Protocol Berg v2: Jeff Strnad - Delegation and Participation in Decentralized Governance

so much. So, uh, they always ask for questions. I'd love to have answers for some of the problems that I'm going to pose here. If you have an answer, well, let's do the answers first if they have an answer. Okay, I'm going to start my clock now.

Um, so, uh, today I'm going to talk about Sorry. Uh I'm going to talk about an epistemic view of delegation and participation in decentralized governance. Um there's a paper that I'm going to have the QR code at the end. There's a lot more material that I can do in 20 minutes. So I'm going to go fast.

I'm going to try to develop some understanding and then I'm going to have to summarize some of the results and not go through them. Okay. So um what's the idea of epistemic? The idea is that when there when there's a correct decision, we want the governance mechanism to maximize the probability of reaching it. and we're leaving out the possibility there's not a correct decision which is probably true for a lot of things and we're just focusing on the epistemic aspect.

So what's our goal? Our goal is to see if we can cooperate in a decentralized environment uh framework to to get good epistemic outcomes. Um now what we're going to do is we're going to make some really strong assumptions which are really unrealistic. The advantage is when we make these assumptions, we have clear results. And furthermore, if we have failure under these assumptions of some of our methods, that's a very powerful result because these assumptions make governance much easier.

On the other hand, if we have success, which we will, that's not that meaningful because the assumptions are so strong. We don't know where that what that success in this theoretical framework means. Okay. So, what are the assumptions? The first assumption is the epistemic assumption.

There's a single correct answer. We're going to look at binary choices, A or B. One of those is true. We're not sure which one. Um, the second assumption, voters share the same epistemic objective.

They all want to reach the correct collective decision. This this has some powerful advantages. We're ruling out malicious actors and actions, which is going to make governance assign a lot easier. Um, we're also ruling out voters just asserting their opinion rather than allowing some difference to other people's opinions. A a third assumption.

We assume that each voter knows the probability, their own probability that they can make the correct social decision on their own. And also any information they have, any items of information, they can take that piece of information and they know the probability that that piece of information alone will lead to the correct uh collective outcome. Okay. Um so now we're going to unfortunately I have to I have the slide navigation is a little funky here. Uh there's two cases.

One is we have independent competencies where voter judgments are statistically independent. That's an easy case. We get really nice baseline there. It's really unrealistic. Why?

Because most people have shared information sources, which is going to put us in the second case of dependent competencies. Um, in that second case, we're going to do another strong assumption. We're going to assume the total information out there can be decomposed into a set of statistically independent signals. the voters can know what signals they have their each voter's information set is a collection of those signals. Okay, now we're going to take a look at the independence case.

In the independence case, we have a really strong baseline, this optimal waiting theorem. Okay, so P subi is voter I's probability of of getting the correct social choice and they know that probability and we can calculate a weight wi, which is an optimal weight. If we weight the voters with those weights, we will maximize the probability of getting the correct collective choice. Okay, so that gives us a nice baseline. That's why we're looking at the independence case.

Okay. Um, so look at the second bullet point, second main bullet point here. A lot of times in Dows, we have token voting. One vote per token. If the proportions of the tokens line up with the optimal weights, then we're in then we're in the perfect situation.

You can just vote all the tokens. The weights will the voters will get the optimal weight. Now that also corresponds, if you look at the bottom line here, to the ratio of the absolute value of the weight to the number of tokens being the same for every voter. If that's true, then the uh proportion of tokens is just going to exactly match the weights and we're going to get the optimal result. Now I'm going to do some uh quick examples, some simple ones to devel.

So this first example is the independence case. Suppose we have 3,000 voters. they have a 52.5% chance of being right. They're independent of each other.

Um that's each that's going to create a weight of 0.1 and in aggregate they're going to have a weight of 300. Okay. If they make the decision on their own without the other voter, the other voter is a more expert voter, 73% chance of being right, a weight of 1.0.

If the if the 3,000 voters make the decision on their own, it's they're going to have a 99.7% chance of getting the right answer. The right collective making the right collective decision. This people know this idea. This is the Condra jury theorem result.

If you have a bunch of voters, they have the same probability of being right. They're independent. As you add voters, you get a bigger higher and higher probability of getting the correct social decision. As the number of voters goes to infinity, it goes to one. It's close to one here.

In this case, adding an expert voter doesn't help you much. It it does help you a little bit. You can't see it because it's it's it's less than the rounding error. Okay. Um so let's look at something else.

