Can AI Strengthen Democracy? | Sandy Pentland

Welcome to Behavioral Grooves, the podcast that explores our human condition. I am Curt Nelson.

And I'm Tim Houlihan. We talk with researchers and other interesting people to unlock the mysteries of our behavior by using a behavioral science lens.

Yes, I love how you say mysteries. There you go. All right, so Tim, Picture this. We're going back 10,000, 15,000 years ago. There's a group of us. We're sitting around a campfire, right? We're telling stories about what happened during the day, what the hunt was, how we worked together, different things, what we encountered.

We're not debating spreadsheets and setting

up times for

meetings.

We're not going over the numbers of the spreadsheets and talking that. No, we're visualizing these things. discussing like what happened and like telling the story and like how you froze when we met up with the big, you know, saber-toothed tiger. And, you know, I had to come with the spear to save you. So next time, maybe not put you in front, maybe put you in the rear, right?

And I'm better in a supportive role.

Yeah, yeah, more, more, more like, you know, bringing us water and doing all this, the good supportive stuff. But this is how we caught— this is, this is what we did. We didn't share the numbers and do this. We told stories and we, we brought in those experiences through those stories. And then those stories evolved and they shifted, but they, they helped us come together as a group. They helped us understand the world that we lived in. So the people who weren't on the hunt actually could start to visualize what the hunt was like, or people who were on the hunt could visualize what was going on in the in the little, you know, community as they were doing the other things.

Over time, these shared stories actually evolve into culture and like social norms and things like that. And our guest today, we met Sandy a couple years ago at MIT Conference on Polarization, but today we're talking to him about his newest book, Shared Wisdom. Now, Sandy Pentland, the author, says that The process that Curt just described really hasn't changed all that much, at least not in our genes, right? I mean, maybe we're talking about spreadsheets and setting up meeting times today, but we still rely on shared wisdom to evolve, to improve, to innovate. And we do it through story sharing and collective sensemaking.

Yeah. And I think that's a really important part is in today's world, we We tend to focus in on some of those other things, and we don't really think about that shared storytelling, that sensemaking of the collective. And so what has changed is the scale. So instead of 20 people sitting around a fire and talking, we now have billions and billions and billions, if I could do my—

Billions.

Carl Sagan, right? Here we go. Connected by digital platforms, and increasingly, uh, shaped and guided by AI, artificial intelligence. And Sandy has really been at the center of AI research, um, for decades. I mean, literally decades. He's one of the first people to be, be studying this. So from early AI waves at MIT's famous Media Lab, which he led— I mean, Sandy is one of the premier guys in this field, right?

Yeah.

All the way to today's large language models. And he talks about that in this book, Collective Wisdom. And he asks a deceptively simple question, I think. Will AI help our shared wisdom evolve, that shared connection, that storytelling, or will it lock us into kind of the past and into elements that aren't necessarily about shared, they're just data. And that, I think, is really interesting.

It is. So Kurt and I also talked to Sandy about the Enlightenment as a network effect and how AI systems trained on the past data can end up creating this cultural stasis and why distributed leadership, whether in the military or in democracy, Might actually be our recipe for the future.

Yeah. And we explore one of the most positively hopeful ideas, um, we hope, right? And that is that AI could actually improve democratic deliberation if it's designed correctly. And so this is not a doom and gloom conversation. It's actually pretty cautiously optimistic. How do you like that?

I, I, I like that because I was afraid it would be too dark. Uh, it's also It's also a reminder that the real question isn't, is AI good or bad?

Yeah, it's, it's what kind of culture do we want it to help us build? Where do we want it to take us?

Exactly. So we encourage you to sit back and relax with a healthy draft of shared wisdom and enjoy our conversation with Sandy Pentland. Sandy Pentland, welcome to Behavioral Grooves.

Happy to be here. Looking forward to it.

Actually, welcome back. We did meet at MIT a couple years ago, back at a polarization conference. So technically, welcome back.

Yeah, well, good to talk again. Yeah, absolutely.

Okay, we're going to loosen things up with our standard speed round questions. And first, we'd like to know, would you prefer to learn a new instrument or a new language?

Probably a language, because I always feel handicapped by not knowing more languages.

Yeah. Any come to mind?

Oh, well, so a little story. When I was young, I learned both Spanish and French in addition to English, and I confuse them all the time. The only place in the world I'm comfortable is Barcelona, where they speak a mixture of French and Spanish, which somehow that works for me.

Right. Oh, that is— I love that. I love that.

Just an anecdote. I was giving a talk in Bogotá and I decided that I would open it with some Spanish. And my Spanish is very bad, so I rehearsed and memorized. And after I was done, one of the guys raised their hand and said, you're the first American that we've ever heard speak Spanish with a French accent. I don't think that was a compliment.

Yeah. Okay. Okay. Yeah.

All right, Sandy, second of our speed round questions here. Are you a coffee drinker or a tea drinker?

Coffee.

Coffee. Coffee. There you go.

Gotta have that caffeine, right? And I have this argument, not argument, discussion with my wife all the time about how much caffeine is good for you. I'm holding out for 4 or 5 cups a day. Wow. No, no.

I thought there was no limit to how, you know, caffeine just keeps getting better and better.

There's science about this, up to like 4 4 or 5 cups, it actually extends your life and fights off dementia. It's like, okay, good. Let's have another cup.

Well, I need to drink more then. So there you go.

Send us those papers. We'll have to use those. Okay. 3rd speed round question. True or false? If we're not careful, AI could actually make us less intelligent as a society.

It could definitely freeze society. make it so that it doesn't evolve, right? Because it's all backward looking. It's all the things we said in the past. It's not the sort of cool new thing, right? And that seems like a recipe just for cultural stasis.

Yeah. Oh, we'll get into that because I think that's a really interesting—

It is, yeah.

Perspective on how we think about this because that's not necessarily the general consensus as we think about this. But 4th and last of our final speed round questions here. Are shared stories the genes in our DNA since they are the building blocks of human culture? Would you consider that an okay statement?

