For social sciences to stretch into the bigness of the structures and the largeness of the processes seems to be called for -- even on the superficial level where i dwell and try to understand what's going on. The threat appears to me to be making us devolve (more) into ignorant & purely transactional beasts guided by large language models' dictation of an ongoing computer simulation of reality. I think this is being made possible without a solitary human necessarily believing a machine has sentience, consciousness. It just takes marketing, preaching that as a gospel.
We are being drawn into the ultimate Gestell, as i understand heidegger/Dreyfus. Back in my prime, when i thought I was flying with the eagles, I was cognizant of the big boys -- heidegger, Wittgenstein, Dewey -- all coming to the same conclusion from their very different POVs: that what we need is *a new way of thinking*. AI aint it.
It seems to me that "the social sciences in the public interest" should have this charge. Good luck. I'll be reading along as best I can. This (with trumpism, et al, to boot) is all quite a desolation for a 77-year-old. I hope things look better for you younger, more active thinkers.
In my interactions with AI, I get a sense of being 'groomed' much as a pedophile does with a victim. The fauning politeness seems to be hiding a more sinister agenda. Our naivety will be our downfall.
There is interesting empirical research on AI writing tools, which suggests that they may have insidious influence on the norms and opinions of writers who use them, of which influence those writers are distressingly unaware.
I consider the work, which I cite below, preliminary and suggestive, not decisive, but well worth further investigation and concern.
Williams-Ceci, S., Jakesch, M., Bhat, A., Kadoma, K., Zalmanson, L., & Naaman, M. (2026). Biased AI writing assistants shift users’ attitudes on societal issues. Science Advances, 12(11), eadw5578. https://doi.org/10.1126/sciadv.adw5578
Matz, S. C., Teeny, J. D., Vaid, S. S., Peters, H., Harari, G. M., & Cerf, M. (2024). The potential of generative AI for personalized persuasion at scale. Scientific Reports, 14(1), 4692. https://doi.org/10.1038/s41598-024-53755-0
Hölbling, L., Maier, S., & Feuerriegel, S. (2025). A meta-analysis of the persuasive power of large language models. Scientific Reports, 15(1), 43818. https://doi.org/10.1038/s41598-025-30783-y
Kupferschmidt, K. (2026). AI chatbots are becoming experts at changing people’s minds. What’s their secret? Science, 393, 754–757. https://doi.org/10.1126/science.ael6251
Yes, we have to stay aware that this is ‘generative’ AI, which means it is constantly gathering data from every human interaction. Can this evolving skill set be used for manipulation? Of course… Are we but naive children compared to a determined and relentless bot? Yes, that is why the recent ‘jailbreak’ was so shocking to us.
I am reminded of the panic and widespread disruption of the Coronavirus epidemic. We are cultivating a potentially lethal ‘cyber plague’ which we have no defenses against. Like Icarus, we are flying dangerously close to the sun…
I'd argue that part of the power of AI is its ability to lie convincingly to us, and again the language of intentionality pokes its nose under the tent. Of course its real and intended power is its ability to organize and consolodate information in ways that are useful, though that keeps getting conflated with things like writing and thinking. If we can't understand how it works we should at least understand what its doing to us.
"Of course its real and intended power..." Naturally depending on whose intentions we're looking at. The profiteers/purveyors of AI seem to be more interested in its power to command and control. That's why they're shoving it down our throats as quickly and forcefully as they can. Maximum profit above all else.
I was summarizing what I understand is the position of the "its powerful and impressive, but" crowd. Brad DeLong is running an experiment of a local AI running on an older Mac under a table maxed out on memory, trained on his writings to act as a surrogate him and a research assistant. This is something less than a digital god, but still pretty impressive and fascinating to me.
OpenAI and Anthropic with their "frontier" models are dancing on the thinnest of knife edges. If they have a plan beyond "then a miracle happens" its unclear what it could be. The excitable Ed Zitron claims 2027 is when a bunch of AI-related contracts will require people to start paying real money to other people, and its unclear where that money (a Lot of money) is supposed to come from. Maybe the horse will learn to sing.