Now we're going to start playing with things. Let's look at some delegation situations. Let's consider a liquid democracy approach where you can delegate as much as you want. We have as many stages as we want. And let's suppose what voters do is they look for a more competent voter to delegate to.

Suppose everyone does that. Suppose this plays out perfectly. What's going to happen? All the votes get delegated to the expert voter. Now the probability of the correct social choice is 73%.

It's down from 99.7 which is a disaster. This is a problem of overdelegation. We see this in in experiments. We see it empirically.

We see it in theory especially with liquid democracy. It tends to have overdelegation. It looks like this. We're going to come back to that later. Um okay let's look at dependence.

Let's suppose that the 3,000 voters are perfectly correlated in their information. They just they all have the same information. Then they're basically the equivalent of one voter as far as information we looked at as a signals. They're just one signal. That should get a weight of 0.

1 total, not 300. Okay. And if we do that correct thing in this this top um piece here, then u then basically the expert outweighs them 10 to one. expert is going to make unilateral decision. We're going to have a 73% chance of being right.

If we screw up and use the weights we should is the weights we would have used in the independence case. Then these voters are have the whole decision process. They're essentially one voter and we're going to have a 52.5% chance of being right. Okay.

So again, this is a disaster. So if we fail to take dependency into in account, that's a big problem. Okay. So now um if we uh have the dependence case, all we have to do is replace the voters with signals, we can use the same formula. Whoops, darn it.

Um sorry about that. We can we can use the same formula here, but we're just with the signals. If we had a um a centralized party or uh or we have a smart contract that that that they could take all the signals, amalgamate it, make a perfect choice. But we're interested in a decentralized solution. What's that going to look like?

It's going to look like um the previous case here um in this example. So suppose these 3,000 voters realizing that they're dependent just each divided their weight by 3,000. That's the number of people who got the signal. If we could coordinate in that way, we could in a decentralized manner get the right result. They would they would downgrade their signal.

Okay. All right. So um let's move forward here. Uh let's move to terminology. Okay, terminology is a mess in this area.

Delegation and participation are vague, undefined, ambiguous. Okay, why is that? Let's look at at at at delegation. What if somebody abstains? Well, that's actually delegation.

What you're delegating to all the other people that vote. So a lot of times we see in the literature, well, we can get people to participate by having them and do delegate to other people. Well, guess what? All we're doing is switching the delegation from a delegation to all the people that otherwise would have voted to another delegation, which I'm going to call transfer delegation. When you transfer the votes of other people, I'm going to call transfer delegation to distinguish it from delegation by abstension.

Okay. Um the other thing I'm going to do, participation, everything is participation. like I delegate as participation. I vote directly as a participation. If I abstain, especially intelligently, why isn't that participation?

So, I need some terminology. I'm going to call direct participation. When I vote on my own, distinguish it from everything else. Okay. Now, we're going to make an important point about um uh the sec the bottom second main u uh point here.

There's a key difference between delegation by extension and transfer delegation. And this is going to be really important. You walk away with just this one thing. I'll be super happy. Okay.

If I delegate by abstaining, I am not interfering with the relative weights of the other voters. All I'm doing is withdrawing part or all of my own votes. If I transfer delegate, I'm interfering with the weights of the other voters. And that's going to be it's it's going to be really hard to do that in a way that improves the epistemic result. As a result, transfer delegation is going to look really weak epistemically.

I'm just uh previewing that uh result. Okay. Now, let's move forward to uh All right. Now, we're going to talk about delegation by partial abstension. This is going to be the winner in these environments.

How is this going to work? Well, remember that if we got if we got the same ratio of weights to tokens for all the voters, then we're going to get we're going to have to maximize the probability of making the correct collective decision. How do we do that? Well, let's let's just there's some voter who has the largest weight, largest ratio. What the other voters can do is they can reduce the number of tokens they vote.

That's going to push their ratio up. We're reducing the t in the denominator. They push it up to r. Everybody pushes up to this level r. We're golden.

However, how the heck do we know what r is? Okay, we don't. But fortunately, um this is very um solvable. Okay. Um, all right.

Um, what we can do is first of all, if we use an R that's too big, if we use if we use a coordinate of a number that's larger than the actual R, that's not a problem. It doesn't matter what that number is if it's higher because we're still going to get the if we just use some number, we're still going to get the right proportions. However, the danger is is we'll set a number and it'll be lower than that highest number. But if we we can do it we can do the following moves. First of all, we can take everybody's tokens.