Yeah, I think that's fair. And we'll talk more about it. But with that analogy, that's actually how humans think and how culture evolves. It's this sort of evolutionary analysis with stories as little atoms that you fit together to make shared wisdom and get people able to act together.

Oh, I'm glad you agreed to that because that was taken right from the book.

Yeah, I thought I recognized

that. Yeah. And just after we had told you we're not gotcha journalists here, I was going to pull a little gotcha on you there if you'd gone the other way. So

it's kind of a poor setup there.

I'm sorry. I apologize in advance.

We are talking to Sandy Pentland about his new book, Shared Wisdom: Cultural Evolution in the Age of AI. Sandy, you've been working with smart computers and artificial intelligence for a lot of years, and so we were particularly excited to read your book, and we're glad that you wrote it. You wrote in the foreword of the book, digital technology can either hurt or help. Now, I don't want to be too binary here because I know that that's not what the book is about, but Is it hurting us or helping us?

Well, both is the answer, right? I mean, it's certainly amazing that it potentiates all sorts of things that we wouldn't have done. I could give you stories. I mean, my first job was actually when, in response to a famine that had occurred in India, and millions of people died. But nobody knew there was a famine because nobody knew the crops were failing. And as a result, people put up a satellite to be able to look at crops so you could predict when there was going to be a problem. That seems sort of second nature to us now, but it saves tens of millions of lives. It's also why we know that the Amazon forest is becoming smaller. It's why we know glaciers are changing. It's like Knowing more about your world can be really beneficial. At the same time, we see polarization and influencers doing crazy stuff and people spying on you and stuff like that. So yeah, there's good and there's bad, but look, it's here. Yeah. It is not going to go away because the additional productivity that it does is What allows us to feed all the people, allows us to have longer life expectancy, et cetera, et cetera, et cetera. We're not going to give that up. What we've got to do is say, okay, we've got this digital stuff. What do we want to do with it? What sort of culture, what sort of life do we want to have that incorporates that? And then sort of work backwards. You need that North Star of what sort of life, what sort of culture do you want to have? And once you've sort of picked something like that, then you work backwards to make sure you don't run off the rails.

Yeah. And I, I love that idea of like that North Star kind of leading you and kind of helping you give direction and different things. Um, I wanna go back to the title of the book, Shared Wisdom. So can you help our listeners understand exactly what you mean by shared wisdom and how does that parlay into this concept of storytelling, but also in the digital age, what that means?

Well, we live in an age where we've been, it's almost like brainwashed, but the standard thing is we're rational, we're logical, we're thinking things through, we're individuals, it's all fact-based. But the truth is that's not the way we're built. A good way to sort of think about this is our history. So there used to be hunter-gatherers around campfires, Studies of hunter-gatherers today show that what they do when they get together around that campfire is they talk about the events of the day, what worked and what didn't work, and they use that to decide what to do the next day. And that's sort of a template for how people learn. And if you look at that process, so they're bringing stories together because they're not always perfectly correct about this, of course, like humans are. And then they come up with sort of a shared idea, shared wisdom about where they can go that's safest and the food will be best and so on and so forth. And that's the basic process that you see over and over again. The most striking example of it in our sort of stories we tell each other is the Enlightenment in the sort of 1500s, 1600s. And people say, well, gosh, there was this blooming of ideas, but you never hear people say, why did that happen then? And they say, well, maybe they invented science. Well, no, science has been around for thousands of years. And actually real science with statistics and stuff like that had to wait a few hundred more years. So it wasn't science. What it was, was there were postal systems that were controlled only by the royalty, and they opened them up. And so people who were concerned about understanding the world, interested in things, started sending letters about science to each other. So Leibniz, famous scientist parallel to Newton, wrote 3 letters a day for his entire life. The letters are not like 2 paragraphs, they're 5 and 10 pages. Yeah. This guy was a maniac, right? And he was very far from alone. And these sorts of trading of letters about how the world works is why they're called men of letters, because they write letters to each other. And that's where the Royal Society in London came from and the Académie Française came from, as the kings and other royalty realized that actually having people that understand how things work is a pretty good thing for the country. Right. They could understand how to make coins and cannons and this and that, and so they gave rewards to people who came up with really good theories. They became fellows and they had a stipend and so on. And that's in some ways the beginning of the modern research enterprise. And it was those incentives for creating new useful ideas and the trading of ideas between people, just like the people around the campfire trading ideas, trying to find things that people can all get on board with that created this blossoming. So it was a shared wisdom. Okay? Now I say wisdom and not truth because the wisdom of 1600 is not the wisdom of today.

Fair enough.

Fair enough. Even in something like physics, back then, they didn't know about different types of atoms. They didn't know about thermodynamics or electricity. And if you think that physics is in sort of better shape today, you should remember that 25 years ago, physics thought they were in really pretty good shape, the Standard Model. And you get the people on TV talking about the beginning of the universe and they're so certain. But about that time, they discovered that everything they knew was only about 20% of everything. And there was this mysterious stuff that was the vast majority. It's called dark energy, dark matter. Dark matter. Yeah. The vast majority of the universe, they haven't got a clue, right? Still, they don't have a clue. So don't call physics truth. Call it, it's a story, right? And it evolves as we get facts and people sort of say, hey, that's useful. And what really it's doing here is it's saying, look, we need to combine the views and experiences of different people. And that's sort of like a collective intelligence or a community intelligence. And then We're gonna sort of debate about what works. We're gonna do little experiments. We're gonna find out what's really helpful. That's the sort of shared wisdom, as it were, and it will evolve over time. And if you compare that to the, we're rational individuals, it's a lot more modest. It says, no, I'm not all that bright. You're not all that bright, but together maybe we can figure it out. And it's our best guess. It's not the truth with a capital T. It's our best guess. And that's the sort of wisdom that if we all to say, yeah, that sounds like the right idea, then we can all act on that together. If we don't agree on that, then we can't act and we're going to get nowhere. Yeah. So that's really the blood pumping through the minds of society. And that evolution of shared wisdom, the incorporation of new stories, that's how culture evolves. And of course, different people, different communities have different experiences, and so they end up having different cultures. And that's not a bad thing. That's evolution, right? The ideas that are good ideas will turn out to be more useful, and the ones that aren't may get dropped, and people will copy other ideas that they see working. And, and it's this sort of evolution of ideas, evolution of shared wisdom that keeps culture Adapting to the current situation and moves us forward. It's not logical deduction.

