16 Comments

Majority opinions are like people. Everybody has one. :)

Expand full comment

People are educated mostly along majority consensus lines however, at least in the basic things. For example, math lessons in school will teach not only the majority oare apinion but the global consensus as doctrine, i.e. as self-evident and objective truth. However, as Math gets more and more complex and further levels of complexity are added, expressions written math for the purpose of physics for example, rearrange the entire canon of mathematics previously understood to be self-evident truth. E=MC² for example, alters a localised view 'Newtonian Physics' into an atomic global view, 'Quantum Physics'.

Expand full comment

You're assuming opinions are normally distributed, but shouldn't we expect a multimodal distribution due to the social nature of beliefs? Think about how that graph would look like for the age of Earth in town where half of the people are evangelical christians.

Expand full comment

Hmmm. You're only better off accepting the average opinion if you are not already in the shaded area.

My view is that your model is information-neutral but people are not. A simple instance would be to ask everyone in the world what the Russian word "piva" means in their language. Dude, I'm not taking the majority view on that! So much for facts.

In the world of less factually based opinion, I'm wondering *why* I would take your stance. Were it true that the majority was wiser than the individual, our views would remain the same in perpetuity. But they don't.

Expand full comment

Can I break out the religion reductio again? Please don't convince me that normative epistemology commits me to believing in the existence of a supernatural deity.

Expand full comment

On Bias, the Wisdom of Crowds, and Philosophical Majoritarianism

Wow. It's one thing to claim bias or conflict of interest every single time you disagree with someone else. It's a whole different beast when you recognize that everyone (including yourself) is biased everywhere, all the time. Hal Finney's post is an i...

Expand full comment

Robin, you can't justify a line of reasoning by saying that it's "required by the whole approach". Either it works or it doesn't. If you need to draw an accurate map of a city, walking through the city works, flipping coins doesn't. Even if flipping coins is "required by the whole approach" for approach X, it still won't work, just like engines can't run on water.

Likewise with saying, "I think the majority is biased so I'm going to reject their opinion." Either that's a reasonable justification or not (I sure think it is), but either way a cognitive procedure is not going to work for drawing accurate maps on just one occasion and then fail at all other times - let alone exhibit such behavior because it's "required by the whole approach".

Expand full comment

Eliezer, it seems to me that Hal is trying to go as far as he can in the direction of not presuming that he knows more than others. So the reason to disagree with the majority on majoritarianism, but not on God, is because the first disagreement is required by the whole approach, while the second is not.

Expand full comment

People vary by orders of magnitude in their ability to perceive reality correctly. One can see this by engaging the man in the street on subjects like physics where the correct answers are known to a near certainty. My personal favorite is the man who after viewing television footage of the process, insisted that when a parachute opens, the parachutist for a few seconds is pulled up, i.e., rises with respect to the ground.

People vary greatly in how confident they are in their own perceptions. I hypothesize that the people who are exceptionally good at perceiving reality correctly got that way by scrutinizing the reliability of their own perceptions much more probingly than most people are are willing to. (But beware the "distinguished scientist" effect.)

Unless doing so is an assignment for a class or a personal favor to Robin, the mere fact that one is a regular reader of this blog vastly increases the probability that one subjects his perceptions to probing scrutiny, which by my hypothesis increases the probability that one is exceptionally good at perceiving reality.

If a reader of this blog hesitates to make some important decision, there is no shortage of people who will boldly make that decision in his or her stead. It is an ethical mistake to refrain from acting until one is sure because that concedes the decision to people who are more confident and less competent.

Expand full comment

Hal, your stated justification for ignoring the majority opinion on majoritarianism - "Well, I think most people have a bias in the opposite direction, so I'm going to ignore them" - is enormously flimsier than the justifications for ignoring the majority opinions on such topics as the nonexistence of God, the expected utility of cryonics, or, hell, the truth of evolutionary theory. If you're willing to ignore the majority opinion in the first case, you ought to be willing to ignore the majority opinion in the others.

Expand full comment

Bias is not bias per se. It is not arbitrary. It doesn’t serve merely to mislead us. It is merely the application of cognitive faculties perfectly suited for tasks they are designed to handle, to tasks which they are not. Like how when Wason’s card problem is posed as it is, people fail miserably. But when posed as problem of catching cheaters, which has the same logical form, people instinctively employed their brain’s cheater detection mechanism, which effectively provided the correct solution. The solution, if we want to arrive at some truth not designed to be handled by our ancient cognitive mechanisms, is not to compensate quantitatively for bias (as in to subtract another opposite bias from it?! which seems to me the basis of your entire proposal, including your justification for it), but to be aware of domains of reasoning where we are susceptible to them, and to keep them in check.

The assumption of a bell-shaped distribution to opinions is plausible. However, there is no need for the truth to be anywhere near or far from the peak. Biases aren’t random, they therefore don’t cancel out. Unlike the distribution of opinions, the probability distribution of truth would more likely be uniform across the entire spectrum of opinions (appealing to the principle of indifference, as we really don’t at all know where the truth should be).

Majoritarianism implies the heeding conclusions, instead of evidence. Thus it will inevitably lead you to conclude what everyone else is concluding. And what everyone else is concluding is what everyone else is concluding, ad infinitum. The entire system is self-deterministic and closed. It is designed to be in ignorance of evidence. In deciding truth, it is no better than rolling a dice. Except, as the system will tend toward a stable equilibrium, you only have one shot at the truth. Which makes this system far worse than even gambling.

Expand full comment

1. How useful is the assumption of a bell-shaped distribution in approximating the range of opinions, esp. given that individuals' opinions are normally interdependent. Specifically, the first observation would not necessarily hold if the distribution is exceptionally skewed - perhaps when one particular opinion is oversubscribed due to media coverage or political propaganda.

2. Secondly, if we turn this into a dynamic analysis and if everyone is to follow 'philosophical majoritarianism' in forming their opinion, wouldn't the distribution itself change? How would subsequent observations change? Would there be a convergence of opinions? If there is, would this point of convergence be the truth, or perhaps a stable equilibrium away from the truth, hence a bias?

Expand full comment

I do take issue with Philosophical Majoritarianism. Your justification is based on several assumptions, two of which are particularly bothersome.

Firstly, you assume that opinions follow a normal distribution, which I think is a vast oversimplification.

Secondly, and more importantly, you assume that there is an actual truth. On scientific issues this is certainly the case, but I put it to you that on those issues you'd be better off going with the majority view among scientists.

On political issues, for instance, the truth is more a question of personal values than of anything else, hence the majority view will reflect the majority values which might not be your own.

Expand full comment

Gah, I am an idiot. Ignore my comment.

Expand full comment

This seems to rely upon opinions on the matter in question having a normal distribution, but I don't think that the central limit theorem necessarily works in this case - can we really assume that majority opinion is the sum of "independent and identically distributed random variables" in most cases? For politically divisive questions, I would expect to see more of an "m-shaped" curve.

Expand full comment

Are there criteria for which majority to choose? The majority of every human being that has ever existed? The majority of educated adults living in our own country? The majority of scientific specialists on the subject?

On any given subject, we have to make carefull decisions as to which majority to select. Unless we have good reasons, independent reasons for which majority to go for, this brings us back to our own biases. I think it may be wiser to just try and establish when the majority is wise (and when it's not, and what majority has the best opinion), rather than use it as a default position.

Expand full comment