Instead of the the number of tokens, we give everyone one vote. It doesn't matter because they're going to partially have state. It doesn't matter how many tokens they have. Okay? And now that means that the ratio the ratio is just exactly equal to the the competence the the the optimal weight for each person.

So somebody has a maximum optimal weight. What we'll do is we'll set a number that's really high compared to what that could be. We'll set it to say a thousand. We set it to a th00and. Okay.

So, we're targeting a number that's a thousand. There's still a possibility that somebody is even more confident than having a weight of a thousand. By the way, the weight of a th00and is way out there. However, if we make that mistake, the maximum loss of failing to make the correct collective decision is 5* 10 to the minus 435th power. Boom.

Okay. So all we have to do is rearrange it like this, create a big number and people only have to know their own competence and we can get the best possible collective decision. So this is very successful in this theoretical framework. Um okay and dependence we can handle by what exactly we've already seen it. We have to assume the voters know how many other voters got the same signal and everybody divides.

But notice what we're doing now. Now I have to know something about the other voters. In the previous case, I didn't. I only had to know my own thing. So my own probability of being right.

Okay. So this is a little more complicated. We're not going to spend any more time on that. Now we're going to skip six slides forward. Two, three, four, five, six.

Okay. Now we're going to talk about direct participation. If we have an optimal epistemic environment, what's that? That's where the uh voter coordination and the voting rule set it up so that we get we have optimal voting weights. If we have optimal voting weights in our decision process, adding people is great.

It can't make it worse and it usually makes it better. This is a wonderful world. We love participation. So this is where it's going going to work well. But if we don't have that then we're in big trouble.

Okay. Um, so a lot of people study majority voting. The sufficient conditions for improving the social choice are really strong. You have to add people with really high confidence to be sure that you're improving things. Furthermore, there's a lot of bad things that can happen.

We're going to look at three of them. One is a flooding danger. We'll look at that first. What is that? Well, I can prove I've proven in the paper that you can add people that are better than a coin flip.

They have a probability higher than 50%. I you tell me how low how low you want how close you want the social choice to be to be a coin flip. I can load voters in to get that result. What we're doing is put a lot of voters in with low competence. The epistemic quality plunges.

Okay, you have a flood. Okay, flooding is a big danger. If we go out there like some people want and say, "Okay, let's get all these people to participate. Get all these people to participate." Well, a lot of people are they're rationally ambivalent or rationally ignorant.

Like, if I own like one token, why should I spend a lot of time trying to figure out this protocol? Now, if you push me to to vote, it's going to be might be close to a coin flip. So, this can be a disaster. Okay? Even though we see a lot of people pushing this.

Okay. U dependency dish. We already saw it. You got a lot of people that have the same signal. You throw them in there, it's going to degrade.

It's going possibly catastrophically degrade the epistemic quality. Uh the third thing, epistemic danger from whales, very straightforward. You have a big token holder. They're a giant token holder. They might be very competent, but they're way overweighted.

Now, a whale has with a big stake has a huge incentive to get things right. They can partially abstain. they can back off to allow the other information to have some power. Okay, it doesn't work the other way. If the whales under if the highly competent whales underweighted, then we have a coordination problem.

All the small people have to coordinate with each other to partially abstain. Okay, that's enough of that. Now, we're going to move um uh seven slides forward. One, two, three, four, five, six, seven. I counted on having the thing where you can type in the number and it we don't have it.

It's I have only plus one and minus one. So I'm really apologize. All right. So this is the main result for transfer delegation. Uh it's really hard.

Okay. Why is that? Because we're interfering with the relative weights of the other voters. What do we have to know to do that? Well, we have to know all the to we have to know all the tokens that are going to be voted by the other voters.

The number we have to know the weights optimal weights of all the other voters. Oh my gosh. like this is not going to happen. Why do we have to do that? Because we're interfering with which coalitions win and we have to compare.

If we have a winning coalition under the tokens, we have to decide whether that would would be a winning coalition under the weights. And if not, we have to do some delegation to fix that. We have to go through all the winning coalitions and losing coalitions to figure this out. This is not going this is just not going to be very viable. So that's the main negative result.

Now I have only four minutes left. So we're going to drastically speed up Okay. Drastically. Okay. Like drastically.

In fact, I I blew this. I needed to count this. What I can use? What? Okay.

All right. Well, I'll take it. All right. Now, I have to figure out what slide I was on. Okay.

Um All right. So uh so now in this in this scenario that we just looked at we also made a simplifying assumption that that I delegated last. I got to see all the other people's delegations. Actually we have a coordination problem. We we're delegating simultaneously.