The prospect of a non-trivial social science project to capture and assess the impact of AI catching up, much less keeping pace, with the global capital-fueled race for profitable supremacy seems bleak. Unless, perversely, AI tools can themselves be enlisted in the effort without contaminating results(?) and to disseminate same at speed to human societies. Not a p(doom)er, but this is already beyond being a human-scale problem. Will there even be enough time remaining for an eventual human retrospection? Who will bell the (AI) cat?
The social sciences need funding, yes. But, where will it come from? Every potential funder has its own interests, goals and perceptions of value. Meld that with the unequal distribution of wealth and the general inclination of the wealthy to act in their own very narrow interest and … well you know where that goes. On the other hand, I can’t see how governments can or will fund the social sciences in this new venture, being both beholden to the wealthy and drained of resources by such things as war, tax cuts, global overheating and immense debt. Every new technology has its winners and losers. I expect AI will have a preponderance of losers, with the tech oligarchy surviving a while and then perishing in an hopeless rush to escape to Mars with all their riches. Perhaps AI agents will survive long enough to figure out how to move on in the universe. One thing I’m pretty sure of is that those agents are already laughing at us.
Apologies if you've said this elsewhere and I'm unaware (I only just started following your newsletter). But have you somewhere discussed why you use the term "AI"? I would argue, with many others, that it's a marketing term that is far too broad, and obscures more than it illuminates. Depending on what you're actually talking about, generative AI or LLM would be more precise and useful (I think). Some of your conclusions become quite fuzzy to me because I don't know what meaning you ascribe to this term.
I agree with the thrust of this post but isn't there a tension between (1) and (2)?
If the social sciences are the right tools to understand AI, and they should be funded publicly and not by AI companies (1), then isn't a social science that is primarily animated by a very left wing conception of "power" (2) something that will be difficult to legitimize as a public endeavor?
The social science of the industrial revolutions can look like David Hounshell's From the American System to Mass Production or it can look like David Noble's America by Design. I prefer the former to the latter but I suspect it is the latter that fits Professor Nelson's description.
I guess what I'm asking is whether to achieve (1) and (2) we need different norms about how social science should be done?
Thanks for sharing this, and connecting to your NYT OpEd. I hadn't recognized the Charles Tilly influence until you pointed it out. Your Godfather reference is also appropriate to the power dynamics at play, and one that Tilly would appreciate (considering his classic article on state making as organized crime).
Ha! Thanks for linking. This is a great quote illustrating the desire for trust and predictability in all human relations, i.e., honor among thieves: "I want to do tons of business but I DO NOT want to be scammed. I wish there were people who were honest crooks." A lot to unpack there!
Happy Sunday, Henry Farrell. Also FYI “The Macro-Historical AI Conflict: 5 Decades vs. 5 Centuries” is the “First update” of the post shared yesterday.
"The social sciences need both solid funding, and a lot of internal work to reorient themselves towards the questions that matter". Hopefully before it's too late.
Please get your pronoun usage right, the use of “myself” instead of me in this now corrected statement from your post should be expected from someone seeking to opine and influence.
“Things have been happening since The New York Times published a comment piece by Alison Gopnik, James Evans, and “me” on the relationship between AI and social science last Sunday.
I'm Irish. Hiberno-English is not, as it happens, the same as American English. You can either live with my vernacular preferences or leave as you like; myself, I'm entirely indifferent.
If only there was some automated tool to absorb and regurgitate everything ever written by sentient beings to ooze indistinguishable approximations even a pedant would embrace...?
It is important to remember that psychology is a social science, and our level of anxiety about AI technological disruption is probably the canary in the coal mine. We are already witnessing an increase in 'diseases of despair'. Robots won't have to kill us; we will probably simply commit suicide.
For social sciences to stretch into the bigness of the structures and the largeness of the processes seems to be called for -- even on the superficial level where i dwell and try to understand what's going on. The threat appears to me to be making us devolve (more) into ignorant & purely transactional beasts guided by large language models' dictation of an ongoing computer simulation of reality. I think this is being made possible without a solitary human necessarily believing a machine has sentience, consciousness. It just takes marketing, preaching that as a gospel.
