In short, instead of all governors having voting weights proportional to their committed algos (i.e. the “1 algo = 1 vote” rule), why not randomly select N governors (with replacement), in a way that the probability of getting selected is proportional to the committed algos, but once selected, each governor will have 1 vote (or more if the same governor was selected more than once)? Everything else could be the same as in the current algorand governance plan.
Note that the above proposal uses sampling with replacement, which means that any governor may be selected more than once, and therefore may get more than 1 vote. This ensures there is no incentive for users to split their algos in multiple wallets, nor is there any harm in doing so.
The primary motivation for proposing this specific change, is to make sure that the initial algorand governance model is one that if needed, can always be changed from within (maybe not very quickly, but eventually), and will not get stuck forever in a status quo where the same ultra-wealthy individuals who make up more than 50% of all algo, are also always the ones who have more than 50% of the governance votes, and never allow governance to be changed to anything less plutocratic.
What exactly N should be is certainly open for debate, but to illustrate the claim that this method does not get stuck the way the foundation’s “1 algo = 1 vote” method can get stuck, consider the following hypothetical scenarios, each involving one ulta-wealthy individual owning 5.1B algos.
If N=1, the ultra-wealthy individual could commit all 5.1B algo to governance, but assuming all other algos were also committed to governance (a total of 10B algos), the ultra-wealthy would be selected as a governor only 51% of the time, and would not make 100% of the governance decisions himself like he would in the “1 algo = 1 vote” method. (Of course with N=1 we are in a sense talking about selecting a dictator, and I wouldn’t actually recommend choosing N=1.)
If N=100, the ultra-wealthy individual would get an average of 51 seats in the governance, but due to the random sampling, he would still have only 54% probability of having majority of the seats, and therefore could still not unilaterally force his will to all others for more than about half of the time. (Of course in practice the ultra-wealthy can get his will more often than that even if he gets only 49 seats for example, as all it takes is two other governors to vote along to win the majority.)
If N->Infinity, the ultra-wealthy would get exactly 51% of the total governance voting weights 100% of the time, which is the exact same situation as we saw with “1 algo = 1 vote”.
So, the method proposed above is actually just a generalized model leading to “1 algo = 1 vote” if parameter N=Infinity is chosen.
My argument therefore is, that we should not choose N=Infinity, but choose something different.
P.S. For those whose initial reaction to any probabilistic governance is that it makes no sense to randomize something you don’t have to randomize, I would recommend you learn more about it in Wikipedia under the term sortition: Sortition - Wikipedia Not only does it have many theoretical benefits over the commonly used non-probabilistic voting, election, and governance methods, but also (direct quote from Wikipedia): “In ancient Athenian democracy, sortition was the traditional and primary method for appointing political officials, and its use was regarded as a principal characteristic of democracy.”
Also, those familiar with machine learning may know that one of the most powerful machine learning methods, Random Forest, is based on the principles of “wisdom of the crowds”, and the way to achieve that in machine learning is via random sampling. It is therefore no coincidence, that I consider my proposal to better achieve “wisdom of the crowds” than the original “1 algo = 1 vote”.
Finally, as a practical point I would mention that the selection of governors can be done ahead of time, as a form of representative democracy where the selected governors can then spend extra time educating themselves on the issues that are being voted on, knowing that their vote really matters. Or, if the risk that not all selected governors end up voting at all, causing bias and uncertainty in their decisions, the random selection can also be made known only after the voting has taken place, forcing all governor candidates to vote first (even if not all votes end up being counted).
Disclaimer: I’m not aware of an existing voting system that is exactly as the proposal described above, but it is based on my intuition on how to solve one of the main problems introduced by “1 algo = 1 vote”, while acknowledging certain restrictions that the governance model must have (e.g. no KYC allowed). Therefore it is certainly possible I have made a mistake and the proposal does not work as intended. If so, please point it out.