Hey Groovers, we want to take a moment away from our conversation to thank you for listening to Behavioral Grooves. If you enjoy the conversations we're having and want to help us keep the groove going, here are a few simple ways that you can support the show.

First off, subscribing to our Substack is a great way to stay connected with us between episodes. The weekly newsletter provides you with cool insights that are beyond the episodes, and they get delivered straight to your inbox.

And if you haven't already, leaving a review or a rating of the podcast on a platform like Apple or Spotify or YouTube helps other curious minds discover us. And there's 2 great things about that. One, it gives us a boost, and 2, it costs Nothing.

And it only takes a second, but it makes a huge difference for us. Plus, we love hearing from you, so don't be shy. Leave us a review or give us a quick thumbs up.

We're coming up on 500 episodes, and we're doing this because we love the conversations we have with our guests.

Yeah, we also want to do it because we love bringing you insightful, behavior-changing content every week, and we hope that some of those insights will help you find your groove.

Oh, Sandy, I find that really interesting, and that, that concept of how culture is that, uh, the response or the stories that get told around that community. And I can see how that then— all right, so somebody living in, uh, uh, you know, sub-Saharan Africa has different stories than somebody, uh, in, you know, North America versus somebody in, you know, somewhere else in the world. And so those are there. Given today and the interconnectedness of the world, how does that play out where, you know, the stories that I'm hearing and that Tim is hearing and that you're hearing are coming from all sorts of different places within the world? Is that creating a larger culture or is it just making it more difficult because those stories might pertain to me, might not help me? Does that make sense?

Yeah, it makes sense. I mean, For instance, I think one of the dynamics that I see is that Far Eastern cultures are much more communal than Western cultures. And I don't know which one is, what level of communality is correct, right? Yeah. Do we all have to have much more agreement in order to act, or could we act much more independently? I think it depends on the situation. And it may be something that the theory is, is that in the Far East, A lot of the food required communal effort. So you had to actually work together. The same thing if you go to a place like Finland, you see people really cooperating with each other. Why? Because it used to be that if you didn't do that, you froze to death. And those people are all gone. Sorry. Whereas in places like America, we're a bunch of escapees. We didn't like all this sort of government type stuff, so we decided we'd move into the forest. And yeah, so that's sort of our shared wisdom is a lot more individual. And it remains to be seen exactly how much of that you need in the modern world, particularly as you get all this change. Yeah. If you have a communal thing, it's a little bit slower to adapt often, The individual thing, you can experiment a lot, but it's hard to get consensus. Okay. Well, the starting gun has happened. Now we're going to find out, right?

So to what degree, I guess I'm thinking about the efficacy of stories is going to have to do with sort of their compelling narrative. It's the way that the story is told and then The content itself, does it make sense and can we agree on it?

Well, let me add that because we have misengineered our social media to have these influencers, the very loud voices who have compelling stories about how these bad guys are going to steal all whatever, right? But it's not the compellingness that really matters. It's does it work in some way? Does it really work for people? It's like, If a story doesn't lead to better outcomes, eventually people will clue into the fact that it's not working and they should do something different. I mean, a little controversial, but for instance, for many, many years, getting a college degree was absolutely something you had to do. But it's not clear exactly what that gives you as much anymore. It's not just a college degree. Maybe it's certain types of college degrees. What about community colleges that give degrees? What about skilled trades? Do those count? Those actually look pretty good. Make good money. AI is not threatening you. That doesn't look bad, right? Yeah. So that's something where there was a shared wisdom that this was the right thing to do, but it wasn't good for everybody in every situation. And now we're getting more stories And beginning to sort of rethink some of that as a society.

So, so with that, with society and technology and everything that is going on, it seems like it's getting faster and faster that the stories that we hear, uh, it's not just a story a day or even 2 or 3 stories a day, it's multitudes of stories a day. And can we Keep up, or can, can the story, you know, can we keep up with that? Is there, you know, the sense that we're feeling like we're bombarded with more and more stories and they come at us so fast that we can barely process one before moving on to the next? So does that— how does that happen?

So, so yeah, there's this sense that the pace of things is getting faster and faster, and it's very disturbing for most people. And it's indeed getting faster. And that's because we can talk to each other better and search for things better and all sorts of things like that, as well as there's just a lot more people, right?

Yeah.

So far, what I see is people with the right sort of attitude, which is a very entrepreneurial type of attitude. It's like, I'm skating over the ice so fast that the cracks are behind me, that sort of person. Tends to be young, tends to be in some sense full of themselves in a way because they're going to go out there and do it. Those people are just having fun, right? And so a question is, can we all become that way? And you might say, oh no, I'm too tired, I can't do that. But I think actually that's a potential for everybody. One of the things that holds you back is that our institutions are all built around a sort of late 1700s, early 1800s model of how people think, which is, Generally, people are supposed to be a gear in a big machine, right? The model is based on armies. You have the general, general says, go do this, and everybody has their little job. And it may not make sense to you, but you have your little job and you're supposed to do it. That's our model. That's what our schools are based on, right? We're going to put everybody into little sort of narrow things. They're going to learn this, this, and this so they can sit at that desk and check the boxes, right? That's the idea. But that's a terrible model for a dynamic world. If the world is changing, it's like exactly what you don't want. Maybe not exactly, but close to exactly what you don't want.

Yeah.