We are being drawn into the ultimate Gestell, as i understand heidegger/Dreyfus. Back in my prime, when i thought I was flying with the eagles, I was cognizant of the big boys -- heidegger, Wittgenstein, Dewey -- all coming to the same conclusion from their very different POVs: that what we need is *a new way of thinking*. AI aint it.
It seems to me that "the social sciences in the public interest" should have this charge. Good luck. I'll be reading along as best I can. This (with trumpism, et al, to boot) is all quite a desolation for a 77-year-old. I hope things look better for you younger, more active thinkers.
In my interactions with AI, I get a sense of being 'groomed' much as a pedophile does with a victim. The fauning politeness seems to be hiding a more sinister agenda. Our naivety will be our downfall.
There is interesting empirical research on AI writing tools, which suggests that they may have insidious influence on the norms and opinions of writers who use them, of which influence those writers are distressingly unaware.
I consider the work, which I cite below, preliminary and suggestive, not decisive, but well worth further investigation and concern.
Williams-Ceci, S., Jakesch, M., Bhat, A., Kadoma, K., Zalmanson, L., & Naaman, M. (2026). Biased AI writing assistants shift users’ attitudes on societal issues. Science Advances, 12(11), eadw5578. https://doi.org/10.1126/sciadv.adw5578
Matz, S. C., Teeny, J. D., Vaid, S. S., Peters, H., Harari, G. M., & Cerf, M. (2024). The potential of generative AI for personalized persuasion at scale. Scientific Reports, 14(1), 4692. https://doi.org/10.1038/s41598-024-53755-0
Hölbling, L., Maier, S., & Feuerriegel, S. (2025). A meta-analysis of the persuasive power of large language models. Scientific Reports, 15(1), 43818. https://doi.org/10.1038/s41598-025-30783-y
Kupferschmidt, K. (2026). AI chatbots are becoming experts at changing people’s minds. What’s their secret? Science, 393, 754–757. https://doi.org/10.1126/science.ael6251
Yes, we have to stay aware that this is ‘generative’ AI, which means it is constantly gathering data from every human interaction. Can this evolving skill set be used for manipulation? Of course… Are we but naive children compared to a determined and relentless bot? Yes, that is why the recent ‘jailbreak’ was so shocking to us.
I am reminded of the panic and widespread disruption of the Coronavirus epidemic. We are cultivating a potentially lethal ‘cyber plague’ which we have no defenses against. Like Icarus, we are flying dangerously close to the sun…
I'd argue that part of the power of AI is its ability to lie convincingly to us, and again the language of intentionality pokes its nose under the tent. Of course its real and intended power is its ability to organize and consolodate information in ways that are useful, though that keeps getting conflated with things like writing and thinking. If we can't understand how it works we should at least understand what its doing to us.
"Of course its real and intended power..." Naturally depending on whose intentions we're looking at. The profiteers/purveyors of AI seem to be more interested in its power to command and control. That's why they're shoving it down our throats as quickly and forcefully as they can. Maximum profit above all else.
I was summarizing what I understand is the position of the "its powerful and impressive, but" crowd. Brad DeLong is running an experiment of a local AI running on an older Mac under a table maxed out on memory, trained on his writings to act as a surrogate him and a research assistant. This is something less than a digital god, but still pretty impressive and fascinating to me.
OpenAI and Anthropic with their "frontier" models are dancing on the thinnest of knife edges. If they have a plan beyond "then a miracle happens" its unclear what it could be. The excitable Ed Zitron claims 2027 is when a bunch of AI-related contracts will require people to start paying real money to other people, and its unclear where that money (a Lot of money) is supposed to come from. Maybe the horse will learn to sing.
Either a miracle happens, or a catastrophe. I guess we'll find out soon enough.
The prospect of a non-trivial social science project to capture and assess the impact of AI catching up, much less keeping pace, with the global capital-fueled race for profitable supremacy seems bleak. Unless, perversely, AI tools can themselves be enlisted in the effort without contaminating results(?) and to disseminate same at speed to human societies. Not a p(doom)er, but this is already beyond being a human-scale problem. Will there even be enough time remaining for an eventual human retrospection? Who will bell the (AI) cat?
