J. E. Landau

One issue I’ve been considering for a while is the future of governance, and how politics could be shaped by new, transhuman technologies. I think that there are some interesting options there that could improve democracy, while destroying most of its conventional institutions.

Mind uploading has been a regular subject of discussion in this blog: it’s the idea that you are not ontologically unique, and that the process of you can be replicated in a computer. This then gives rise to the concept of “forking”; if you can exist in a computer, there is no reason that you cannot be copied within or between computers and run on multiple machines simultaneously.

Sortition is the practice of randomly selecting members of the population to make decisions. Americans, at least, will likely be familiar with this from the practice of assigning people “jury duty”, where one is randomly assigned membership on the jury for a trial (after a further selection process). This is, however, something that could be applied more broadly. Instead of having people vote on issues or candidates, proposals for sortition have randomly selected members of the public form committees to decide certain issues.

With forks, this process may become far easier and cheaper. If it is cheap to duplicate people, then you could theoretically form sortition panels including every person in your polity, all the time. This would make it possible to perfom total direct democracy, where all issues are decided by everyone voting, because you could have the entire public informed and focused on every issue at once, assuming you have the computational resources available.

Is this a good idea? I’m not sure. Larger groups tend to have more trouble coming to a consensus and discussing issues, and the end result might not be better than traditional methods of representative or direct democracy. Further, these techniques would eat up significant computer time, which might be better used supporting a larger population, or a population where each individual has more resources at their disposal, both of which are often seen as the default by science fiction writers and transhumanists. Is it better to have a population of one billion, where everyone is being polled on everything all the time, or a population of one trillion with the same hardware resources? The latter sounds, perhaps, more advantageous.

But this is assuming that you need to run a full fork for voting. There are good reasons to make that assumption. If voting is a right that all persons are entitled to, then it’s probably reasonable to say that running anything less than a full copy of that person is at least somewhat abridging their rights as a voter. Of course, that raises its own issues: if each person is entitled to a vote, are forks entitled to votes separately from their progenitors? Are forks deserving of a full set of rights, and how are they treated separately or the same as the “originals”? These are thorny questions, and in general I am inclined to say that two instances of one person are, ethically, one person, but as I have discussed earlier this is a complicated problem that has not been given full philosophical treatment and I believe that one can come to a variety of sound conclusions, even from fairly similar assumptions and reasoning.

But, what if you can’t? After all, the fork doing the voting doesn’t really need to be fully-featured. The real purpose of this experiment is to come up with someone that will give the same voting result to you, and if there’s no debate, the goal is just to find your opinion on any given issue.

This raises the possibility that you could have some sort of “stripped down” fork that just votes, or, more plausibly, you could have an external statistical model that seeks to emulate your voter behavior.

This is, in a way, the goal of public opinion polling. A president who wants to assess some hypothetical policy before enacting it could have a poll done, which is based on asking a group their opinions on any given issue, and then using statistical techniques to extrapolate this data to the whole population by studying the demographics of the people in the original sample. This is something that has been done in real life, and it is a heavily politicized and controversial process. Polls don’t always agree, and the statistical techniques used to extrapolate data out can be manipulated or accidentally incorrect.

One proposal has been something called “synthetic polling data”, which has AI chatbots pretend to be members of the population with different views, and then polls them on their “opinions” on various topics. This method is, frankly, crap, and it has been shown to “eliminate a great deal of [opinion] diversity” while experts have raised concerns as to “how the chatbots account for shifts in public opinion”. But these techniques will likely get better.

One way that they could be applied is by training a chatbot on data related to each specific person in a country, and then polling those chatbots. Ubiquitous surveillance and better Large Language Model technology could be used to carry out this procedure, without worries about personhood or the computational overhead of emulating fully-featured forks. Of course, these chatbots will not reflect the entire diversity of thought and experience of this person, and so will only have a limited “voting fidelity”, but people already (begrudgingly) accept measures that dilute their voting power, like gerrymandering, so this might be surprisingly palatable.

Taken to the extreme, this sort of simulation might replace all traditional participation in the political process. I would call this situation “algocratic democracy”, because it’s “algocratic” (in that you are ruled by computer systems) but it is is still “democratic” in the sense that these algorithms are theoretically representative of the population, even on an individual level, and the power theoretically rests with the individual people to make decisions, even if very abstractly.

I, personally, don’t think I’d like this very much, since I value my privacy and like to think that I have a deep personality and my decisions would be difficult to predict. This is probably not actually the case. There is very little about me that could not be predicted with enough data and I’m unlikely to be particularly anomalous. Still, it feels rather slimy, and I certainly wouldn’t vote to take away my voting.

Another interesting prospect for algorithmic democracy is the idea of creating some kind of “national leviathan”, which is an artificial intelligence based either on an idealized “spirit of the nation” (think Uncle Sam or Lady Liberty) or an average of the views of the national population. Presumably you could give this leviathan some role in government, perhaps as a national executive or an advisor to the government.

While this could, in some sense, be nominally democratic, the construction of such a being implies a very strong sense of what the people in charge believe their nation is, and what it ought to be. I believe that it is also a fundamentally conservative idea, as in, the national leviathan would necessarily be a conservative factor in society, it is one that prevents change in the future. After all, the national leviathan is now a being with its own will, that will presumably resist being altered, and it’s based on a snapshot in time of one’s national idea.

To moderate this, perhaps national leviathans could be constructed with each successive government, sort of like electing a new figure, which would return a bit of democracy to the concept. But I think that this sort of conservatism is actually the main reason someone could construct a leviathan. If you are a fascist, for example, this national leviathan might be a very powerful guiding light for your nation into the future, which will survive the death of your leader. Of course, in such a case, it would probably be preferable for the leader to arrange for their own immortality, but I can imagine specific technological and political situations making a leviathan feasible and desirable.

Through either a national leviathan or algocratic democracy, one can imagine a government where all legislative, executive, and judicial functions are handled by computers, for a total algocracy, where all government functions are computerized. This is not a new idea at all, people have been thinking about computer government for a long time.

These cases are interesting because they represent a form of totally-computerized government that, theoretically, isn’t divorced from the public at all.