I have heard pessimism lately about whether democracy can produce good decisions frequently enough to stop everything rapidly going to Hell. Primary concerns are that voters are ignorant and that voters are evil. Supposing voters are evil, arguably any good system of government should bring about Hell. However the ignorance seems like a real issue.
A popular response to all complaints about democracy is, ‘Well, what else are we going to do? Do you want dictatorship?’
I think this ignores the potential for mild variations on democracy. ‘Democracy’ is not very specific. Wikipedia lists a bunch of variations. I’d like to suggest a different one.
The basic problem I want to solve is that the people voting for policies (directly or indirectly) are ignorant about the likely consequences of policies.
But first I’d like to point out that this is a problem for everyone, not a conflict between an ignorant team and an informed team. That ignorant people vote for destructive policies is at least as bad for the ignorant people as it is for everyone else. That is, if people truly vote for bad policies due to ignorance, they would presumably prefer the outcomes that they voted against.
II. A (hand-wavy) solution
My proposal is for people to vote on what they want to happen, and then for someone else to put in the hard work of figuring out which policies correspond to which outcomes. That is, to vote on values.
Robin Hanson suggested this in Shall we Vote on Values, but Bet on Beliefs? (2013), as a component of Futarchy—a system where people elect representatives to stipulate their values, and use prediction markets to judge which policies will satisfy those values. Robin is mostly excited about the prediction markets aspect, but I think the idea of separating out values from policies is important on its own. Prediction markets are but one thing a population might use to figure out what to do, once they knew what their group as a whole wanted. Arguably a pretty good thing, but still. Any kind of voting on values then doing something else about beliefs seems like it would have a number of benefits unrelated to prediction markets.
We can think of selecting policy as something like:
Everyone has their own values and empirical beliefs. Everyone has to share policies. Values and empirical beliefs together determine the best policies. We basically want everyone’s values to be represented in the policies. It is not important that everyone’s empirical beliefs are all represented though—if we had a good way of just using the most accurate beliefs to bring about everyone’s values, the people with the least accurate beliefs would still be better off.
Usually the combining of values and empirical beliefs into policy recommendations happens within each person’s head. Then we aggregate policies, via voting on them directly or voting for representatives who agree with us on policy. Instead, we could aggregate values alone, and combine the aggregated values with empirical data gleaned some other way.
III. Good things
I claim this would achieve the good things about democracy—e.g. accounting for everyone’s interests, fairness, avoiding extreme evils, reducing reasons for conflict—at least not much worse than the current system, while mostly mitigating the problem that most people are ignorant or misinformed about most things.
I think there are also a lot of other benefits. Here are benefits of voting on values that I can think of:
More accurate sources of empirical belief. There are lots of better ways to get accurate empirical views than taking a national vote. The problem is often summarized as ‘people are stupid’ and ‘people are uneducated’, but even smart, educated people are probably very ignorant about the policies they vote on a lot of the time, relative to experts. It would just be an infeasibly huge amount of work to have informed views about the myriad policy questions a person has some tiny amount of political influence over.
Much less effort. Instead of every person in the country figuring out which policies lead to which outcomes (a very tricky problem), it only has to be done once.
More efficient use of information. If everyone’s beliefs constitute noisy evidence about the true state of the world, and each person uses only their own beliefs to choose their favorite policy, most information that could be used for each choice about policy preference is not being used. If beliefs are aggregated in some way and and then applied to aggregated values, this uses all of the information.
More fairness to those with few resources. The status quo means that uneducated people are less likely to get the outcomes they want, because they are more likely to vote for policies that don’t support those outcomes, due to misunderstanding. This proposal should avoid that bias.
Less destruction from voting to express values. Arguably, most of the consequences of a person’s political positions are on friends’ and acquaintances’ perceptions of the person. So we might expect political choices to be partly optimized for signaling values and qualities, rather than for optimal policy consequences. If people voted on values rather than policies, this would superficially seem to make advertising your values and qualities more straightforward, and less destructive, because expressing your values is just what you are supposed to be doing.
Several of these seem pretty big.
