Control markets are systems where the control of resources and the system is determined by a market, the currency of that market is given out dependent upon how well the system as a whole is doing.
They have been discussed and explored in computer systems for a while with Agorics, Learning Classifier systems and Eric Baum’s work being two notable examples (ZCS being the closest LCS to it). These are limited and constrained markets, in that the communication and computational expressiveness of them are limited. However unlimited control markets may be of use in the real world to control an organization in a flexible fashion. For example it might be useful for shareholders to control a board or possibly the whole company. Even if it isn’t compatible with human motivational systems and real world conditions, discussing and thinking about these sorts of systems may enable us to find better organizational structures than our current ones.
Control market can be seen as trying to embed reinforcement learning into an organization.
When we create organizations we don’t tend to design their control structures from a theoretical basis. We adopt the prevailing structures, such as a representatives, boards of directors or trustees. So are there organizational systems that can do better at keeping the organizations to their purpose1? Any exploration will have to be a mix of theory and practice.
Let us define a few terms:
Stakeholders: People with some information about how the organization should act. E.g. Voters.
Feedback: That which stakeholders can provide to change the organization. E.g. Voting would count.
Actors: The people that act within the organization. Generally considered at least somewhat selfish.
Resources: Parts of the organization that can be controlled by an Actor.
We have little control on what stakeholders and actors are like, so we can only control the nature of feedback. So what desiderata do we want for feedback?
Feedback should count: People have argued that voting is generally irrational or pointless. Lets avoid that.
Feedback should be easy: A choice, a single number or small group of numbers.
Feedback should work in a predictable fashion: Stakeholders and Actors are limited computational agents so cannot evaluate complex mechanisms.
Negative feedback shouldn’t hurt the organization: Some forms of feedback, such as voting with your wallet when choosing which charity to donate to, hurt the organization’s ability to perform their goal. You may not wish to do this, but simply change a small facet of the organization.
Making feedback about issues not people: A priori feedback, such as voting, means that you only have the promises and personality of the person to make a decision upon. Post-hoc feedback allows you to see how an Actor has performed during a time period.
Actors should be encouraged to show rationality: Politicians are generally optimistic about things like the economy. If the system encouraged them to show what they thought the future held and be accurate, this would be valuable.
Checks and balances: No actor or group of actors should be able to get too much power
Ignoring checks and balances for a moment, a simple Control Market would be this:
There is one resource.
There is an internal currency called funge (short for fungible, rhymes with sponge) that is used by Actors to bid upon control of the resource for a time period. The one that bids the highest wins (I favour vickrey auctions2).
Stakeholders evaluate how the resource has been used in the time period and give a number. That number is aggregated in some simple fashion (say summed) and given as funges to the controller of that resource.
There are any number of ways of designing the auctions and the system. You probably want to try and avoid inflation. Some theory for this would be good to develop.
So funge is cycled around, sunk into bidding for a resource and sourced from the feedback. So how are our desiderata?
Feedback should count: Changing your feedback will make a slight change in the amount of funges the Actor gets. Even an individual.
Feedback should be easy: Yes. Just a single number.
Feedback should work in a predictable fashion: Just a few rules, so pretty simple.
Negative feedback shouldn’t hurt the organization: If you give on average 5 funges giving less than that will be considered negative (due to expectations)
Making feedback about issues not people: It is post-hoc so you can see how people have done. There is still an issue of cult of personality, you might give more negative feedback to someone you don’t like vs someone you do. You don’t need to know who is in control, so that might offer a way out.
Actors should be encouraged to show rationality: If a bad period is expected in the future, then people will bid less for a resource as the expected funge they will get back will be lower. If someone over bids they won’t do so well.
However it gives ultimate power to one person. So fails the checks and balances desideratum. Checks and balances can be achieved somewhat by adding in more Resources that can be controlled by multiple Actors. How many resources you need to create to make sufficient checks and balances, is a question I don’t know how to tackle, especially if each Resource gets individual Feedback which adds more complexity for the Stakeholder. You might have non-Feedback giving Resources and say that they cannot be controlled by the same Actor so that power is spread.
There is at least one other issue: Bootstrapping, how does an person become an Actor and get funges, if you need funge to buy resources to get funge? It seems likely that parties will spring up and loan funge to up and coming potential Actors in expectation that they pay it back with interest. Whether these parties will be poisonous and powerful as a current ones are, is worth investigating. Aspiring Actors may also help Actors perform their jobs, with out any control of a Resource, with funge as a reward. This would allow them to save it up to be able to get control of a resource.
If funge is transferable to money, there is an issue, Actors might simply get money and then use that as an incentive for people, so that they will never have to give funge to someone (and potentially create a competitor for them).
If you are getting a feeling of deja vu I linked to my blog about this a while back. The reason why I am posting this now is that I am somewhat in a position to put some time/money towards creating infrastructure for testing a Control Market. I’ll do another post with a call for participation.
Coda: So what is wrong with our current currencies? Fractional Reserve banking and money printing. This gives the power to people who can inflate the money supply, devaluing the effort people have previously put in to the system.
1 A group of highly motivated people with a common vision are probably the best way. But most companies and charities probably can’t hope for this. Also if an organization starts off like this it may not stay like it over time, e.g. USA.