What you want is you want something where it's people able to sort of assess things very broadly, take actions, work together, have a really very explicit idea of that North Star, where it is we're going. And what's interesting is you see, I'll tell you a story about the US Army in Iraq. So the US Army in Iraq— Yeah. Army's like the classic sort of bureaucracy, the classic business, right? Levels of hierarchy, you do what the guy says, et cetera. They were getting their butt kicked. They really were, because all of the people they were opposed to were acting locally and could just turn on a dime. And the way that all these hierarchical organizations are done is everything has to funnel up to get a decision made. Okay, so it's very slow. Signals don't get all the way up there, right? And so what the leader did— McChrystal is his name— is he said, okay, We're gonna do this differently. Every time we have a problem someplace, every locality, we're gonna get the people who are directly involved. They're gonna come up with a solution. They're not gonna tell their boss. They're just gonna do it, okay? And if it works out well, they get a pat on the back. And if it doesn't work out well, they get hell. Excuse the language. And guess what? That sort of worked. And one of the reasons it worked was that there's another thing other than orders It's commander's intent. So they said, look, this is the North Star. We'd like this area to be quiet and passive. Nobody gets killed. You take over, figure it out, and do it in a way that's appropriate to this particular little area. And you're the guy that runs intelligence, and the guy that does supplies, and the guy who's the medical guy, all get together and decide how to do it. That's a really distributed management. Where people on the ground are in charge of making stuff happen. Now you can imagine that happening, for instance, in social services in the US where the people who are working in that office in the poor place have the ability to make decisions, to adapt to people, to cause things to happen, to spend money, et cetera. I think you'd end up with a lot better situation than what we have today where you have these uniform rules that are supposed to apply to everybody. And of course you have to audit things. You have to make sure nobody's taking stuff out of the till, but we know how to do that. And that's what all these digital things are really, really good at. So giving more power to the people on the spot empowers communities, which means you can be much more adaptable. Things can change. If you see communities that are being very successful at something, other communities should be able to copy them almost directly. And over time, what you do is you're doing the version of sort of the social services version of science. You're trying different experiments, you're sharing results, you're looking at what actually looked, and eventually you get quite a bit better. And the interesting thing is that way of learning, is much more powerful than what we're taught, this business of logical deduction and reduction and so forth, because it explores a much broader range of alternatives and it's evidence-based. It's something that where there are a lot of things we don't really know how they'll work and you need to try it. Okay, so you need people who are willing to try and live that way. Yeah. And then you see what happens, right? And they won't all be well. They won't all turn out well, but you will discover things that work really much better than today's systems. So I hope that made sense. I sort of went on

a— It does. Thank you. I appreciate that. Yeah.

I love your optimism in all this, by the way. The book is absolutely optimistic. We'll get back to that in a bit. But I'm just thinking about to what degree are these— is our successful testing and trialing and working through things predicated on an accepted North Star or an agreed-upon North Star.

Yeah, you need to have a fair amount of agreement about that. Otherwise, there's fun. I mean, it's like if the notions of virtue and morality and the good life are different, then those communities have to be operating separately. Generally, they have to be interoperable. That's not as bad as it sounds. Everybody goes, oh, that's why we have uniform rules. Well, in the United States, for instance, the reason that ownership rights exist across states is not the federal government. It's that the states figured out that if they weren't interoperable, it would hurt their economy, basically.

Yeah.

So about About 10 to 15% of all the laws in the United States come from this not-for-profit collaborative group of volunteers who get together from all the different states called the Uniform Law Commission. They say, well, gosh, it doesn't make any sense that if you adopt a kid in this state that you no longer have the kid in this other state. We have to do something about that.

Yeah.

Then they bring it back to the states and the states say, yeah, that's sort of screwy. Let's change the law just a little bit, right? And that, since 1870, has been the thing that kept us going through the robber barons, through the advent of telephone, through the stock market, stock market crash, digital things, World War I, World War II. I mean, it's very adaptable. It's really very successful. We should do more of that where the communities don't have to be the same, but they have to be interoperable. Yeah. And they have to have— so they have to have North Stars that are similar but not the same, and they have to be smart enough to realize that being interoperable and collaborative is better than fighting. So that's typically the core problem there is having good data about what's actually happening. So if you have just sort of— Yeah. Rumors and big voices and so forth, that's not going to cut it because they're going to talk about, ah, the sky is falling. And so the work I did at the UN and other sort of places is all about, let's find out the truth of this. Let's actually measure how many people are poor. Let's measure how good life expectancy is. Let's measure the food supply, things like that. Because otherwise, what are you talking about? You're talking about fantasies.

Yeah. And you can have competing stories that are out there, and which to— if you're just the average person, how do you discern which of those competing stories is the truthful one?

It's difficult in that it's a skill, it's a process. This is what science is, right? Science has these hypotheses. They're different stories about what's happening. People debate about it sometimes with a lot of passion. They try to get more evidence and share it around and be able to make the evidence really convincing. That's not a bad process. That actually is sort of fact-based education of people and trying to get people to have a shared wisdom that actually aligns with what's happening.

Yeah. So you relate a very interesting history of AI. Let's get into the AI part of the book here, talking about AI booms and AI winters and how we've cycled through this for 70-plus years now or more. Yeah. So what are some of the lessons that we can learn from those different kind of iterations and those cycles? And where are we currently in that? What do we know?

Well, just to sort of recap it just very briefly, the term AI was invented in the mid-'50s. A guy whose Gave me his endowed chair at MIT is the one. Marvin Minsky is one of the people who helped name it. The original things, there's many different sort of AIs at the beginning. The one that was most successful won a Nobel Prize in Economics, and it's called optimal allocation. And it's a method of doing, essentially solving linear systems with constraints. You don't need to understand that. You need to know that Soviet Union decided to run their entire economy on that. And the rest is history. And the reason that didn't work is that things change. And the numbers are bad because people want to fudge a little bit to look better, and they didn't really understand the laws of things. There really aren't laws. There's things that change and are different. But that way of solving things actually is the most common type of computation in the whole world today.