So I can't even see that. Well maybe if we have multiple steps that would help. I can see you know we have multiple steps and each time I can see the delegations of the other people. Maybe that's going to solve this and that's going to get us into liquid democracy. Okay.

Um All right. And u that has an idealism. You hear this all the time. Liquid democracy. It combines the best parts of direct democracy which is I I vote I get accountability.

Downside of direct democracy, I may not have expertise. I have no incentive to develop it. And representative democracy where I have people that have more expertise but have accountability problem. Okay. people like these politicians they they they have their own agenda that's not ours etc.

We all know that well and the idea of liquid markets I'm going to get the both best of both best of both worlds. Okay. But if we look at the actual if we look at it actually first of all there's a big problem with whether it stops. Liquid democracy there's an unlimited number of steps. Well first problem is there could be cycles.

A delegates to B. B delegates to C. C delegates back to A. That's never going to end. So with the theoretical literature we assume there are no cycles.

kind of like, you know, or we or we or or we say we're not going to count people that are voting in cycles. Okay, that's a cheat. But even after we cheat, we're in deep trouble because first of all, in general, there's no Nash equilibrium. That means we never get to a state where people no longer want to delegate more. Um, if we make assumptions that make a national equilibrium exist, we have to make an assumption about the social network.

Even if we make those assumptions under reasonable delegation dynamics where for example I'm delegating to improve what I the results for me then a lot of times there's no u equilibrium and if there is an equilibrium a lot of times it's crummy this is pretty bad but it's not it's not the worst the worst is yet to come the worst is the overdelegation problem and basically we see this um massively um we see this massively in a lot of research I'm just going to show you some uh things. So, this article, what they do is they say, "Okay, what we're going to do is people know they know some people around. They don't know everybody's weights, but they know some people's weights. They're only going to delegate to people that that beat their own level of competence by a certain amount, and there's going to be a set of those people. They're just going to pick one of those people and delegate to them, get massive overdelegation.

So, how do you fix it?" So what they do is they say okay let's have a mechanism where we cap the number of voters that a delegate can have. The cap is extreme. It's the order of the square root of log n where n is the number of voters. If you have a million voters that's that's less than four for you know and it does and it scales up really slowly.

Okay. So what what does this look like? It looks like representative democracy. Why? Because we have a bunch we have a large number of delegates like a large number of representatives in a parliament.

Each one has a fixed set number of people. It looks like that. And guess what? It it kicks butt. So this is an argument that representative democracy better than liquid democracy.

Okay, that's one loss. Now I'm not have time to go through these other things, but I want to emphasize one result from this is an excellent article by Morris Edall. uh what they show is that that majority voting dominates liquid democracy in an experimental setting where liquid democracy theoretically should win is a terrible bad result. Okay. And now we have the double now we've killed the idealism because majority voting itself beats liquid democracy and so does uh representative democracy.

Oh my god, our idealism is deflated from the literature. Okay. Um, now what? Um, all right. Now, now I have only five minutes left.

Okay. Holy crap. Sorry. Um, so there is a method I can improve liquid democracy. I can create a method of coordination that for so certain so certain social networks is going to work.

And I'm not going to have to go through I don't have time to go through it. But it's inferior to what partial abstension in all of these papers and in this thing. I know my own competence. Well, if I know my own compass, why are we doing all this delegation stuff? I can just partially abstain.

We can pick a large number and get the right result. And this only works if we have the network being connected. If it's disconnected, it looks like this. This is kind of each of those middle dots is the voter and the other dots are the voters that they know about their competence. the disconnect is not going to work.

Okay, the at the delegation for partial extension always works. You don't have to know anything about the other votes. Okay, here you have to know a lot. Okay, that's the end of that. And now what do we do?

Well, we're going to start looking at other things that we can add on. Well, sort I don't have time to cover it. It has epistemic problems. Okay. Um we're going to look at um we're going to look at um other other methods like futarchy and things like that to try to save us and I'm going to only look at there's several of them in the paper.

I'm going to look at two futuri and I'm also going to look at contestable control auctions which is another I've done a lot of work on that okay so futuri we know what that is you have a prediction market use the prediction mark to choose an action okay we predict the stock will go up the prediction market says the stock will go up and CEO quits the CEO is fired we fire the CEO okay now the problem with futur is now I have to go forward I don't know n slides here we go Um the problem with futuri is that um it it does not match up with the optimal voting rule. Even under idealized conditions for example if everybody has logarithmic utility we get uh we get a nice result which if everybody has logarithmic utility with few turkey what happens is the average belief is reflected in the price of the market. That's wonderful but it it produces a set of weights that's not the optimal set of weights. Let's see if I can pull that slide here quickly. Um, here it is.