IV. Tricky things
There are also several obvious difficulties. A first difficulty is converting values into policies without the interference of the values of those people involved in doing the conversion. To put it less abstractly, if my nation decides that it values jobs a certain amount, and I am in charge of figuring out how to best create some jobs, and I don’t like people having jobs in forestry, you have to somehow stop me from just lying about whether forestry is a good place for creating jobs.
While this seems hard to prevent entirely, there is already a lot of indirection between what people vote on and what happens in our current system. And probably this already biases outcomes far in favor of what intermediaries want. So the bar for improvement is not very high. I expect we could make a system of voting on values that was better than the current system in this regard.
Another difficulty is that there isn’t a clearly good format for values to take while they are being voted on. Do you tick a box next to ‘people should be richer’? Probably not. Your vote would need to indicate how much some values are worth relative to others, and there are just a lot of things to value, and they don’t come in convenient units. Robin’s paper proposes a solution involving representatives, which at least demonstrates that this can be solved. I expect there are other ways to do it.
A related difficulty is deciding what kinds of things can be values. If you are going to aggregate everyone’s values, they will probably need to be in some common and easy to vote on format, which will probably restrict expressiveness. Again the bar for improving on the current system is not high however—choosing between two representatives whose level of agreement with your policy preferences is mostly explained by your being of the same species as them also probably reduces expressiveness.
There are probably also heaps of other problems that I haven’t thought of now. I’m mostly suggesting that this is worth thinking about, rather than presenting a detailed proposal that doesn’t have terrible problems.
V. Alternatives (that are worse)
Sometimes people suggest that citizens just not be allowed to vote unless they meet certain intellectual standards, such as basic knowledge of the part of the world they are voting about. This would have the dreadful downside that the people who know less about the broader world—perhaps because they don’t have the resources to invest in reading about such things—have their interests completely ignored. Yet people find such proposals perennially appealing. I think voting on values is the natural resolution to this conflict between wanting to represent the interests of people who are not deeply educated about all policy-relevant aspects of the world, without absorbing their empirical misunderstandings.
Vote on values
I. A problem
I have heard pessimism lately about whether democracy can produce good decisions frequently enough to stop everything rapidly going to Hell. Primary concerns are that voters are ignorant and that voters are evil. Supposing voters are evil, arguably any good system of government should bring about Hell. However the ignorance seems like a real issue.
A popular response to all complaints about democracy is, ‘Well, what else are we going to do? Do you want dictatorship?’
I think this ignores the potential for mild variations on democracy. ‘Democracy’ is not very specific. Wikipedia lists a bunch of variations. I’d like to suggest a different one.
The basic problem I want to solve is that the people voting for policies (directly or indirectly) are ignorant about the likely consequences of policies.
But first I’d like to point out that this is a problem for everyone, not a conflict between an ignorant team and an informed team. That ignorant people vote for destructive policies is at least as bad for the ignorant people as it is for everyone else. That is, if people truly vote for bad policies due to ignorance, they would presumably prefer the outcomes that they voted against.
II. A (hand-wavy) solution
My proposal is for people to vote on what they want to happen, and then for someone else to put in the hard work of figuring out which policies correspond to which outcomes. That is, to vote on values.
Robin Hanson suggested this in Shall we Vote on Values, but Bet on Beliefs? (2013), as a component of Futarchy—a system where people elect representatives to stipulate their values, and use prediction markets to judge which policies will satisfy those values. Robin is mostly excited about the prediction markets aspect, but I think the idea of separating out values from policies is important on its own. Prediction markets are but one thing a population might use to figure out what to do, once they knew what their group as a whole wanted. Arguably a pretty good thing, but still. Any kind of voting on values then doing something else about beliefs seems like it would have a number of benefits unrelated to prediction markets.
We can think of selecting policy as something like:
Everyone has their own values and empirical beliefs. Everyone has to share policies. Values and empirical beliefs together determine the best policies. We basically want everyone’s values to be represented in the policies. It is not important that everyone’s empirical beliefs are all represented though—if we had a good way of just using the most accurate beliefs to bring about everyone’s values, the people with the least accurate beliefs would still be better off.