Okay.

Every spreadsheet, Every delivery vehicle, every, you know, how do I drive someplace, uses exactly that method today. The chips and the phones are designed for doing that sort of thing. So it really was hugely successful, but only for small problems where it's pretty well-defined. So people began to learn about that, and they came up with another generation in the '60s, which was early '70s, which was expert systems. They said, well, we don't have laws about this stuff, but we have some guys who look like They're really experts, so we're just going to ask them what to do. And they would have these sort of rules of thumb that they would put into the computer, and it was really pretty good, but it had another sort of fault, which is some company like a bank would say, well, let's make our loans by these expert systems. There's heuristics for giving loans to consumers. And they'd say, well, we're just going to make it uniform across the entire country, right? But every community is a little bit different. And as a consequence, the local loan officer disappeared, the local knowledge disappeared about money and things like that. It all became very abstract. And for communities that weren't typical communities, it was pretty bad. And so then you got machine learning a generation later where we were actually using the data from the internet to be able to do things. Which is good, except that it's evidence-based. It actually is adaptive to things, except the companies are only incentivized to make money. They're not incentivized to have good service, right? They're not incentivized for privacy. It's like, yeah, okay, good, but you didn't hit it. And now we have current AI, which uses everything on the internet to make those same sorts of decisions. And it could be pretty good. But of course, it's trained on everything. So it's trained on the crazy people. It's trained on, you know, all sorts of stuff that's just, you know, garbage. And it's not the same for everybody. All the data in these experts, in these AIs today, interestingly, like about 80, 90% of it comes from Americans that live on the East Coast or West Coast. So you say, well, but it speaks Turkish. Yeah, people who speak Turkish in New York, not people who speak Turkish in Turkey, because those are the only people that post things in such overwhelming numbers. So it's really sort of a crazy thing, and it's not representative of most places where— so you hear this stuff called sovereign, phrase sovereign AI, which is people want to have AI tools that represent their culture and the ways that they do things, which are not—

Right.

And I think that's actually probably all to the good, because now we're going to have different ways of doing things and we can compare what works to what doesn't work. We're not so good at this prediction type stuff, right? So that's a lot of what's wrong. Again, a little bit of a diatribe there, but it's useful to know that this is not our first go-around and we screwed up the first few, but we screwed up the first few in fairly standard ways that we could be more aware of today. Yeah.

Yeah. And you think about, you talk about more individualized AI getting not Not even from a community level, but on an individual level as consumer champions.

Yeah.

Right? I mean, this is a fantastically optimistic view of how AI could be an advocate. Is that a fair term? Is that pushing too far?

Yeah. I mean, we did a big sort of survey, AI survey actually, in Washington, DC, not the politicians, the people with Single moms with 2 jobs and 3 kids, those sorts of people. And the thing they wanted, I mean, the question was, what sort of AI do you want in government? And the thing they came back is, we want something that'll defend us from government. We want something that acts for us to be able to fill out those forms and know what's available and interested with, because government's just too complicated. Yeah. And we have a project called Loyal Agents, which is just loyalagents.org. People can look it up. It's all— it's with Consumer Reports. It's like, you know, so when you go on the web, you have all these people selling you stuff. And with agents and AI, they're going to be better at it. Oh my God, what are we going to do? Well, what we need is we need something that acts for us. Now, in legal terms, that's called a duty of loyalty. Right. It's like what doctors have to do. Doctors have to act for you, not for them. They may not always do it, but that's the law. And we want our agents, our AI, to act for us, not for whatever other sort of company, et cetera. And so what we're doing is designing the legal framework to make sure that that happens. And an interesting thing is What that means is you don't give all your data to the companies anymore. You have perhaps a group of people that share data among themselves, a cooperative, say, like a credit union, except it's a data union. And you maintain an AI that makes only the requests and allows the companies only to see the things that they need to see to have a transaction. And that can actually work. You see places around the world where people are actually beginning to do this. For instance, at the last G20 meeting, this was like a really interesting thing to people, particularly people in countries like India where there are cooperatives for everything. I mean, it really is not like this country at all. It's cooperatives all the way down. Everything's a cooperative.

Yeah.

And so what they do is they come together to be able to get advantage for their members. And, this is an example of something. You can actually now build an AI agent for you for a couple thousand bucks. And, that's today, and the price is dropping. And so, you can begin doing that sort of thing to be able to buy groceries, buy cars, do your taxes, whatever, those sorts of things. So, they're very enthusiastic about that. I think we should be enthusiastic We're going to see versions of it, but we want to make sure that these things represent us and not somebody who's just trying to scam us.

Curt, do you mind if I keep going? Go for it. Okay. I wanted to thank you for that, Sandy. That was really great and thoroughly enjoying this discussion about the impacts of AI. What about on democracy? What's your thoughts about the impact of AI on the state of democracy?

Well, and you have a chapter like Reinventing Democracy, chapter 6.

