I think it would be simplest to use an existing game. Players would send you their moves and you'd make the changes to the game by hand (at least at first). For possible games you could look for suitable turn-based games in this list of multiplayer browser games. Would poker or chess be a possibility?
Poker or chess may be a possibility, if you tweak them to require teamwork. Lets say each person gets a slightly distorted and partial view of the game state(this works better for chess than poker). People would have to share information to synthesize the likely true world state before the ace poker player/chess player (or bot people have made to play that game) could pick the moves.
I suspect that one answer would be creating a scoring mechanism for Artemis. One problem I foresee is that you need a game where the decision-making is nontrivial given perfect communication, or the game becomes some variant of Space Team.
A rules light game such as poker or chess would give you a lot of leeway in designing a scoring system and implementing the social systems, but probably has an insufficiently complex game state to allow for a large team size while still minimizing redundancy. If you want to develop for large teams (which is almost required to create a difference between true democracy and a representative system), I would suggest a highly customizable, complex game such as Civilization 5, perhaps by allowing each player to control and receive data from an initial unit with socially selected ability to control cities and subsequently produced units within the team.
I think that limiting information turns the entire game into maximizing strategic miscommunication instead of distributing decision-making.
This would avoid camaraderie, team spirit and reputation management being organisational factors.
Er, why would you want to do this? Do you have a specific management domain in mind, where these things actually don't matter?
If not, perhaps you can just watch what happens in carefully-selected, massively-multiplayer games? Eve, maybe?
Most people don't know the reputations or the personalities of the local councillors or board members they elect. It is these types of political situations I want to improve.
Is this a game about getting elected; about testing social structure within an existing framework?
Pretty much. You want to find a social incentive structure such that it is at least an evolutionarily stable strategy for the participants to try and maximise their teams score while being selfish agents. And is compatible with human psychology and mental limits.
Your summary does not appear to agree with mine; it seems like you want to create a game where the strategy which is personally selfish is also ideal for a group, which is wildly different from a game where getting elected scores points regardless of one's competence.
One solution would be an incentive system that managed to reward each individual proportionately to their contribution to the group; that roughly reduces to free-market capitalism, which appears to be bad at the intended goal (based on previous results). Does that issue boil down to difficulty determining the value of each person's contribution to total welfare?
I didn't quite get your message, so I think I interpretted it incorrectly.
My goal is a game where different organizational structures can be compared for effectiveness.
One of organizational structure would be voting. In this structure getting elected would be a win of some sorts. In another structure there would be a different win condition. Seeing which type of win conditions and rules motivated people the best, is the reason I want to make the game.
I agree there is problems with free market capitilism as it is currently practised. However there are lots of knobs we could twiddled about how they work. E.g. fractional reserve banking or long term land ownership and other monopoly issues. If we could try out these changes on a smaller scale and they are successful compared to our current systems, then we might be able to convince people to make real changes later on.
My point was that a system where people want to maximize their chances of getting elected is wildly different from a system in which people want to elect the person which maximizes group utility.
The bonus for getting elected in a democracy would have to come either out of a higher-sum total or at the cost of someone else in the group, not be free. Assuming all candidates are equally qualified and every voter has full knowledge, the person who believably promised the best kickbacks would end up elected, right? Any leader who took kickbacks for himself could be outbid by one that took smaller kickbacks- but at some point it would be better to be on the receiving side of the pork.
To find the winner in a democracy (with perfect knowledge, identical values, and fungible utility), determine how much total utility each person will generate if elected; the winner is the person who can maximize the total score; he distributes to half of the voters, excluding himself, as much as the second-place leader could have, plus epsilon, and takes the remainder for himself. The second-place leader and half the voters earn epsilon more than he would have if he were elected, and just under half the voters get nothing.
If we define the total score to be equal to the sum of the square roots of each individual's effort put forth, and the effort put forth by an individual to be equal to the log of their final expected score, (forcing a lower bound of 1 effort), that makes the total wealth generated by a democracy dependent on how it is distributed; can the leader of a democracy outperform the electorate under those rules?
Formal proposal: Teams consist of n characters, each of which understands the rules. The total score of the team is equal to the sum of the square roots of the 'effort' produced by each team member, and the effort produced by each team member is proportional to the expected log10 of the score assigned to that member by the leader. (Production is exponentially more expensive, and rewards are logarithmically less rewarding) (Method of determining score need not be deterministic) (individuals need not have the same proportionality constant relating score received and effort, but each team has an identical set of members) A) What form of distribution results in the highest maximum score for the team? Is it possible to have a team of n score higher than n times what a team of one scores? B) What method of selecting a dictator/distribution method results in the form of distribution that maximizes the team score, given that every individual is selfish and wishes only to maximize their own score?
Not sure if this is fruitful path (we would need to justify the logarithm and square root empirically). But it is an interesting problem. Assuming each person is equally productive for now. In pseudo code
S = score
P = vector of proportions
sum ( sqrt (log (p*S)) ) = S
This can be simplified (if my rarely used math muscles are correct) to
sum ( sqrt( log pi + log S) /S) = 1
I can't see anything to solve it analytically easily. So let us assume that we have 3 people and they are equally distributed to for now. As I expect this is the maxima?
sqrt(log(1/3) + log(S) ) / S - 1/3 = 0
Newton's method isn't being very helpful at the moment. I'll try some other numerical methods tomorrow.
