Let's say we want to find an objectively true answer to a question, or at the very least, find a probability distribution for what the likely answer is. There are a few methods we could use to go about achieving this.
1) Intuition
Basically system 1 thinking, which works for a lot of things but isn't always reliable. Probably the most time efficient of all of these methods, but it's also the laziest. Also the most easily distorted by cognitive biases.
2) Opinion Poll
We could take an opinion poll of a lot of people, which has the benefit of a larger sample size but again is often prone to error. For example, if a topic is complex and most people are uninformed, this might just be measuring biases.
3) Poll of Experts
Probably a better approach if we can reliably identify who the experts are, but we could also be measuring selection bias insofar as some people might be more likely to become experts, and then hold opinions they already held. And it's often as difficult to agree on who the experts are and how broad their expertise is as it is to agree on the topic itself.
4) Research
We could research and then use or intuitions based on the new information. But this is the most time-intensive method, and there's the issue of filtering through bad arguments for each side.
5) Debate
We could debate the topic. But this can be pretty time consuming and also depends on each debater being a good representative of the strongest arguments for their viewpoint.
6) Social Platforms
Ideally, this allows for a bunch of people all researching and debating the topic and then figuring out where the majority of people stand. But propaganda, bots, and bias make it far from ideal. As well as most arguments being a few sentences long at most and largely the same ones repeated over and over, which seems inefficient.
So, how can we take the best of these approaches while optimizing them for efficiency?
Phase 1: Argument Mapping
One way to save time might be storing all the arguments in one place so that people don't have to start the argument over from first principles each time. If the chain of argument goes something like point -> counterpoint -> counterpoint, people could be allowed to add arguments at any point along the chain. This could get very big very fast, but if we ask people to rank arguments for the same side against each other, then bias is largely eliminated, since when we ask someone "what is the best counterargument against argument X?" their answer won't really be influenced by whether or not they agree with argument X. So if I want to know the strongest arguments for and against a position, I can just read the strongest argument for, then the strongest counter, then the strongest counter to that, etc.
This system is intentionally adversarial, in that each side is trying to beat the other, yet this does not actually harm the efficacy of the system. If we want to form the strongest argument -> counter chain possible, then it doesn't matter if the CIA is making arguments in favor of Israel and the IRGC is making arguments in favor of Palestine. Insofar as they want to support their chosen side, they will be making the strongest arguments possible for their position, and that's what we want—the strongest arguments for each side. Making bad arguments for the other side would be pointless, since the system is designed such that we read the best arguments, not the bad ones. The only way I could see gaming the system working would be rating bad arguments for the other side high to bury their good arguments, but praising a bad argument for a position you disagree with seems so contrary to human nature that I don't see it being a huge problem.
Phase 2: Acquiring Knowledge from Arguments
So once we have a database of the strongest arguments for each side, what is the best way to derive knowledge from it? Maybe we could have some group read the argument chain and then take an opinion poll, but there's still the question of selection bias. And even if we polled everyone on Earth, is that the optimal sample? If one ideology has more kids than another, might that unfairly skew the poll results?
But there's a better way in my opinion. We sample a bunch of people with preexisting views for either side, ignoring how many of each we get. Then we measure what percent of each group flips their view. Let's say, for example, 2% of people who initially favor the resolution change their mind after reading the arguments, and 1% of people who initially oppose it change their mind.
We then determine what view distribution would be most efficient in terms of information theory—that is, what would the view distribution of the population need to be such that nothing new can be learned from the arguments? In this case, if two thirds of the population opposes the resolution and one third supports it, then the overall poll results won't change after they are exposed to the argument chain—two percent of the smaller group and one percent of the larger group changing their minds will cancel each other out.
To think of this another way, suppose that some small percent of the uninformed population bases their opinion solely on the new information provided by the strongest arguments, and the rest of the uninformed population just keeps the view they had before. Regardless of what that percentage is, presuming it is equivalent regardless of the starting view, we can mathematically estimate the optimal view distribution by sampling each viewpoint group before and after.
Another method might be having people give their percentage certainty in the resolution being true both before and after, then see if they directionally converge toward a specific value. For example, graph people's final certainty vs initial certainty in the resolution being true, and find a line of best fit, then find where the line of best fit intersects y=x. (So for instance, the line of best fit might estimate that if someone is 50% sure the resolution is true beforehand, they are 60% sure afterward. But if they were 80% sure beforehand, then they were 70% sure afterward. But at one point, in this case 65%, we would predict that the certainty remains the same before and after. So we assume an estimate of 65% certainty in the resolution being true is informationally efficient, since on average someone who made that estimate wouldn't change their mind after reading the strongest arguments for each side.)
This phase cannot be adversarial, because it would be easy to skew in favor of one side, but filling out surveys takes less effort and people than making a bunch of arguments, so we could get a sample group of 100 people or so with each viewpoint and verify that they are real people. I think it goes against human nature to voice support for a view one disagrees with, so I think most people will be honest.
Basis for Efficiency
The theory behind this system isn't that it always arrives at an objectively true answer, but that it arrives as close to the objectively correct answer as the population possibly can—that is, there are no additional sources of evidence, arguments, or intuitions that can be exhausted once this system has been used that would aid in reaching on objectively true answer. In practice, people won't submit every possible argument and response, but they might get close enough to be effectively efficient.