You see this top bullet point here? If you uh create logarithmic utilities, you get this nice result about the average beliefs being reflected, but the weights are equivalent to 2p minus one, which is not optimal. It's underweing people with high competence. We can fix that in a in a very kind of somewhat esoteric way. uh we can uh have we can put a tax function in here and uh what we can show is that if we put the tax function in there and we u and we also force if the voters also are price takers and they're acting as if they're in a one-shot game and they have logarithmic utility we can get as close as we want to the optimal voting result which is a really cool result.

It's linking futur and a modified version of futarchy to optimal voting. However, people are not going to follow. People don't do this. Not everybody has logic utility. Most people are more are less riskaverse.

So, what do we do? AI agents. We put AI agents in. We force them to have this utility. We force them to be price takers.

Um, now we have to worry about where they can get information enough information and stuff. I have only 30 seconds left. Uh, so now we're going to jump into just temporarily. We'll jump into the last contestable control auction. What is this?

We create an auction for control. You can get temporary control of the DAO. It has some great properties. It has a really strong epistemic property. You're going to pick the best possible project and distribute most of the surplus to the existing token holders, not to the person doing the project.

That's wonderful. Also, it has another property which is really valuable given what we've just seen, which is if we can put this in here and still have a voting mechanism. We know a voting mechanism is likely to mess up. When it does, we have an auction temporary control. We reset things, fix the epistic problem, go back to the voting mechanisms.

We have voting mixes. We like participation or whatever. We can have it all potentially. Uh that's the end. And um this is the if you want to actually read some of this the actual details, many of which I omitted here.

This is the QR code of the archive thing. I'm going to revise this. It say has an addition here. I'm going to revise it probably in the next month. Then you'll have another one.

Then you'll have another one. You can keep if you really like this, you can keep reading it. Uh you might fall asleep after the first 10 pages and give up, but that's up to you. That's that's all I have to say. Thank you for your patience.

Sorry I had such a messy slide presentation. Thank you Jeff. Um thank you for this great talk. Since this is the last talk before the lunch break, I'll take the freedom to ask some questions if you like. Sure.

Um, so the first question is, how does the challenge of disinformation play into the waters being able to estimate the necessary probabilities? We don't have it on. Sorry. Okay, that's an incredibly powerful assumption. Probably totally unrealistic.

And it's even worse than that. I may, for example, be a fool. I have less than a 50% chance of having the right answer. Many people think that I don't know why they would think this that voters in like big countries that many voters are like this. I'm not going to make a judgment there, but that's going to be a disaster because then you get the reverse conversate jury theorem.

If voters are less than 50%. Then the more we pile on the closer we get to the wrong answer. Okay? And so um this is a strong assumption and this is what the p the way paper's designed. I've loaded it to try to make governance easy.

And even when we do that, transfer delegation, things like that fail dramatically. The the partial abstension doesn't fail. It succeeds. But we don't know if it's going to work in the real world because these assumptions are so strong. It might work.

It might be if we if we have a DAO together and we all say, "Oh, we're just going to try this. We're going to partially abstain. We're going to try to think about, you know, if I go on the internet, get my information." dude, like a lot of other people are doing that, so I want to I want to downgrade my weight. Maybe there's an approximate result.

We'd have to do experiments to figure this out or things like that. And so I don't know the answer to that. The next one, um, what is in your opinion the best governance approach that fits both communities and governments in which representative can be held accountable by for example pulling my vote from them instantly? Uh I think the best one is is is a contestable control auction. Why?

Because um no it eliminates entrenched control powerfully. Founders can't be have entrenched control. People do empty votes that borrow tokens and vote them. They can't have control. Why?

Because they have to win the auction. And if you win the auction, you put all kinds of safeguards in. When you win to win the auction, you have to say, "We're going to get to this token value under under under our leadership. You don't get to that token value. You pay.

You pay the the difference to the other token holders. If the token value goes down, you pay. We do deposits. So, we make them do that." Okay?

And we do we have an auction where the the the the winner of the auction is going to promise the most surplus to the other token holders. There is nothing like this in like securities, regulation, anything like that. Have you ever had had invested in something where you had this kind of protection against the expert people? You're going to get you're guaranteed to get what they're promising and you're guaranteed protected against losses. There's nothing like this.

