Opinions on Opinions

Opinions are quite a scam. They can, in almost all cases, be broken down into one or more of a few functionally distinct categories. Either they are:

1. A difference in preference that arises out of individually unique phenomena, e.g. ‘It’s my opinion that bananas are a great fruit.’ This is not asserting the claim that ‘bananas are a great fruit,’ but rather that ‘my set of taste buds react very well to banana.’
2. A difference in world model that arises out of asymmetric evidence, e.g. a battle-hardened veteran may claim that ‘it is human nature to be violent,’ after witnessing many such cases, while a peace-loving hippie may claim that ‘it is human nature to be kind to your fellow man.’ In both cases, the same failure mode has occurred, with each hypothesis having received overwhelming evidence within a limited reference class that its probability has been effectively rounded up to one, thereby trapping it. If the scope of each claim was properly asserted, i.e. ‘When in battle, humans are often violent,’ and ‘When in social situations, humans are often kind,’ then the world model most plausible under all evidence emerges: ‘Human behavior is often context-dependent.’
3. A difference in optimization targets, most commonly in the case of valuing one’s self highest, and steeply discounting with distance. The suburban homeowner arguing against the development of high-density housing units in their neighborhood likely has a similar model of the market dynamics and net utility of such a project as the looking-to-buy yuppie, yet they state different beliefs. Often some degree of self-deception is involved in the process of making arguments which do not hinge on self-interest, resulting in a clouded world model. However, such beliefs rarely get entangled with one’s foundational world model, and can be effortlessly flipped should it not longer benefit their interests. I doubt that many NIMBYs would retain their stance after losing their house in a divorce and being put in the position of purchasing a new one.

There is a notion of opinions as belonging to some higher ontology, to which all people are entitled to possess, and which in some sense cannot be falsified. Yet the categories above do not seem to rely on any unifying ontology, and they certainly do not adhere to the above properties ascribed to ‘opinions.’

The first category is overwhelmingly common (as evidenced by the quantity of people that consider the validity of pineapple on pizza an open question), yet often harmless on its own due to how quickly it bottoms out - the variation of phenomena is limited and so few foreign policy decisions hinge on the taste of fruit. A more serious failure mode emerges when it bleeds into the other categories. Taste in particular is easily identifiable as a non-universal phenomena, and so few wars are fought over it, but other members of aesthetics are often conflated with some higher sense of the ‘Good,’ with disastrous consequences.

A clear-cut example of this exists in the divergence between what I’ll call the ‘green’ movement and the ‘climate’ movement. In many cases, both movements advocate identical actions: shutting down coal mines, stopping the construction of natural gas pipelines, and accelerating the move to electric cars. However, they clearly differ in their attitudes towards nuclear and dense cities, as fog over the Manhattan skyline or steam coming out of a reactors’ cooling tower pattern matches too closely to smog filled industrial-era London or a coal-burning plant — aesthetics that many have been strongly conditioned against.

An antiparallel tension arises out of postmodernism, in which art, synonymous with aesthetics, explicitly rejects aestheticism. If an artist tapes a banana to a wall, a viewer may complain that it does not satisfy their expectations of a visual phenomena, and thus fails within their definition of art. To the artist this is not an issue, as they have replaced aesthetics with some other optimization target, most likely signaling. It no longer has to look good just because it’s made of paint on a canvas. Just as the aesthetics of eco-friendliness bled into and mutated the rankings of carbon emissions, the optimizations of signaling bled into and mutated the rankings of aesthetic phenomena.

Politics is the clearest example of individuals’ world models and optimization targets clashing with one another’s, but strangely enough it provides the strongest use case for the notion of the ‘Opinion’ as a sacred category. When an individual casts a vote for their desired candidate, they do so because they believe it best optimizes their interests, given their limited world model. Not only are their interests being represented, but so is their world model, and in turn their evidence for it. And so when many individuals cast their votes, their world models are aggregated along with their interests.

If everyone has the same world model but different interests, then the aggregation of their votes will produce an outcome that roughly optimizes for their total net utility. In the same fashion, if everyone has the same interests but different models of the world, then the aggregation of their votes will once again produce an outcome that roughly optimizes for their total net utility, not because it finds some compromise between their interests, but because the aggregate world model implied by the aggregated votes will likely be more accurate, and so will better fulfill their individual interests.

The role that the sacred ontology of ‘Opinions’ plays in this aggregation process is ensuring that the aggregation occurs at the polls, by roping off world models to not be influenced by one another. If one believes that their opinion has some special significance and should not be changed, this significantly reduces the influence of the vocal and persuasive. This does not seem like a bad system given the assumptions that:

1. The optimal policies can be found through a myopic process which merely requires a robust world model.
2. The intelligence distribution of the populace is generally narrow.

I suspect that these requirements are rarely met for any poll of sufficient size. Having a generally robust world model is quite often not sufficient for the optimal solution to be reached, as such solutions typically require multiple inferential steps, during which the computational complexity explodes and a brute force search is not sufficient. No amount of aggregating the evidence from individuals about the state of the world in 1939 would have provided the insight to begin preparing for the nuclear era.

This does of course break down once the total voter base reaches the point at which the expected value of an individual casting their vote a certain way is less than the value they get from signaling allegiance to their desired in-group. At such a point the notion of an ‘Opinion’ only serves as an excuse to vote with the in-group without bothering to come up with any reasoning, should they need to defend themself to the out-group. This is also assuming that they are sufficiently WEIRD to take issue with lying to those around them, as otherwise the tipping point occurs when the EV of their vote dips below the value of their time spent going to the polls, as the party they signal allegiance to need not match their vote.

I suspect though that this role as an information aggregator was not the true source of the optimization pressure that allowed the meme of ‘Opinions’ to spread. On Occam’s razor, it seems far more likely that it arose simply from the strong signaling power of a defense for poor reasoning, which was patched together as justification for self-interested actions.