How to not talk about probability estimates
post by boggler · 2018-02-21T03:19:19.645Z · LW · GW · 4 commentsContents
4 comments
I have recently begun asking for people's credence in interesting beliefs.
I was speaking with two of my peers at dinner yesterday, and our discussion chanced upon the topic of the existence of extraterrestrial life. Curious, I asked them what probability they would assign to the existence of life beyond Earth, and they both replied along the lines of "Definitely 100%. The universe is massive."
Surprised, I brought up the problematic corollary of having a 100% credence. "A belief of 100% probability in extraterrestrial life is equivalent to being unable to design a universe with all of our current knowledge of the world and which lacks extraterrestrial life. Yet, it is clearly possible to design such a universe."
"Wait, why are you designing a universe?"
"I'm not, but the claims are equivalent. If such a universe exists, then you cannot reasonably have a 100% belief in the existence of extraterrestrial life because what if our universe is that very universe you designed to not have extraterrestrial life."
"I still don't understand why you're designing universes."
The conversation unfortunately ended there, and in post-conversation analysis, I've been working out how to better articulate this claim.
I believe part of the problem is the verb "design". They are correct in that one cannot literally "design" a universe, and furthermore "design" may be conflated with "intelligent design". A clearer phrase would be "think of". Can you "think of" a universe such that extraterrestrial life does not exist? This is a concrete question. One could think up of such a universe, unlike the ability to design one.
Yet, even with this word replacement, how does one clearly articulate that "thinking of" an alternate, plausible universe implies it is unreasonable to hold a 100% credence in the existence of extraterrestrial life?
My natural impulse is to justify it by saying, "Well, what about the tiny chance that your hypothetical conception is the correct universe? That means that you can't hold a 100% belief because there's that small probability that we are currently in your life-only-on-Earth universe."
Unfortunately, I believe that this may not be effective. I have a sense that our disagreement lies in our different models of what it means to have a 100% credence.
Perhaps, in their model a 100% credence means a very high likelihood, while to me it is absurd epistemological arrogance.
Maybe it would be best to ignore the notion of probability all together, and rather ask the related question: Can you think of a universe in which Earth life is the only life?
After a “yes,” I can very reasonably model their credence as not being 100% without their explicit signaling.
I could follow-up by asking what evidence supports this thought-up universe. In the context of their evidence for-and-against, I predict it might be possible I can reach a more accurate model of their credence than if I had explicitly asked them for a probabilistic estimate.
This intuitively feels epistemologically arrogant, but it succeeds in solving the probability language discrepancy.
Thus, my hypothesis:
When speaking to non-rationalists or the non-mathematically-inclined, indirectly asking about credence through thought-experiments and evidence may be more conducive than directly asking about probabilistic estimates.
Suggestions on how to improve discussing probability with lay people, thoughts, counterarguments, and more welcomed below!
Thank you for reading my first post 😊
P.S. After asking for feedback on the post, a rationalist friend inquired as to why I was interested in the beliefs of my peers with regards to extraterrestrial life. This is a very valid question, and I had asked myself the same as I wrote this.
The answer is that my primary goal is to not discuss beliefs, but rather to understand ways to bridge linguistic and model-based gaps between myself and others, so that I may become a better communicator of my beliefs.
4 comments
Comments sorted by top scores.
comment by Qiaochu_Yuan · 2018-02-21T05:52:33.558Z · LW(p) · GW(p)
Arguing 0 and 1 are not probabilities with most people is a waste of time, and eliciting probabilities is also mostly a waste of time. Step 1 is to teach the fundamental move of considering the other possibility at all; asking whether they can mentally inhabit the world in which life only exists on Earth.
comment by Allen Kim · 2018-02-21T08:30:34.084Z · LW(p) · GW(p)
I think usually the best way is to ask them what kind of betting odds they would take. Although in this case the difficulty in verifying might be a problem. But maybe if you argued that a 100% certainty means they should be willing to lay infinite odds that might get them thinking.
comment by zlrth · 2018-02-21T03:56:03.465Z · LW(p) · GW(p)
This intuitively feels epistemologically arrogant, but it succeeds in solving the probability language discrepancy.
In general I support the thought that you avoid a lot of pitfalls if you're really precise and really upfront about what kinds of evidence you'll accept and not. I suspect that that kind of planning is not discussed enough in rationalist-circles, so I appreciate this post! You're upfront about the fact that you'll accept a non-explicit signal. I see nothing wrong with that, given that you're many inferential steps from a shared understanding of probability.