And there's a website that has software and examples and everything that's all open source. It's called deliberation.io. Okay? So those are the 2 things, deliberation.io, and the other one was loyalagents.com. That's the Consumer Reports one. The deliberation one, we decided all the things that we do with democracy are under threat from some of these things, from the big voices, the things that subtly manipulate you and so forth. Actually, the state of the democracy is not, in my view and in many people's view, not that good. The reason is that people just don't have voice. Right? You have, uh, like in the US, you have a House of Representatives member who represents millions of people. Nobody actually knows the guy, right? No, he comes from a very different culture, or she comes from a very different culture than you typically. I mean, it's like, this is not really representative the way it was back in 1780 when, you know, there was really only one town and it had 20,000 people and everybody sort of knew each other. Or at least knew a friend of a friend. That was the original design. And so what we need is we need ways of connecting people, and AI gives some ways to do that. So let me give you just a little story, the story that inspired us. In Taiwan, about 12, 13 years ago, there was a survey about how many people trusted government, and they were worse than us. There was about 7% of people trusted government. We're still like in the 10, 12 range, right? So much better. So much better. Way better. So much better. And what they did is they set up community discussion groups with digital tools, a thing called Polis, that allowed you to make comments. No influencers, no loud voices, no sort of rude comments were allowed. But you could see what everybody was saying. That was the big thing, is being able to see what the whole community said, not just the community, the whole country was saying. In a period of about 10 years, they went from 7% trust in government to 70% in government. We haven't had that for most of a century, right? Just imagine what we could do if we actually sort of were all on the same page. And so forth. And so we designed a tool that would help the average person, the person, like I say, single mom, 2 jobs, 3 kids. That person has just as right to participate as somebody who comes from a wealthy family and has the time to go to the meetings and shake hands and stuff like that. And so that's what it is. And we have, I think, a lot of evidence that that type of thing can really work here. One of the little cute things is we showed that having a mediator AI, something that does not contribute content, it just sort of reminds people about where the discussion is at and what we're doing, just like keep people on the same page a little bit, right? I hear people saying this and that, that sort of thing. That doubled the subjective quality of meetings. The meetings were twice as good because this little AI just kept people on track. No content, just, you know, I hear this or that. What about, you know, that sort of, you know? And so I think that's really hopeful. We could do stuff like that, like Taiwan did. We could do it in all the meetings we go to in organizations potentially. I think it's a pretty interesting thing. And so now, We're exploring how to deploy this more broadly. People can take it, do what you want with it, with no sense of ownership here. And I think that using that sort of thing, there's real hope for the future. Taiwan is sort of the shining example. Now, we're not going to be just like Taiwan, but it shows that there's hope. You can do things, and the experiments we've done have shown that too. One of the major things, let me just add one more thing, is I said in Taiwan they had this digital platform that showed what everybody thought. So it's like a graphical thing that shows, oh, here's the people think this about this question and that and so forth like that. So you get a sense. We did a big experiment here at Stanford. Rob Willer was the guy that ran it. It took everything we could find that was a suggestion for depolarizing American society and tested them on 32,000 people. And the ones that allowed you to understand what the other side actually thinks without the influencers did by far the best. It changed people's opinions by about 30%, just like a 5-minute exposure. Wow. There's all these topics that we're so divided by. Take gun control. Turns out almost everybody in America believes pretty much the same thing except for some people on the extreme ends. Same thing with abortion. The US in general is a little bit more liberal than Europe, but just a little bit, except for the people on the extreme ends. Right? And topic after topic is like that. Now, there are some things that are genuinely different, but you have to know that to have a constructive conversation about it. And again, the influencers, the extreme people, the voices like that are the things that continually derail our conversations.

Well, and you talk about that too, and I know we don't have a lot of time left here, but the rich get richer feedback loop that you talk about within This idea of how data gets presented out there. I think one of the bigger fears that I have is, all right, we are consolidating. Much of the technology is being owned and operated by just a few entities. Given how they structure algorithms and everything else, is there hope? moving forward that they're not going to have their own personal kind of— and maybe it's not even intentional, it's just like we're protecting our stuff and doesn't necessarily help in building what you're just talking about here.

Yeah, there's obviously a lot of concern about this. It's a valid concern, of course. I mean, just look at social media and so forth. One of the hopeful things is that the price of AI, so a given amount of intelligence, is dropping by a factor of 2 every 3.5 months. So if you look at all the open-source AIs in the world, they drop by a factor of 10 each year, the cost. And in the last—

So way different than Moore's Law, right? I mean, this is dramatically faster.

So much faster than Moore's Law. I mean, Moore's Law has transformed the world. This is almost 12 times faster, something like that, right? So in the past 3 years, while there's been sort of all these big open source things, the price has dropped by a factor of 1,000. It used to be this really expensive thing that used all this electricity and blah, blah, blah. No, not so much anymore, and it's continuing to drop. And so the question that makes it possible for communities, for other sort of peoples to have their own AI. And I think what we need to do is we need to encourage having, you know, not this sort of central thing which sucks up all your data and sort of biases you.

Yeah.

But people need to be able to defend themselves. And that's what that loyalagents.org is. Interestingly, this is the way almost every country in the world thinks about this. They say, look, today we have either the American hyperscalers or the Chinese. We don't want either of those, but we want something that represents our culture, helps us in trade and education of our kids, and it's not going to be what the Americans say and it's not going to be what the Chinese say. Fortunately, that price has dropped by 1,000, and so they really have more and more plausible ability to be able to do this. In which case, what you get is you get this sort of smorgasbord of different sort of AI capabilities. You want to have rating agencies, things like that. You want to have things to certify them because there's going to be a lot of scams. Interestingly, places like Singapore have already set up stuff like that to be able to validate the claims of these things, and it has legal force behind it. And other people are beginning to sort of notice, yeah, this is actually not a bad idea, is to be able to audit what happens and be able to hold people responsible and to be able to test these things. Right. We know what they're doing. And it's a lot of the things that we didn't do with social media and haven't done before that, you know, sort of ended up causing all the problems. And so, I'm at least somewhat hopeful that people are waking up to the fact that, you know, data, AI things, you got to think of it a little bit like money. You got to know where it came from. You got to know what it's doing. You got to know that it's based on factual claims and things. We do that with money all the time. That's part of the infrastructure. We got to do that with things like AI and probably social media also, which is going to upset some people, but it will result in a lot more sensible behavior.

Wouldn't we love that? Sandy, we have thoroughly enjoyed this conversation. We wish we had another hour with you, but right now we just want to say thank you so much for being a guest on Behavioral Grooves.

My pleasure. It's always great to have these conversations with smart people who ask interesting questions and have opinions and so forth. So great. Thank you very much.

You must

know who you're talking about. Tim or me are intelligent. That's a first we've ever heard.

But it really is. It really is fun.

