So far we’ve covered four of the major so-called V factions that science fiction writer Radivis has come up: The Exuberance, the Freedom, the Unity, and the Exaltation. There’s one major V faction missing, and that’s the Balance, right?
Sockrad Radivis Interview XI: Balance, synthetic value, Balance factors, frowney and spiral simulations, odysseys, variblissies
Correct. Previously we’ve discussed the four major V factions that are based on a simple “pure” concept, like happiness, freedom, or knowledge. The Exaltation may be special in the regard that their concept of “recellence” doesn’t seem to be simple, even if it may be based on the idea of the perfection of the self. But then the same argument can be made against happiness, which is really challenging to formalize in any simple way. But anyway, all of these are irreducible values in the sense that you cannot separate aspects out of them and still have a functioning value. The Balance has a different approach. Instead of trying to reduce value to a single concept, it takes many concepts and tries to produce some kind of synthesis of those. And one of the easiest ways to combine different values is to add them up. So, the Balance takes the base values of the other V factions and adds them up, with weighting factors attached to each value, in order to create a synthetic value that it tries to maximize then.
How does that work? Would a Balancist compare two options and for example compute that option A generates 10 units of happiness, 5 units of freedom, 15 units of knowledge and 5 units of recellence to get a total value of 35 units, while option B may generate -10 units of happiness, 10 units of freedom, 30 units of knowledge, and 10 units of recellence, for a total value of 40 units, resulting in deciding for option B?
Basically, yes. The reason for that is that if the want to consider all the four basic values of the other V factions, then we would have a 4-dimensional value space for any evaluation we make. We can’t say whether any point in that space is better than any other point, because there is no obvious way to compare them (except for the rare situation that one point is greater than another in all dimensions). To make a clear comparison, we need to project that 4-dimensional value space onto a single linear dimension. Adding up those values is the simplest way of doing that. We could also do other operations like multiplying them, or taking the root of the sum of squares of each component value, but those operations are less natural to the problem we are faced here, so a simple addition is preferable.
That sounds like a really cheap solution for a V faction that’s supposedly created by artificial superintelligences.
Right, the missing elegance of that approach is one of the most common criticism of the Balance. However, there is some beauty in simplicity. Unfortunately, this approach is not as simple as it appears at first, because to make the four vastly different values appropriately comparable, they need to be transformed to a similar order of magnitude. This is where the weighting factors come into play, which do exactly that. But this creates the problem of coming up with 4 (or at least 3) weighting factors for the different values in a way that is “balanced”. There was the hope in the Balance that there is a set of “obviously correct” weighting factors, which would then turn the synthetic value with these factors in the “obviously correct” core value. That hope never materialized, so the Balance had to live with internal disagreements of these weighting factors, which also turned the Balance into the V faction with the least internal philosophical cohesion. There wasn’t one underlying value system, but numerous variants, each with their own weighting factors.
How could the Balance succeed with these internal problems?
With a really simple operation: Taking the weighted average of all relevant weighting factors prevalent within the Balance. These weighted average factors then become the “Balance factors ”, the weighting factors officially valid for the Balance. Of course, this introduces the problem that these Balance factors are not stable, but fluctuate in accordance with the popularity of certain Balance sub-factions with their own weighting factors, but at least that approach represents a workable compromise that enables the Balance to operate as a more or less cohesive unit. To make these factors more stable and prevent manipulation, actually the median of the weighting factors is taken.
But isn’t that a deference to populism in what should be a purely philosophical V faction? How is such an approach justified?
Oh, it was really just a political compromise. The Balance could have ended up as loose alliance of Balance sub-factions, but the situation of the Black War encouraged the different sub-factions to work together in order to make victory more likely. There was also the hope that until the end of the Black War a better solution would be found. After the Black War the Balance tried to maintain the balance between the V factions. Only a strongly cohesive Balance could fulfill that role, which is why that initial compromise was continued.
Why had the Balance succeeded at maintaining its role as one of the five major V factions, if its philosophical foundation seems to be so weak?
The Balance fulfilled the important role of providing a place for those who considered the monomaniacal focus on a single value as flawed. For them, it’s obviously a bad idea to exclude all the other rather reasonable values from your considerations. Even though the philosophical basis for the Balance was weak, it was considered to be less problematic than trying to reduce everything to one single value.
Fair point! The other V factions do appear quite reductionist. But weren’t there other approaches to find a fair compromise between the values of the different V factions?
Yes, there were a lot of alternative sub-factions of the Balance which tried more complex combinations of the other values than simply adding them up. During the Black War, these approaches were seen to too complex to be of practical use, and were more or less considered to be distractions from the war effort. After the war, the situation changed and these alternative sub-factions became more popular. Eventually one of them turned out to be quite special, and was later known as the Coherence .
How can philosophies be too complex for superintelligences?
Making far reaching decisions within the framework of consequentialism is inherently resource intensive, because for each different option the long term consequences need to be computed and evaluated. Due to the problem of combinatorial explosion , the computational effort to make predictions rises exponentially or even faster with the increase of the time frame one wants to consider. And if you combine that with values which are inherently hard to compute, like the values that the V factions are based on, the computation effort to do consequentialism right quickly becomes a challenge even for superintelligences. Adding even more complexity on top of that appeared like a bad idea under the threat of extinction by the Seraphim System. After all, it was obvious to all V factions that allowing the victory of the Seraphim System was a really bad idea.
So the end of the Black War allowed for more philosophical depth, which in the end turned out to be worth it. Or is that a misconception?
The case wasn’t so clear before the proof of the UVS theorem. After all, the Coherence had the rank of any other V faction, which could have just as well turned out to be provably correct with its value system. It’s not so much the proof of the correctness of the Coherence that made the big difference, but the proof of the deficiency of the other V factions. That actually caused their collapse, or rather their transformation into the post V factions.
In what way were the other V factions deficient?
For the most V factions the core deficiency was their incompleteness that arises from focusing on a certain value, which then leaves out other essential dimensions of value. For the Balance the case is more complex, since it has always valued many different dimensions of value. After all, it could be argued that the system of the Balance with perfect weighting factors might be the best possible value system, or equivalent to it. Showing that this is actually not the case was one of the hardest parts of the proof of the UVS theorem.