In October I will be writing a Ruby on rails web app for experimenting with control markets to control a real world organisation.
Simple control markets* are used in Learning Classifier systems to make the system goal oriented. I’m heavily modifying it from that, but the basic idea remains the same, control of parts of the system is based on bidding a currency which is gained from feedback. I’m curious to see if they will work in the real world, or whether aspects of human psychology will make them untenable.
The theory behind them needs to be tightened up a lot as well. What you need to do is prove that the Evolutionary Stable Strategies for your control markets are non-malicious. But that is hard for me, at the moment.
*Warning—Neologism. I chose it for the allusion to prediction markets. Related: Agorics and market-based control.
Edit: An article not designed for lesswrong, can be found here please put any notes of what you would like in a lesswrong article there as well.
Heh. I’ve switched track; I’m working on another idea. Having code modified at load time to make exploiting it harder (due to return-oriented code not knowing where to exploit). That is going slow, mainly due to having to understand other peoples code and machine code, which I am rusty on. I’ve been meaning to contact developers of as and the elf-loaders.
The reasoning behind swtiching track. It’ll look better on my CV. It is easier to get people to adopt. And may help reduce existential risk/ world takeover of computers. It’ll take up less of my time when it is done, so it makes sense to do it first.
If anyone is interested in playing a part in the experimental organisation, let me know. I’ll write a post on it, once I have the software up and running.
Since I identify politically as a futarchist, I can hardly turn down an opportunity to participate in a real world experiment. I’m looking forward to the full post.
It is not strictly a futarchy (else I would have said), No voting is involved anywhere, and the only predictions the market makes are in the expected ability of the organisation to please the feedback mechanism.
So it has a different range of expected benefits and costs to futarchy.
Have you considered a test vessel for a control market? It seems to me a system with a known scale for success—like a Food Chain player—would be useful.
I’ve been thinking about that sort of test a bit. As the system is meant to be to control for incompetent/corrupt leadership, I doubt a group would do any better than the best normal players. So you would have to introduce corrupt players into the team for testing purposes. It doesn’t lend itself to well controlled testing.
In October I will be writing a Ruby on rails web app for experimenting with control markets to control a real world organisation.
Simple control markets* are used in Learning Classifier systems to make the system goal oriented. I’m heavily modifying it from that, but the basic idea remains the same, control of parts of the system is based on bidding a currency which is gained from feedback. I’m curious to see if they will work in the real world, or whether aspects of human psychology will make them untenable.
The theory behind them needs to be tightened up a lot as well. What you need to do is prove that the Evolutionary Stable Strategies for your control markets are non-malicious. But that is hard for me, at the moment.
*Warning—Neologism. I chose it for the allusion to prediction markets. Related: Agorics and market-based control.
Edit: An article not designed for lesswrong, can be found here please put any notes of what you would like in a lesswrong article there as well.
How has progress been?
Heh. I’ve switched track; I’m working on another idea. Having code modified at load time to make exploiting it harder (due to return-oriented code not knowing where to exploit). That is going slow, mainly due to having to understand other peoples code and machine code, which I am rusty on. I’ve been meaning to contact developers of as and the elf-loaders.
The reasoning behind swtiching track. It’ll look better on my CV. It is easier to get people to adopt. And may help reduce existential risk/ world takeover of computers. It’ll take up less of my time when it is done, so it makes sense to do it first.
I’ll get back on the web-app once that is done.
If anyone is interested in playing a part in the experimental organisation, let me know. I’ll write a post on it, once I have the software up and running.
Since I identify politically as a futarchist, I can hardly turn down an opportunity to participate in a real world experiment. I’m looking forward to the full post.
Thanks.
It is not strictly a futarchy (else I would have said), No voting is involved anywhere, and the only predictions the market makes are in the expected ability of the organisation to please the feedback mechanism.
So it has a different range of expected benefits and costs to futarchy.
Have you considered a test vessel for a control market? It seems to me a system with a known scale for success—like a Food Chain player—would be useful.
I’ve been thinking about that sort of test a bit. As the system is meant to be to control for incompetent/corrupt leadership, I doubt a group would do any better than the best normal players. So you would have to introduce corrupt players into the team for testing purposes. It doesn’t lend itself to well controlled testing.