Hmm. Wolfram Alpha suggests it doesn't have any positive solutions. Have I made an error in the maths?
Generalizing a bit more: Each player has an effort function E(s) which determines how much effort they exert based on their expected score; they also have a production function P(e) which determines how much they add to the team production based on their effort. (These two functions can probably be combined for all intents and purposes) Further, the team has a score function S(p), describing the total score of the team based on the total of the team member's production.
With a few constraints on those functions, I think I can guarantee at least one solution: All three functions are continuous, strictly increasing, and their first derivatives approach zero as the independent variable tends towards infinity. The form of group leadership divides the group score S among the individuals according to the methodology of leadership: a dictator chooses the distribution which maximizes his own score; each type of democracy selects the distribution which maximizes the score of enough of the team to win the election; a pure socialism divides the team score evenly between the members; a different system divides the score proportionately to each member's production; and the ideal system divides the score in such a manner as to maximize the team score.
Another possible formulation is if E is a function of expected proportion of score. People seem to be interested in relative status and also this would also stop the crazy feedback loop and possible unintuitive things like working harder when someone else is working with you than you do on your own (if they are a lot better at getting score than you and so increase your expected score, even taking a share).
I still favour empirical testing to see how people actually behave.
I tried, and failed, to account for my observation that if rewards are independent of effort, very low effort is expended.
How would you perform empirical testing with various budgets?
So lets start with the most convincing evidence that our theories are correct and work our way backwards to what might help us achieve that proof.
Best proof: There are many "real world" charities and businesses with the our significantly better structure we developed and they are out-competing (charitable outcome and profit, respectively) orgs with more conventional structures. Orgs with more conventional structures are forced to adopt the ideal structure to survive. This is likely to take many millions and require some changes to the law (charities having an elected board of directors, is baked into UK law I think). So how do we bootstrap to that? I suppose they could still fail if the "better" organisations interacted in some bad way with each other.
The idea is significantly viral (a better way of organising! minimising corruption!) that it will take off on its own if it gets even a modicum of proof in the real world. Sufficient proof might be something like: create multiple organisations with the same purpose, with different control mechanisms. Create them where people's work is cheaper for the low millions? Charitable organisations seem like the best bet here. They would have the same external economy and influences as organisations in the developed world. It would also be somewhat blind, there would not be the issue of the people involved with the organisation believing it was better and therefore working harder.
We would need to get some wealthy backers for this plan. They would probably want to see how our better type of organisation worked on a day to day at scale, so we would need one example of it at least. This I suggest would best be the Centre for Organisational Experiments that I suggested earlier. It would need a score of some variety, this would be how well the people funding it thought that it was experimenting and raising awareness of organisational experimentation. The sorts of things it might do: Funding theoretical examinations and occasionally the changing control structure and observing its own productivity. Depending on the enthusiasm of the people working there, this could be done for a 40-200k or so? You probably need at least one person working on it full time and then a number of other expenses, servers and the like. Motivation would not be entirely selfish, people would need to want to participate for the love of the idea. However, the more money involved the purer the experiment.
The experimental game. What this post was about. The biggest costs would probably be promotion and design of a compelling game. The more people that play, the larger scale organisations that can be tested. It also helps for funding for later experiments. If I code it all in my spare time and attempt to market it myself I could probably do it, although it would take a lot of time. I'd need to iterate through different game designs as well.
My budget is 1K now and about 50 a month. Ideally I want to be able to treat this as an expensive hobby. at least to start with.
I am in favour of theory. It will allow us to get better starting off points to experiment with. And then we can refine our theories.
So lets start with the most convincing evidence that our theories are correct and work our way backwards to what might help us achieve that proof.
Why not search for the evidence which falsifies our theories, and change our theories to the one with the strongest evidence?
My reference proposal would be a retail coffee shop, wherein the owners would receive a smaller percentage of 'net profits' than traditional, and the employees would be paid a minimum base rate (complying with minimum wage regulations), plus a prorated part of a significant percentage of the net profit of the business. Rather than try to prorate based on production, I was expecting to use a proxy such as 'hours worked'.
Hypothesis: A business which quarterly distributes a portion of its net profits back to all employees will have higher net profits than a control.
Proposed experimental evidence: Gather the most reliable information about the expected performance of a 'traditional' business, and create the experimental business instead. Compare the actual results with the traditional estimate. This can also be done by projecting the performance of an existing traditional business model and changing the model.
Potential outcomes:
Null: Net profit does not vary with portion of profits distributed to employees
Counter: Net profit diminishes with increasing portion of profits distributed to employees.
Variant: Net profit varies in a non-simple manner with portion of profits distributed to employees.
Hypothesis: Net profit increases with portion of profits distributed to employees.
