Casual anthropic principles

Convenience-sampling lived human experience

An anthropic principle in physics states that we can derive certain observations of the world to be likely based on the fact of our existence in it. The world must be one that can support life or we, as living things, would not be around to see it. This, for example, allows us to deduce that if there were many worlds and we know no good mechanism to choose among them, perhaps all possible worlds do exist but the ones that do not allow us to live are the ones that we never see because those universes are barren of anyone to see.

There is a whole philosophical body of research about the implications of this for Bayesian reasoning and ethics and all that stuff. I avoid the heavy epistemological end of that here, largely.

The more lazy layperson kind of anthropic principle that arises every day is what I'm after here:

Put another way, the sampling process that produces us observers is important in constraining the types of observations we are likely to make. In fact, if we do put it that way, then this idea is nothing but a fairly standard tool in statistics, but it is one that we seem to forget fairly often. In some systems, the fact that we are likely, or even able, to make observations of those systems, says something significant about which (part of the) system we observe. Observations through a keyhole are likely to be of something that was be hidden behind locked doors.

Alternative name for this kind of observational bias problem, also from physics: geocentric models. Before Copernicus, it was frequent and natural to regard the earth as the centre of the universe because certainly al the people who cared about it seemed to be on the earth. This was the geocentric perspective. We have leveraged various evidence since then to understand that possibly out viewpoint might be typical from the perspective of humans, but from the perspective of mass in general, not so much. When we admit that the earth might orbit the sun, that is a heliocentric perspective, and we need to collect the evidence to support it despite the fact that we cannot see things directly from the perspective of the sun.

A popular geocentric cosmology before Copernicus.

On this page is a collection of systems that have this kind of observational bias baked in, for my own interest.

It turns out, of course, that I am not the only person to have become curious about this: See, e.g. Katja Grace’s Anthropic principles page Nick Bostrom’s Anthropic Bias for a published book on this theme. There are many small variations. if you want to get philsophica, e.g. the sleeping beauty problem.

Any argument you find yourself in is an unresolvable one

Anthropic principles meets memetics.

Any statistics you do is at a particular level of uncertainty

The anthropic principle in statistics. Ken Rice said

Re: the anthropic idea, that (I think) statisticians never see really large effects relative to standard errors, because problems involving them are too easy to require a statistician, this sounds a lot like the study of local alternatives, e.g., here and here.

This is old news, of course, but an interesting (and unpublished) variation on it might be of interest to you. When models are only locally wrong from what we assume, it’s possible we can’t ever reliably detect the model mis-specificiation based on the data, but yet that the mis-specification really does matter for inference. Thomas Lumley writes about this here and here.

Any philosophical debate you are in is a semantic quagmire

C’mon, why are we still arguing about free will? Is it because it these words identify something profound or just a deep hole? How about free speech?

You have the perspective of someone who is like you

Because much of our experience of the world is based on how other respond to us, and that is based on features of our appearance that are hard to change, we all live in our own private filter bubble based on the world view. e.g. because I am a confident, white, English-speaking male of a particular generational cohort, I elicit particular responses from the people around me, that people who are not confident, not male, or not white, are less likely to elicit, even in a situation which is otherwise identical. I will not generally experience that alternative experience without role-playing, cross-dressing or blackface, all of which are either taboo or difficult. Vice versa, people who look different to me are less likely to have a good insight into my perspective of the world. The term white privilege, as it is commonly used, presumably describes a particular manifestation of this general phenomenon.

The intriguing phenomenon for me is that, experientially, it seems hard for us to learn to compensate for the difference in world perspective, and we do tend to either erroneously assume our own experience is universal, or at least to do a bad job of understanding the differences between our experiences.

You have the perspective of your peers

TBD. See filter bubbles and/or inference on social graphs.

Hype cycles and Ponzi schemes

Where in the hype cycle do you encounter a product? Which is the most likely level that you will join a Ponzi scheme?

From the original Hobart essay, Sin, Secret, Series A. Every startup needs to know something:

A social media site might turn out to be the reductio ad absurdum of the brand-as-lie/lie-as-Schelling-Point phenomenon, since the entire point of user interaction on the site is to make the lie true. If a site markets itself as the place where a certain kind of cool person hangs out, and says it boldly enough to the right audience, it becomes exactly that.

A corollary to this is that for you, every social media site peaks in utility right after you join. When I was barely cool enough to qualify for Quora, Quora was pretty cool to me — but to anyone who’d been on the site for six months, Quora was a formerly cool site now populated by lamers.

Reverse all advice as necessary

See adverse advice selection.

No comments yet. Why not leave one?

GitHub-flavored Markdown & a sane subset of HTML is supported.