AI has reproduced, yet again, the idea that the median view of the US political class has some normative claim to centrality. Both Grok (far-right) and DeepSeek (CCP) have shown that it is perfectly possible to give answers reflecting different background assumptions. But people still "Ask ChatGPT" and expect to receive objective truth.
A central question here is: which country is being represented? On the assumption that the US is already a write-off, it's important to develop digital representations of the democratic world which don't depend on the whims of fascist billionaires.
Well, thanks; this was eye-opening for me in this way: that there are idealists endeavoring to have this powerful thig do good.
My impression from reading the statements of the 'big important men' in ai is that there's a thinking that they are creating, even incarnating, a powerful thing that's under their control. It will consume terabytes of electric energy to do so, with consequences for most of our fellow rate-base members. As in higher costs or routine disconnects from the grid for those who've found some cost relief by signing up for the peak- load- period disconnect rate. There's not enough capacity to keep grandma warmed or cooled to feed Andreesen's new theology.
To civic representation: I've known for twenty+ years that there was sufficient computing power in ESRI's ArcView GIS software that any reasonably- sized civil engineering firm or local or state government agency could take Census survey data and crank out rightly apportioned, logical, easily understood voter districts to more equitably insure a representative distribution by area and population. A map, if you will. And it could be done by a small team or teams within a year, and updated every ten years in a matter of days.
But that doesn't allow for the jiggery-pokery of people who wish to concentrate power for their particular purposes- which more and more appear to be about the accumulation of unholy concentrations of money. Which is what it appears that the owners of this new cool thing want now for themselves.
Sincerely, thanks for this discussion. Think I'm going to pick up another good Aladdin kero lamp from the Amish hardware store, and another five gallons of K-1 in addition to finishing up my woodshed project and split and stack the two cords of free -fall oak and ash. Winter's coming.
The lossy compression-descriptive representation comparison is fascinating. I’d add “representative samples” in polling to that as well. Using non-random samples to create a more “representative” result is a Sisyphean task. Every polling sample is metaphorically an eye dropper full out of a free-flowing data river at a point in time, so prone to error, bias, and unidentified independent variables as to make it meaningless in isolation. I’m not informed enough to flesh this out.
Also, instead of the “big data” we heard so much about a decade ago, maybe we need a “small data” revolution to get at nuance and variation. What’s the point of an AI that gives a half-assed big picture when the whole point of digital computation is to do all the small calculations that inform the whole? IMO, AI is going to suck at thinking specifically until it reaches into small data rather than just LLMs.
So I read through the paper and found references to "AI" to be very vague and speculative. Which is appropriate, I suppose, for a very vague and speculative technology. If AI is doing the sortition and defining the terms of the reasoned debate, then the process is at the mercy of the black box at its heart synthesizing its input data. As it always turns out, the issue is not AI but the how it's output is consumed by the humans. In a way it's totally appropriate and clever for Silicon Valley to pitch AI as a benevolent objective God. There's a deep rooted human need to kick all of our most vexing problems upstairs to a Great Tribal Chieftan in the Sky to solve. That's the basic definition of religion.
There's no question the purveyors/profiteers of AI want to do away with any democracy that doesn't put them in complete control. The "Dark Enlightenment" is upon us.
Doesn't this apply to democracy in general? At best, there is a very limited coupling between voters and their elected representatives. At worst, votes are mere proxies for campaign finances and political capital. If democracy pins its legitimacy on representatives "acting for" voters, it seems that legitimacy is weakening day by day.
There is a large literature on this and related issues in Science and Technology Studies e.g. the work of Sheila Jasanoff and Brian Wynne. Weirdly narrow democratic debate related to science and technology is not new.
See for example: Welsh, Ian, and Brian Wynne. 2013. “Science, Scientism and Imaginaries of Publics in the UK: Passive Objects, Incipient Threats.” Science as Culture 22 (4): 540–66. https://doi.org/10.1080/14636778.2013.764072.
I don't think this limited-view-of-democracy problem is limited to AI field, though I think your point around engineers' skills and personalities making it bigger there is right on.
I have a close friend who does community engagement consulting (e.g. for municipalities or large-scale NGO's), and that world has a similar blind spot. The entities contracting for community engagement simply don't question WHO makes decisions, and only focus on what information the decision-makers (often themselves) have. I think of it as a general cultural tendency to assume that bad decisions are the result of poor information or technical expertise rather than the result of (for example) simply caring more about what real-estate developers want than what poor & working class residents want.