(In all cases, I'm counting the base wages as an expense and a reduction to profit, but the profit redistributed remains in the profit column; this is probably wrong by accounting standards. As a result, it is expected that there is some point where increasing profit-sharing increases net profit while decreasing stockholder returns.)
Not considered: Distributing an amount of money which is a strictly increasing nonlinear function of net profit.
I've been researching this while making a website to do with experimenting with organizations to reduce corruption in them. I came across this reference. From the way it was quoted it suggested that profit sharing wasn't effective, but random checking with a very low punishment pay if insufficient effort was. I've not read it, but thought you might find it interesting. It is going on my pile of things to read.
Why not search for the evidence which falsifies our theories, and change our theories to the one with tIghe strongest evidence?
How much money do you have to experiment with? If it is not very large, we have to consider the ability of whatever experiments we do to enable us to raise money for more experiments.
About 20% of small businesses fail in the first year, what happens if our coffee shop, for some reason, is one of them? Just having a better organisational structure does not mean it will be free of accidents or illnesses. And I am not worried about the loss of money, but that a single business failing or succeeding won't allow us to falsify a hypothesis. A small business with a better business structure may only have a 10% chance of failing but we would need more trials to tease out the confounding variables (of which there are many).
I would also need to look into the history of cooperatively owned and other profit sharing businesses, but as they have not taken over the world I doubt they are strictly better than non-profit sharing businesses.
Hmm. There are probably at least 3 things involved in the low score low effort issue.
1). Energy returned on energy invested. Effort is energy, if you can't make appreciably more energy (or things that can be converted into energy) by expending your energy for the organisation, you may as well save your energy.
2). Other opportunities: In the real world there is often other ways of getting score/energy so you should use your effort to do those rather than things with a poor pay off.
3) Other players in different organisations: Even if the game is the only opportunity available to you, you still may be competing against people playing different games with different payoffs. Take buying a house for example (the biggest relative pricing issue people generally face), even if you are a dictator in an organisation with a small score you may still not be able to buy a house if you are competing in the same market as communist in an organisation with a very large score. So the dictator might not be motivated to expend all his effort if it still can't get his dream house.
Ignoring EROEI, perhaps proportion of total score (of every player) might be better as the input. This brings back the difficulties of the feedback loop though.
I'll think about different testing methods over dinner.
I was thinking about modeling effort as negative utility, and reward as positive utility, but that only works to model rational agents that share those assumptions.
taking the log of something less than one gives a negative result. I'm too tired to do the math right now, but I'll give it a shot when I can.
In a game with a small set of fixed rules innovation is relatively unimportant. In the real world it matters a great deal.
I had thought about including some programming element in the game, for this sort of reason. Getting people to program different subsystems of a machine that would fight with other people's machine, for example. But this would cut down on the number of people who felt they would enjoy the game. I'll probably start simple and try and drive interest and then do something more complex.
Say we want to try out new organizational structures. Zaine suggests that a game might be a good method. However rather than a game to test a specific method of organizing people, I'm going to make a game where different organizational structures can be pitted against each other and statistics about their operation over time can be collected to inform new organisation designs.
Some organizational structures that might be tested include Democracy, Futarchy, Control Markets, Histocracy, some form of Meritocracy and Direct Democracy.
The conditions under which organizations suffer from corruption of purpose more frequently are when the people inside the organization are generally selfish and only moderately interested in the goals of the organization. So it makes sense to concentrate on these sorts of conditions.
I will be using the terminology defined in this article to talk about different facets of an organization.
One other bit of terminology: Team, a group of players given an organizational structure to test.
Although to simplify things we shall ignore Stakeholders, unless they are strictly necessary, instead relying on how well the teams perform in the game as Feedback.
Social Condition Creation
In order to make people selfish we need at least an individual high score table. Also people should be anonymous, assigned their teams randomly and communication restricted between them so that they interact with each other like strangers. This would avoid camaraderie, team spirit and reputation management being organisational factors.
Game play
The design of the game is a tricky subject in itself.
It would need to be:
Scoring
I envision the high score table being an average of how you do during each game, with people having to not be below a couple of standard deviations of the average number of games played to be ranked. You couldn't play one game, ace it and retire, you would have to be consistently good.
Each organisation type would have a different scoring method and different high score tables.
Control Markets would naturally have the amount of funge acquired by a team member as a score.
Members of a Futarchy would have their remaining money as a score, perhaps scaled by the score of the team.
Democracies might allow the leader(s) to pick a percentage of the score acquired for a round to disburse to the general team members as an incentive for them to help out and pick a good leader, the rest being kept by the leader(s). Or more cynically the score for a democracy match would be the number of times you got elected. Perhaps both scores could be tracked.
Metrics
The simplest metric to collect would be which organization types did the best on average. But in depth information could be collected on why teams fail in different organization types. The reasons might include in such as lack of engagement, infighting or underhand sabotage. Actor behaviour over different organization types could be analysed.
Downsides
It is a pretty artificial setting, so even if one structure did well in the game, it might not do well in real life. When you add in Stakeholders or an external economy the dynamics may well change a lot.
Comments and ideas appreciated!