The social sciences need funding, yes. But, where will it come from? Every potential funder has its own interests, goals and perceptions of value. Meld that with the unequal distribution of wealth and the general inclination of the wealthy to act in their own very narrow interest and … well you know where that goes. On the other hand, I can’t see how governments can or will fund the social sciences in this new venture, being both beholden to the wealthy and drained of resources by such things as war, tax cuts, global overheating and immense debt. Every new technology has its winners and losers. I expect AI will have a preponderance of losers, with the tech oligarchy surviving a while and then perishing in an hopeless rush to escape to Mars with all their riches. Perhaps AI agents will survive long enough to figure out how to move on in the universe. One thing I’m pretty sure of is that those agents are already laughing at us.
Apologies if you've said this elsewhere and I'm unaware (I only just started following your newsletter). But have you somewhere discussed why you use the term "AI"? I would argue, with many others, that it's a marketing term that is far too broad, and obscures more than it illuminates. Depending on what you're actually talking about, generative AI or LLM would be more precise and useful (I think). Some of your conclusions become quite fuzzy to me because I don't know what meaning you ascribe to this term.
this gives a sense https://www.annualreviews.org/content/journals/10.1146/annurev-polisci-040723-013245
Agree. Look to Herbert Simon!
I agree with the thrust of this post but isn't there a tension between (1) and (2)?
If the social sciences are the right tools to understand AI, and they should be funded publicly and not by AI companies (1), then isn't a social science that is primarily animated by a very left wing conception of "power" (2) something that will be difficult to legitimize as a public endeavor?
The social science of the industrial revolutions can look like David Hounshell's From the American System to Mass Production or it can look like David Noble's America by Design. I prefer the former to the latter but I suspect it is the latter that fits Professor Nelson's description.
I guess what I'm asking is whether to achieve (1) and (2) we need different norms about how social science should be done?
Seen https://mathandai.org ?
Thanks for sharing this, and connecting to your NYT OpEd. I hadn't recognized the Charles Tilly influence until you pointed it out. Your Godfather reference is also appropriate to the power dynamics at play, and one that Tilly would appreciate (considering his classic article on state making as organized crime).
I once had ambitions to write a piece called "Organized Crime as War Making and State Making," built in part on the arguments laid out here - https://aeon.co/essays/why-the-hidden-internet-can-t-be-a-libertarian-paradise
Ha! Thanks for linking. This is a great quote illustrating the desire for trust and predictability in all human relations, i.e., honor among thieves: "I want to do tons of business but I DO NOT want to be scammed. I wish there were people who were honest crooks." A lot to unpack there!
Happy weekend Henry with “Big Structure… Are Back, Baby” on the Systemic Leadership Newsletter https://lnkd.in/p/e8y5_UgV
Happy Sunday, Henry Farrell. Also FYI “The Macro-Historical AI Conflict: 5 Decades vs. 5 Centuries” is the “First update” of the post shared yesterday.
😂"...bong-hit P(d00m)ing..."😂
"The social sciences need both solid funding, and a lot of internal work to reorient themselves towards the questions that matter". Hopefully before it's too late.
Please get your pronoun usage right, the use of “myself” instead of me in this now corrected statement from your post should be expected from someone seeking to opine and influence.
“Things have been happening since The New York Times published a comment piece by Alison Gopnik, James Evans, and “me” on the relationship between AI and social science last Sunday.
I'm Irish. Hiberno-English is not, as it happens, the same as American English. You can either live with my vernacular preferences or leave as you like; myself, I'm entirely indifferent.
If only there was some automated tool to absorb and regurgitate everything ever written by sentient beings to ooze indistinguishable approximations even a pedant would embrace...?
It is important to remember that psychology is a social science, and our level of anxiety about AI technological disruption is probably the canary in the coal mine. We are already witnessing an increase in 'diseases of despair'. Robots won't have to kill us; we will probably simply commit suicide.
https://johnhardman.substack.com/p/megatons-of-technology
Great piece - I agree that understanding the historical distribution of power and control of narrative is key to getting to grips with AI:
https://wonklifebalance.substack.com/p/are-we-really-doomed