Thanks for writing and sharing this. A note: it would help some of us readers if you explained abbreviations the first time through (LLM/large language model) and identify the “it” when explaining the asterisk at the end (Hanna Pitkin’s The Concept of Representation). Again thank you for sharing. I sincerely appreciate it.
I see no use for AI in my garden or in any endeavor. Send AI to Mars. Let it sort through whatever data its idealistic makers provide, keep itself busy factoring out the weird and significant deviations that some of us cherish, and languish on that planet. Get AI out of my energy grid. Colossal waste of space and resources. Let's feed people.
You are correct to say that the AI democracy debate is weirdly narrow, but I think it’s even narrower than what you imagine.
I’m part of a group that believes we can actually measure public opinion, and we also believe we have worked out the problem of implementation. it’s not direct democracy, but it’s a big step in that direction.
In regard to changing the political landscape, unfortunately relying on our present day political systems does not work. The track record is bad. We can’t even get a small incremental change like rank choice voting.
The problems we have in the world today cannot wait for the slow pace of today’s political systems, even if they make incremental changes, and big changes like sortition are extremely unlikely.
AI is obviously going to play a big role in our future, but there is no way to implement it without bias. We are solving this problem by isolating the data in its own publicly owned institution.
All analysis of the data is controlled by the users. Users will use all methods made available to them by the free market.
We have a website, but it’s still a little clunky right now, it’s a work in progress, so I would suggest that if you want to know more about what we’re working on, check us out on Reddit. simply look for KAOSNOW.
This brings back ancient memories of reading and arguing against Habermas on democratic deliberation. There were two missing pieces, IMO, both of which are foregrounded in this post: power and party dynamics. Democracy is shaped by power concentrations, especially financial interests. You can't promote a more ideal democracy without promoting a more democratic -- and pluralist -- distribution of wealth. Meanwhile, political parties, organized around some combination of interest and ideology, are inescapable and, in some form, necessary. How they articulate with democratic procedures is huge. Indeed, the core political flaw in legacy socialism is its inability to accommodate multiple, contentious parties.
I don't see how folding in AI changes any of this.
Reminds me of the way some medical professionals generalize from studies without much attention to the distracting complexities and contradictions of the individual patient.
AI has reproduced, yet again, the idea that the median view of the US political class has some normative claim to centrality. Both Grok (far-right) and DeepSeek (CCP) have shown that it is perfectly possible to give answers reflecting different background assumptions. But people still "Ask ChatGPT" and expect to receive objective truth.
A central question here is: which country is being represented? On the assumption that the US is already a write-off, it's important to develop digital representations of the democratic world which don't depend on the whims of fascist billionaires.
No filter = far right. (And it still has kind of a filter.)
Applying a very meticulous filter = progressive.
Well, thanks; this was eye-opening for me in this way: that there are idealists endeavoring to have this powerful thig do good.
My impression from reading the statements of the 'big important men' in ai is that there's a thinking that they are creating, even incarnating, a powerful thing that's under their control. It will consume terabytes of electric energy to do so, with consequences for most of our fellow rate-base members. As in higher costs or routine disconnects from the grid for those who've found some cost relief by signing up for the peak- load- period disconnect rate. There's not enough capacity to keep grandma warmed or cooled to feed Andreesen's new theology.
To civic representation: I've known for twenty+ years that there was sufficient computing power in ESRI's ArcView GIS software that any reasonably- sized civil engineering firm or local or state government agency could take Census survey data and crank out rightly apportioned, logical, easily understood voter districts to more equitably insure a representative distribution by area and population. A map, if you will. And it could be done by a small team or teams within a year, and updated every ten years in a matter of days.
But that doesn't allow for the jiggery-pokery of people who wish to concentrate power for their particular purposes- which more and more appear to be about the accumulation of unholy concentrations of money. Which is what it appears that the owners of this new cool thing want now for themselves.
Sincerely, thanks for this discussion. Think I'm going to pick up another good Aladdin kero lamp from the Amish hardware store, and another five gallons of K-1 in addition to finishing up my woodshed project and split and stack the two cords of free -fall oak and ash. Winter's coming.
Tim Long, just up the hill from Lock 15.
The lossy compression-descriptive representation comparison is fascinating. I’d add “representative samples” in polling to that as well. Using non-random samples to create a more “representative” result is a Sisyphean task. Every polling sample is metaphorically an eye dropper full out of a free-flowing data river at a point in time, so prone to error, bias, and unidentified independent variables as to make it meaningless in isolation. I’m not informed enough to flesh this out.
Also, instead of the “big data” we heard so much about a decade ago, maybe we need a “small data” revolution to get at nuance and variation. What’s the point of an AI that gives a half-assed big picture when the whole point of digital computation is to do all the small calculations that inform the whole? IMO, AI is going to suck at thinking specifically until it reaches into small data rather than just LLMs.