Welcome to our Grooving Session where Tim and I share ideas on what we learned from our discussion with Sandy, have a free-flowing conversation, and groove on whatever else comes into our AI-assisted brains.

You're getting off to a dark stop there.

I am a robot and I make all my decisions based on AI. Yes, I do.

2 things flashed in my head right there. One was the robot in the Robinson family Lost in Space show, right? Will Robinson's best buddy was the robot.

Danger, Will Robinson, danger.

Exactly. And then the second thing was Devo, the early punk rock band from the late '70s and early '80s.

Who wore flower pots on their heads. Yes.

They did. Yes. They did. Mark Mothersbaugh.

Devo. There you go.

Yeah. Okay.

All righty, guys. Oh, God. What is going on this morning? This is going to be in some bad outtakes, isn't it? There we go. All right. Let's talk about our conversation with Sandy. What did you take, Mr. Houlihan?

Something that in thinking about this, I was just really pleased to sort of be reminded that humans are just not all rational individuals.

What?

Yeah. From a behavioral science perspective, we get this every day, but that storytelling really emphasizes this idea that we are social learners. And I really like that idea that when we engage in deciding this is what I'm going to do, in part I'm deciding that because of what I've learned through the stories that I've been told. I've said this 1,000 times. How do I know the Earth is round? My mom told me.

Yeah.

My mom told me. It was a story that my mom told me. And stories stick. They really stick.

There's been this— again, how many years have we been doing this, Tim? This podcast thing that we've done? A year?

6 months?

Like 150 years, something like that? It feels like it sometimes. No, we've been doing this for 8-plus years. And in those 8 years, this theme comes up over and over and over again, that—

Right.

That we are narrative-seeking creatures, that we like the social connectivity of others. And that is enhanced and it's built through the stories that we tell, not just ourselves, but each other. And that's how culture is created. That's how wisdom gets shared. And as you said, That's how we learn. We learn, we retain stories much more than just facts and figures. And that keeps coming up over and over and over again.

Yeah. The campfire story is— the campfire metaphor, I should say, is so important because it's been around. This whole idea of sitting around telling stories has been a part of the human condition for a long time. And that's why it sticks so much. That's why it's so deeply ingrained. in who we are, that we are just not this logical fact-processing automatons. We are not robots. We love a story because it's an effective way to communicate and to persuade.

And the research points this out, right? I mean, there's cultural evolutionary theory, which talks about culture has evolved through storytelling, through this connectivity, through imitation. Through group consensus. And that was what Henrik Boyden Richardson, when they were working on that. Right. And so you have this factual-based components that's saying we have to tell stories in order to build our culture.

That's right.

And what I loved about Sandy's conversation, his book, and the conversation he had with us is this idea that we can reframe a lot of how we think about things as stories, and that then makes them more powerful in this. So theories in his mind are basically just stories. Yeah. They're the stories we tell ourselves about this theory of how the mind works, or this theory of how we learn, are just the stories about how we learn that we have come as a group to collectively share. And when we get some additional information, those stories shift, the stories adapt. Like over the campfire, like in the intro, like when I learned that, oh, Tim froze with the saber-toothed tiger, but then over time, you know, you actually get out there with the spear and you do some really great work. Well, then that theory has evolved, the story has evolved.

I'm totally a spear guy. Exactly. And that, that this evolution of the story, I think, is a really important I think that's an important thing. The way that we tend to frame a theory is that it's very rigid and locked in. But if we think about these things as stories, they are more adaptable, they're more flexible. We have the ability to change the story.

Yeah.

Whereas if it's a theory, if it's this deeply seated value system or this belief, we can get into rigidity that doesn't necessarily help us as a species evolve. Okay, I don't want to say any more about that. But he also helps emphasize this idea of just allowing the collective wisdom to win.

Yes.

Right? Like, so why not just think about what is it that we're doing that's working and do more of that? Right? And if it's not working, let's change it. But the collective wisdom, the things that people end up doing over and over again, can be really powerful and really good for us. And those are the stories that we should continue to amplify. Yeah.

And so that shared wisdom isn't about perfect truth. It isn't, right? We're not the perfect truth-seeking people. Well, we are. We're seeking. We want to have that perfect truth. But the collective wisdom isn't about that. It's about getting good enough. It's the satisficing component of this. It's enough Consensus around this to coordinate this. And again, this goes back to my kind of worry about this is because AI has gotten so good to— like, we aren't having that national consensus anymore. We're having consensus within smaller disparate groups that aren't separated Physically within the world. And yet we're having, we're looking at 2 different worlds because AI is so good and social media is such a, the algorithms we have to segment us. So we're not having this shared wisdom anymore. We're having pockets of what so-called shared wisdom is. And that can lead to issues, I think. Yeah.

In some ways, that's just a reflection of kind of the way the world is, the corporate world is. I think about some of the work we've done together. Oh, yeah. A lot of the work that you've done reminds me of Anita Woolley's work on cognitive and collective intelligence, this idea that we build this knowledge together. Think about the number of projects that we've done where we had 2 departments that were at each other's throat, right? And yes, like, they were just like, that department is so blah blah blah, and the other department's like saying, no, it's those guys there, blah blah blah. It's like, you know what, team building exercises actually make a difference in that case because it creates unified stories, right?

Right. You're, you're building a collective memory that brings the human side of people together. So no longer are you Tim from accounting and I'm Kurt from sales. You're Tim who has a kid who's in the circus and has, you know, a dog. And— Oh, I do have a dog. Yeah, and I have a dog. And I, you know, my kids went to circus camp as they were kids, and now we have this collective connection that, that, that's on a personal level, which allows us to understand that you're not just a roboton from, you know, accounting and you're evil and horrible person. You're actually a pretty decent guy. And oh, we're both trying to achieve positive outcomes here. Now we just have to take a look at this and figure out how we can do it together. And those I mean, and again, any one team-building initiative can kind of start to crack that. It doesn't solve it. But if you put a program in place and you're intentional about making sure you're creating the environment and the situation, and we've done that work with— and you see it. You can see it in body language. You can hear it in the way that people talk about it. It really makes a difference.

