Just as Steve Balmer said "Developers Developers Developers",  I like to say to the crypto community  "mechanism design mechanism design mechanism design". What is the basis behind the mechanism design of Bitcoin? Essentially game theory is what is behind the mechanism design of Bitcoin. 

In the paper which is called "Bitcoin mining pools: A cooperative game theoretic analysis" we can learn in my opinion how Bitcoin actually works. In my opinion it is this game theory which is the secret sauce behind the Bitcoin mechanism design. This game theory of Bitcoin was based on some assumptions and I wonder whether mining pools were a predicted outcome. If the game theory was based on the assumption that miners will always compete as in classic game theory and that conglomerates would not emerge as cooperative game theory would expect then something about Bitcoin security itself could be gleaned from studying the cooperative game theoretic analysis which in my opinion need not be duplicated by Tauchain.

In my post I will even make a case and offer some ideas based on my current thoughts on how we could perhaps avoid some of the assumptions made by Bitcoin. I will say though that no solution is going to be a 100% guarantee and even if you avoid some assumptions you always have to make other assumptions in exchange as all decisions of this kind of complexity are going to be based on some assumptions. The sort of assumptions I'm talking about include the nature and motivations of the participants in the network and I think one flawed assumption is to assume all participants are culturally the same, are competitive rather than cooperative, are motivated primarily by monetary gain, etc. 

My first thought based on my previous post on the impact of testosterone on the blockchain community is that maybe we need to go with a broader and more developed model of the participant in the network. For example, mining assumes something culturally about the participants in that particular game. It may in fact be true that a lot of people are motivated by money, or by greed, in such a way that they'll mine competitively and not create coalitions, gangs, or tribes, but this doesn't seem to be true in every case. In fact I would say if we look carefully we can see tribalism here and now in the crypto community between different blockchains, based around different thought leaders, or based around differences in political views, all which could allow for behaviors which aren't necessarily fueled by monetary gain.

I will now quote Matthew Finestone:


"The goal of game theory for cryptocurrency is to model human reasoning to build networks that need no oversight yet have positive outcomes for the greater good. Unfortunately, planning for unpredictable human decisions first requires that we understand what motivates people, which is easier said than done."

My argument is that perhaps we need to apply a better model of human reasoning. The model of human reasoning we saw in Bitcoin or even the model of human reasoning we see in Steem seems to be short sighted in my opinion. If the goal for example is to get as many participants as possible involved (for decentralization) then you would have to develop a model of human reasoning beyond just monetary or greed based incentives.

If we look at Steem as an example of what I think went wrong well just look at Steem Power? Steem Power doesn't represent any sort of social power on Steem or in the real world. Does having more Steem power get a person into the most exclusive clubs? Does it get people higher up on the status ladder of Steem or in the crypto community? No? And how does Steem power function as a store of wealth? As you can see it's not currently very effective because from a social dynamics point of view it simply doesn't add very much to someone's life if they have a lot of Steem Power. If a person is a multi millionaire or billionaire they may want to buy power on a platform but what kind of power does Steem Power give right now? 

Using Applied Algorithmic Game Theory to arrive at a mechanism design for Tauchain

First I want to make it known that I'm no expert at algorithmic game theory. I'm merely someone who did enough research to understand the critical importance of using algorithmic game theory in the context of improving Tauchain, Steem, or any blockchain project.  When I stress the importance of getting the economics right it is this field which if applied can help with getting the economics right. 

Algorithmic game theory is like a form of computational game theory, and part of computational economics.  Algorithmic mechanism design is essentially part of the solution where the market mechanisms can be studied in detail, in a manner accessible to computer scientists.  For example topics such as auctions (Vickrey Auctions) involve and implement mechanism design, and mechanisms for price discovery also deal with mechanism design. To have for example a stable token or not, is in my opinion also important to mechanism design, and I might add that it is also important to understand the psychology of the average participant (most research indicates that the average person is more prone to loss aversion in their decision making.

Having a correct or close to correct model of the average participant in my opinion is important to actually coming up with the right assumptions early on. One mistake I see often in crypto communities is that the initial members of the communities design the project around their own psychology, their own preferences, their own feelings and views, and then later on when it's time to try to take it mainstream become shocked that it's not able to go mainstream. To bring something mainstream in my opinion requires putting our own feelings, views, preferences, to the side at least to some extent, so that the opinions, views and preferences of the unknown quantity of future participants could be catered to on some level.

So for example the earliest participants in the crypto space were risk takers. These people may have enjoyed the opportunities that were offered by the "wild west" like environment of ICOs and easy fundraising. We know for entrepreneurs, for risk takers, for people willing to lose in order to win, having the ability to lose is critical to having the chance to win. What if the majority of the people aren't risk takers like this, don't concern themselves so much about winning big, and are more concerned about just not losing their life savings? This is what the loss aversion research shows is that loss is much more powerful psychologically than winning, on most people. Do we take this into account when trying to model the average participant and when trying to come up with a mechanism design?


Applied algorithmic game theory allows for the use of computer tools to make asking and answering some of these questions easier.  Now this quote below shall illustrate why Tau can make sense for some of this:

"The prototypical problem in mechanism design is to design a system for multiple selfish participants, such that the participants' selfish actions at equilibrium lead to good system performance. Typical objectives studied include revenue maximization and social welfare maximization. Algorithmic mechanism design differs from classical economic mechanism design in several respects. It typically employs the analytic tools of theoretical computer science, such as worst case analysis and approximation ratios, in contrast to classical mechanism design in economics which often makes distributional assumptions about the agents. It also considers computational constraints to be of central importance: mechanisms that cannot be efficiently implemented in polynomial time are not considered to be viable solutions to a mechanism design problem."

That quote from WIkipedia has several key points which make a big difference in my opinion. The first point is that it applies the methods of computer science such as considering computational constraints. Tau for example works on the basis of being able to compute in what is known as p-space, which I translate to mean polynomial time. If this is the case then anything efficeint enough for polynomial time would be at least theoretically doable over Tau and this would mean it may be possible to do distributed algorithmic mechanism design over Tau (in theory).  The other key point in the quote is the reason why we need mechanism design in the first place which is given the assumption that participants are selfish actors, and given that there are incentives to motivate them, how can their selfish actions at equilibrium lead to good performance for the network?

In my opinion Tau will need to solve these problems and the first step in my opinion would be to implement some sort of distributed or decentralized way of studying the problem, analyzing the problem, and doing the applied game theory and or algorithmic mechanism design. If that can be accomplished then during the alpha the mechanism design could be part of the initial discussions. If this is not feasible then it may require more traditional means such as to do things manually, but this in my opinion is going to be very important for the success of Agoras.

References

Lewenberg, Y., Bachrach, Y., Sompolinsky, Y., Zohar, A., & Rosenschein, J. S. (2015, May). Bitcoin mining pools: A cooperative game theoretic analysis. In Proceedings of the 2015 International Conference on Autonomous Agents and Multiagent Systems (pp. 919-927). International Foundation for Autonomous Agents and Multiagent Systems.

Roughgarden, T. (2009). Algorithmic Game Theory.

1. https://en.wikipedia.org/wiki/Algorithmic_mechanism_design

2. https://www.theregister.co.uk/2018/02/05/developers_developers_developers_developers/