An entrenchment is over forever. It's a new kind of property. We did this for regular corporations. Zuckerberg cannot keep control of Meta. He can lose an auction.

Musk can lose an auction. It's a whole different world. And it's a world that I think would be a really good world for Dows. We're trying it here in Europe, not in the United States. We're the first instances of trying this out in reality.

So I I'm strongly biased. I it's kind of an idea that I like and uh you can you should take that into account. There's a huge amount of bias in what I'm saying. But the paper that's based on is going to be published in the next few days. It's already in press in blockchain um research and applications.

that's coming out. It's also on archive. If you type go to archive, type in strernad, which is my weird name, and type in economic dows, you'll pull up the archive version. So now's going to be superseded by the published version. What do you think about sortion in parenthesis random appointment of representatives as alternative to representative democracy?

Yeah. Okay. So uh sorry about this. We have to suffer for this again here. Sortician.

Okay. Yeah. We pick a random sample of people and let them do things and it's a powerful idea. It goes back to the Greeks. The Greeks thought this is the real democracy and we Algarand proof of stake massively successful use of sort okay every time we there's a step in the consensus process like an algorand we choose a new random set of people so nobody can be bribed okay and um and those pe that works perfectly for that but it's a very simple application all the people are economically incentivized to know what they're doing and they're doing a a task over and over again.

It I think it's interesting possibility for um governance in the in in the regular world. Epistemically, it has some problems. Okay, consider this. Suppose there's a lot of diffuse information. For example, let's say we're all users of something and each of us has a view that's important.

We each have our own view of it. We'd like to get all that information. It's a conversation jury theorem type of idea. But if we just pick a small sample, guess what? We're cutting out a lot of information.

And if we if we and we can't pick too large a sample, why? Because these people have to talk to each other and figure out what to do. So this is a trade-off. Now, there's other problems with cert uh you can't force people to participate like we do this in a in a regular population. So we get a skewed sample.

We get the only people that are going to be there are going to people that either have the money or time to do it or activists. How would you like to do a sort where we have the most extreme people in the United States, for example, in the Democratic Republican party and those are the only representatives? Oh, I think I'd rather I'll move to Canada. Maybe I'll move to Canada. Maybe I'll move to Switzerland.

Maybe I'll move to Germany. Okay. Anyway, um so certain has epistemically it's difficult. You can read my the paper has a section on this and you can read it if you want if you want or you can skip it. Freedom.

Okay. Okay. Maybe the last quick one. Let's say uh do we ever see techniques like whales partially abstaining abstaining used in the real world for example do we see black rockck downweing themselves to improve the quality of corporate governant decisions maybe quickly uh yes we do and and we have a brilliant example possibly in um in the crypto world unis swap a6z giant token holder they hold at least 10% they might be 20% we have all these pseudonyms we don't know what do they do they have a token delegate program. They delegate most of their tokens to other people including the Stanford blockchain club.

2.5 million tokens. Maybe they can make proposals. Okay. And they create a degree.

They create technically independence of those voters from and it actually played out in the wormhole controversy. The blockchain clubs voted against uh A16Z. A6Z lost. Okay. Um so this is an example.

And how do they do that? If you read what their description, it's just like the optimal voting rule is we're going to pick a lot of diverse opinions. We're going to get a lot of information. We're going to pick people with high expertise. It sounds like now they don't say anything about the optimal voting rule, but it sounds like it.

Now, there's weaknesses in it, but that's a really strong example. And it's where it's where is it from? It's from Dows. It's from us. Okay.

Isn't that cool? Okay. I think so. But I think it's a and and in the corporate world, the main paper on this is a corporate paper, a corporate governance paper that talks about this. Okay, the the main theoretical paper and and we all do this.

You're a leader in something. Guess what? Am I going to go in and just assert my opinion? No. If I'm a good leader, maybe I'm listening to everybody.

I'm delegating. I'm delegating. A good leader is a good delegator, right? Duh. Okay.

So, we all do this. We do this in our life. I go home to my family, you know. Guess what? Guess what?

You know, all people, we've been in families, either what children or adults, doesn't matter. Yeah, I'm going to, you know, yeah, I'm, you know, I'm going to back off in a lot of situations. And it's rational. Some things I don't know anything about. You know, my wife knows more like, hey, you know, I'm I'm my I'm my my absention is going down to a very small number of tokens here.

And that's appropriate. Okay. Sorry, that's kind of academics. They talk too much.