Usually the combining of values and empirical beliefs into policy recommendations happens within each person’s head. Then we aggregate policies, via voting on them directly or voting for representatives who agree with us on policy. Instead, we could aggregate values alone, and combine the aggregated values with empirical data gleaned some other way.
III. Good things
I claim this would achieve the good things about democracy—e.g. accounting for everyone’s interests, fairness, avoiding extreme evils, reducing reasons for conflict—at least not much worse than the current system, while mostly mitigating the problem that most people are ignorant or misinformed about most things.
I think there are also a lot of other benefits. Here are benefits of voting on values that I can think of:
More accurate sources of empirical belief. There are lots of better ways to get accurate empirical views than taking a national vote. The problem is often summarized as ‘people are stupid’ and ‘people are uneducated’, but even smart, educated people are probably very ignorant about the policies they vote on a lot of the time, relative to experts. It would just be an infeasibly huge amount of work to have informed views about the myriad policy questions a person has some tiny amount of political influence over.
Much less effort. Instead of every person in the country figuring out which policies lead to which outcomes (a very tricky problem), it only has to be done once.
More efficient use of information. If everyone’s beliefs constitute noisy evidence about the true state of the world, and each person uses only their own beliefs to choose their favorite policy, most information that could be used for each choice about policy preference is not being used. If beliefs are aggregated in some way and and then applied to aggregated values, this uses all of the information.
More fairness to those with few resources. The status quo means that uneducated people are less likely to get the outcomes they want, because they are more likely to vote for policies that don’t support those outcomes, due to misunderstanding. This proposal should avoid that bias.
Less destruction from voting to express values. Arguably, most of the consequences of a person’s political positions are on friends’ and acquaintances’ perceptions of the person. So we might expect political choices to be partly optimized for signaling values and qualities, rather than for optimal policy consequences. If people voted on values rather than policies, this would superficially seem to make advertising your values and qualities more straightforward, and less destructive, because expressing your values is just what you are supposed to be doing.
Several of these seem pretty big.
IV. Tricky things
There are also several obvious difficulties. A first difficulty is converting values into policies without the interference of the values of those people involved in doing the conversion. To put it less abstractly, if my nation decides that it values jobs a certain amount, and I am in charge of figuring out how to best create some jobs, and I don’t like people having jobs in forestry, you have to somehow stop me from just lying about whether forestry is a good place for creating jobs.
While this seems hard to prevent entirely, there is already a lot of indirection between what people vote on and what happens in our current system. And probably this already biases outcomes far in favor of what intermediaries want. So the bar for improvement is not very high. I expect we could make a system of voting on values that was better than the current system in this regard.
Another difficulty is that there isn’t a clearly good format for values to take while they are being voted on. Do you tick a box next to ‘people should be richer’? Probably not. Your vote would need to indicate how much some values are worth relative to others, and there are just a lot of things to value, and they don’t come in convenient units. Robin’s paper proposes a solution involving representatives, which at least demonstrates that this can be solved. I expect there are other ways to do it.
A related difficulty is deciding what kinds of things can be values. If you are going to aggregate everyone’s values, they will probably need to be in some common and easy to vote on format, which will probably restrict expressiveness. Again the bar for improving on the current system is not high however—choosing between two representatives whose level of agreement with your policy preferences is mostly explained by your being of the same species as them also probably reduces expressiveness.
There are probably also heaps of other problems that I haven’t thought of now. I’m mostly suggesting that this is worth thinking about, rather than presenting a detailed proposal that doesn’t have terrible problems.
V. Alternatives (that are worse)
Sometimes people suggest that citizens just not be allowed to vote unless they meet certain intellectual standards, such as basic knowledge of the part of the world they are voting about. This would have the dreadful downside that the people who know less about the broader world—perhaps because they don’t have the resources to invest in reading about such things—have their interests completely ignored. Yet people find such proposals perennially appealing. I think voting on values is the natural resolution to this conflict between wanting to represent the interests of people who are not deeply educated about all policy-relevant aspects of the world, without absorbing their empirical misunderstandings.