There will be more on that comparison I hope - worth reading this too, https://www.annualreviews.org/content/journals/10.1146/annurev.soc.012809.102659, which has influenced my thinking on this.
So I read through the paper and found references to "AI" to be very vague and speculative. Which is appropriate, I suppose, for a very vague and speculative technology. If AI is doing the sortition and defining the terms of the reasoned debate, then the process is at the mercy of the black box at its heart synthesizing its input data. As it always turns out, the issue is not AI but the how it's output is consumed by the humans. In a way it's totally appropriate and clever for Silicon Valley to pitch AI as a benevolent objective God. There's a deep rooted human need to kick all of our most vexing problems upstairs to a Great Tribal Chieftan in the Sky to solve. That's the basic definition of religion.
There's no question the purveyors/profiteers of AI want to do away with any democracy that doesn't put them in complete control. The "Dark Enlightenment" is upon us.
Doesn't this apply to democracy in general? At best, there is a very limited coupling between voters and their elected representatives. At worst, votes are mere proxies for campaign finances and political capital. If democracy pins its legitimacy on representatives "acting for" voters, it seems that legitimacy is weakening day by day.
Basically the current state of "AI democracy" is little different than being invited to the studio audience of Dr. Phil Primetime.
There is a large literature on this and related issues in Science and Technology Studies e.g. the work of Sheila Jasanoff and Brian Wynne. Weirdly narrow democratic debate related to science and technology is not new.
See for example: Welsh, Ian, and Brian Wynne. 2013. “Science, Scientism and Imaginaries of Publics in the UK: Passive Objects, Incipient Threats.” Science as Culture 22 (4): 540–66. https://doi.org/10.1080/14636778.2013.764072.
I don't think this limited-view-of-democracy problem is limited to AI field, though I think your point around engineers' skills and personalities making it bigger there is right on.
I have a close friend who does community engagement consulting (e.g. for municipalities or large-scale NGO's), and that world has a similar blind spot. The entities contracting for community engagement simply don't question WHO makes decisions, and only focus on what information the decision-makers (often themselves) have. I think of it as a general cultural tendency to assume that bad decisions are the result of poor information or technical expertise rather than the result of (for example) simply caring more about what real-estate developers want than what poor & working class residents want.
Thanks for writing and sharing this. A note: it would help some of us readers if you explained abbreviations the first time through (LLM/large language model) and identify the “it” when explaining the asterisk at the end (Hanna Pitkin’s The Concept of Representation). Again thank you for sharing. I sincerely appreciate it.
I see no use for AI in my garden or in any endeavor. Send AI to Mars. Let it sort through whatever data its idealistic makers provide, keep itself busy factoring out the weird and significant deviations that some of us cherish, and languish on that planet. Get AI out of my energy grid. Colossal waste of space and resources. Let's feed people.
You are correct to say that the AI democracy debate is weirdly narrow, but I think it’s even narrower than what you imagine.
I’m part of a group that believes we can actually measure public opinion, and we also believe we have worked out the problem of implementation. it’s not direct democracy, but it’s a big step in that direction.
In regard to changing the political landscape, unfortunately relying on our present day political systems does not work. The track record is bad. We can’t even get a small incremental change like rank choice voting.
The problems we have in the world today cannot wait for the slow pace of today’s political systems, even if they make incremental changes, and big changes like sortition are extremely unlikely.
AI is obviously going to play a big role in our future, but there is no way to implement it without bias. We are solving this problem by isolating the data in its own publicly owned institution.
All analysis of the data is controlled by the users. Users will use all methods made available to them by the free market.
We have a website, but it’s still a little clunky right now, it’s a work in progress, so I would suggest that if you want to know more about what we’re working on, check us out on Reddit. simply look for KAOSNOW.
This brings back ancient memories of reading and arguing against Habermas on democratic deliberation. There were two missing pieces, IMO, both of which are foregrounded in this post: power and party dynamics. Democracy is shaped by power concentrations, especially financial interests. You can't promote a more ideal democracy without promoting a more democratic -- and pluralist -- distribution of wealth. Meanwhile, political parties, organized around some combination of interest and ideology, are inescapable and, in some form, necessary. How they articulate with democratic procedures is huge. Indeed, the core political flaw in legacy socialism is its inability to accommodate multiple, contentious parties.
I don't see how folding in AI changes any of this.
Reminds me of the way some medical professionals generalize from studies without much attention to the distracting complexities and contradictions of the individual patient.