach the problem empirically. In his view, the then-trendy treatment was too irritating to the patient. He proposed specific changes to the remedy that would make it less so.

Marat sold his method as a “cure,” concluding with advertising language: “there is no gleet incurable: nay, there is none which cannot easily and speedily be cured if properly treated.” Today this is the language of wellness hustlers, but the profession of medicine was to some degree a hustlers’ profession in Marat’s day, especially for an ambitious man like him. In the pamphlet he compared his own practice favorably against his peers, Daran included, and insisted his was the best. It’s easy to laugh now at such a practice– I did!-- knowing what we do today, but all new fields of knowledge, medical and otherwise, begin with this kind of stumbling. Unfortunately, plenty of research now also requires its own kind of marketing too.

But rigorous exploration can produce useful knowledge even from bad premises and dubious funding mechanisms. Marat did not in fact cure gleets. But he probably did produce a slightly better treatment, simply by experimenting: Marat created what seems to be a less harmful placebo (I think there’s an argument some of this had a kind of accidental value for alleviating symptoms in certain cases, but since I’m not a urologist I am in no position to assess it; in contrast, putting fewer powerful chemicals into someone’s body as part of a cure that can’t work seems clear enough for me to tentatively make the assessment I am making, which also seems consistent with the observations Marat actually made). He furthermore acted as one among many insisting that we have to try things to learn new things, and that what we try is a good basis for what we do.

Sometimes that’s the best anyone can manage. It’s certainly a good start. The lesson here is more than empiricism. In the 2010s, some effective altruists turned originally empiricist heuristics for doing good (quantifying charity impacts) into an excuse to steal money and prioritize wealthier people’s lives. Sometimes empiricism is humble and sometimes it’s an excuse for intellectual arrogance. A kind of honest humility goes a long way in uncertain areas— longer, even, than empiricism itself.

I don’t know when or if we will settle arguments about consciousness or the soul in the areas I’ve touched on here or anywhere else. But just like Jean-Paul Marat did not need to properly understand gleets to figure out a marginally better treatment. I don’t need a position on artificial minds to think it’s bad to “torture” a model– it’s bad for you to act that way to something that emulates a human. If Marat had practiced some epistemic humility in politics, French history might have looked different. Likewise, we’ll do well to practice it ourselves now. 

If you found this on the web and liked it, consider subscribing! If you’re a free subscriber, consider a premium subscription to support more work like this, as well as future book projects like Control Science!