2 Vickery is sealed second bid. So people can’t see what other people are bidding but it encourages truthful bidding.
An Introduction to Control Markets
Control markets are systems where the control of resources and the system is determined by a market, the currency of that market is given out dependent upon how well the system as a whole is doing.
They have been discussed and explored in computer systems for a while with Agorics, Learning Classifier systems and Eric Baum’s work being two notable examples (ZCS being the closest LCS to it). These are limited and constrained markets, in that the communication and computational expressiveness of them are limited. However unlimited control markets may be of use in the real world to control an organization in a flexible fashion. For example it might be useful for shareholders to control a board or possibly the whole company. Even if it isn’t compatible with human motivational systems and real world conditions, discussing and thinking about these sorts of systems may enable us to find better organizational structures than our current ones.
Control market can be seen as trying to embed reinforcement learning into an organization.
When we create organizations we don’t tend to design their control structures from a theoretical basis. We adopt the prevailing structures, such as a representatives, boards of directors or trustees. So are there organizational systems that can do better at keeping the organizations to their purpose1? Any exploration will have to be a mix of theory and practice.
Let us define a few terms:
Stakeholders: People with some information about how the organization should act. E.g. Voters.
Feedback: That which stakeholders can provide to change the organization. E.g. Voting would count.
Actors: The people that act within the organization. Generally considered at least somewhat selfish.
Resources: Parts of the organization that can be controlled by an Actor.
We have little control on what stakeholders and actors are like, so we can only control the nature of feedback. So what desiderata do we want for feedback?
Feedback should count: People have argued that voting is generally irrational or pointless. Lets avoid that.
Feedback should be easy: A choice, a single number or small group of numbers.
Feedback should work in a predictable fashion: Stakeholders and Actors are limited computational agents so cannot evaluate complex mechanisms.
Negative feedback shouldn’t hurt the organization: Some forms of feedback, such as voting with your wallet when choosing which charity to donate to, hurt the organization’s ability to perform their goal. You may not wish to do this, but simply change a small facet of the organization.
Making feedback about issues not people: A priori feedback, such as voting, means that you only have the promises and personality of the person to make a decision upon. Post-hoc feedback allows you to see how an Actor has performed during a time period.
Actors should be encouraged to show rationality: Politicians are generally optimistic about things like the economy. If the system encouraged them to show what they thought the future held and be accurate, this would be valuable.
Checks and balances: No actor or group of actors should be able to get too much power
Ignoring checks and balances for a moment, a simple Control Market would be this:
There is one resource.
There is an internal currency called funge (short for fungible, rhymes with sponge) that is used by Actors to bid upon control of the resource for a time period. The one that bids the highest wins (I favour vickrey auctions2).
Stakeholders evaluate how the resource has been used in the time period and give a number. That number is aggregated in some simple fashion (say summed) and given as funges to the controller of that resource.
There are any number of ways of designing the auctions and the system. You probably want to try and avoid inflation. Some theory for this would be good to develop.
So funge is cycled around, sunk into bidding for a resource and sourced from the feedback. So how are our desiderata?
Feedback should count: Changing your feedback will make a slight change in the amount of funges the Actor gets. Even an individual.
Feedback should be easy: Yes. Just a single number.
Feedback should work in a predictable fashion: Just a few rules, so pretty simple.
Negative feedback shouldn’t hurt the organization: If you give on average 5 funges giving less than that will be considered negative (due to expectations)
Making feedback about issues not people: It is post-hoc so you can see how people have done. There is still an issue of cult of personality, you might give more negative feedback to someone you don’t like vs someone you do. You don’t need to know who is in control, so that might offer a way out.
Actors should be encouraged to show rationality: If a bad period is expected in the future, then people will bid less for a resource as the expected funge they will get back will be lower. If someone over bids they won’t do so well.
However it gives ultimate power to one person. So fails the checks and balances desideratum. Checks and balances can be achieved somewhat by adding in more Resources that can be controlled by multiple Actors. How many resources you need to create to make sufficient checks and balances, is a question I don’t know how to tackle, especially if each Resource gets individual Feedback which adds more complexity for the Stakeholder. You might have non-Feedback giving Resources and say that they cannot be controlled by the same Actor so that power is spread.
There is at least one other issue: Bootstrapping, how does an person become an Actor and get funges, if you need funge to buy resources to get funge? It seems likely that parties will spring up and loan funge to up and coming potential Actors in expectation that they pay it back with interest. Whether these parties will be poisonous and powerful as a current ones are, is worth investigating. Aspiring Actors may also help Actors perform their jobs, with out any control of a Resource, with funge as a reward. This would allow them to save it up to be able to get control of a resource.
If funge is transferable to money, there is an issue, Actors might simply get money and then use that as an incentive for people, so that they will never have to give funge to someone (and potentially create a competitor for them).
If you are getting a feeling of deja vu I linked to my blog about this a while back. The reason why I am posting this now is that I am somewhat in a position to put some time/money towards creating infrastructure for testing a Control Market. I’ll do another post with a call for participation.
Coda: So what is wrong with our current currencies? Fractional Reserve banking and money printing. This gives the power to people who can inflate the money supply, devaluing the effort people have previously put in to the system.
1 A group of highly motivated people with a common vision are probably the best way. But most companies and charities probably can’t hope for this. Also if an organization starts off like this it may not stay like it over time, e.g. USA.
2 Vickery is sealed second bid. So people can’t see what other people are bidding but it encourages truthful bidding.