I love that. I really do love it because I love seeing people come out of those team-building exercises, Like they're new people, like with truly a fresh disposition. And I think that that's fantastic. You alluded to this earlier, but I wanted to come back to expand on this idea that AI has the opportunity, right? Actually, right now, AI is backwards looking, right? Yeah. That it's basically frozen in what the world has been and not what the world will be, right? And this could kind of create this problem for a status quo bias where If AI keeps creating, that it just keeps creating stuff that's already been created. And

with all of the inherent biases and horrible things and the segments that we already have. Yes.

Yeah. Yeah. So I think that Sandy, one of the things that inspired me about Sandy's conversation was, or our conversation with Sandy is this, we need a more adaptable AI. We actually need something that could help us get past, with expert learning systems and optimization models and machine learning, that gets us past just being stuck in the old stuff. That's the thing that I actually am hopeful that the technology could be created to do that.

Yeah. I mean, and Sandy talked about the distributed decision-making and all Yeah. The components that he talked about with, uh, Stanley McChrystal and some of the, the stuff that the, the Army is doing and different pieces.

Great story.

Yeah, yeah, I get that. And again, I'll go back to my, my— so you're, you're being optimistic and I'm being the downer here, um, because I'll go back to that component where it's, you know, AI is not forward-thinking. And if it was forward-thinking, is it forward-thinking? What is What's the underlying purpose of AI, right? And the people who are developing AI, is it for them to continue to use more of that AI so they can get revenue and various different things? And the way that they do that is by segregating people and, and having this AI be the AI for, you know, the flat earthers and this AI be the AI for the round earthers and this AI be The new AI for the triangle earthers and all of those pieces of it. And that worries me that it, you know, what is the underlying component? And it gets into, I think, some of the things Sandy talked about too, is, all right, if we're using AI as this intelligent piece, but we're not bringing new intelligence into the equation,

Yes.

And it is only this, this retroactive-looking piece, and we are expanding the knowledge of the human, you know, humanity as a whole.

Yeah, I'd like to be more optimistic. Honestly, I would like to just feel that maybe there is something that we, we can do better, that especially with the sort of distributed deliberative democracy research of, I think, Rob Willer's work on putting correct information in front of people, accurate representations of what's actually happening as opposed to the extremism kind of stuff actually is helpful. And AI could help us do that.

There was an interesting— I just saw it on the internet, so take this with a grain of salt. Oh, the internet. But there was research Where they paid people to watch different news stations than what they typically do. And then they say they paid them $15 an hour. They had them do it at least 7 hours a week, I believe. Don't quote me on this.

Wow.

And then they looked at change in perceptions and their perceptions changed. Like, how is— I forget what they measured exactly, But this idea of having a more accurate representation, a more plural representation of the world, I think is a really positive thing. And so there is hope. I mean, we understand the issues, and if we put policies in place and we understand that, you know, the people who are designing these systems may not have our best interest at heart, and so we might need to regulate them, and we might need to put some safeguards in place to do that. That's— there's, there's hope, right?

I think we can be hopeful. I think so.

And all right, we're gonna end on a hopeful thing.

Good, good. That makes— that'll make me feel better, I'll tell you that. And another place to feel good is if you do use the Facebook, uh, head out to the Facebook and check out our Behavioral Grooves community page because it's, uh, it's a great place to interact and to exchange ideas and to have a conversation as best as we can on social media.

And even if you don't use the Facebook, you can still go out and join our— use the Facebook just to join our group and only have it be part of our group because that's a fantastic part. It's a community of people out there telling stories. Like we're around a campfire, we're asking questions of people and getting their first perspective on how this question impacts them or how they respond to it. So it is not facts and figures, it's story-based components of that. Another place to get good stories is our Substack. We tell stories in those Substack components, right?

Every week. Every week. There's a new story. And you don't even have to have a story to go to Patreon and just—

You can just be a wonderful person who likes to sit around a campfire and give us some money because we're sharing campfire stories with you.

Because it's great that musicians would show up and sing American Pie for hours on end at the campfire while other people are cooking the marshmallows, but it's also a nice idea to put a quarter in their hat every now and then too. I'm just gonna say that.

American Pie for hours. I mean, I know it's a long song, but I'm not—

It is, but people like it at campfires. I'm just—

that's just why you have to replay it over and over and over again.

You have to play it several times, which I don't—

no, I don't need it now. Drove my Chevy to the levee, but the levee was— all right, nobody wants to listen to Kurt sing. All right, last thing. Um, you know, you're building a campfire. What do you need at campfires? You need friends. And so Share this or our other episodes with a friend so that we can make the behavioral groups campfire all that much bigger. So it's a bonfire, not just a little campfire. It's a huge bonfire that we're all out there and you can roast those marshmallows 15 feet away.

Yeah. So just be careful not to torch yourself. If culture evolves through shared stories, Then the real question isn't whether AI is good or bad. It's really about whether or not it enhances or corrupts the story-sharing process. Whoa. And we have the opportunity to navigate toward a positive story-sharing process.

Whoa. That was good. Can we—

Good enough?

No, that was more than good enough. It's whether it enhances or corrupts what? Our

story sharing process.

Yeah, that's good. You know, and I think we can build that. We can, we can bring that in. I mean, if we're intentional, we're at the point where it's not too late. We have hope, right? We can have this distributed system, shared intent, you know, not an overall, you know, Big Brother kind of thing. We can, we can make this happen. And so then our North Star question becomes, what type of culture do we want AI to help us build? Hmm. Hmm.

That's good. That's, that's better than just good enough.

That's because you wrote it. There you go. I just read it.

And we hope, Groovers, that you are able to take some of these ideas and some of these questions and some of the fantastic ideas that Sandy shared with us And use them this week as you go out and find your groove. Thank you.