This market is a proof of concept on how prediction markets could avoid perverse incentives on questions about x-risk. You can already make a good case that there is a very strong incentive to provide accurate information when that information may help avert the destruction of everything you and everyone else who could ever exist values, but nonetheless I think this is an interesting idea to explore and may be of some utility.

When this question closes in the year 2100, a panel of experts will be asked to estimate a Beta distribution* for the probability that humanity survives from 1st Jan 2024 until 1st Jan 2100. Given that humanity must have survived for this question to resolve, and we don't know whether that will happen, the parameters of the distribution which this set of questions will resolve to is modified. The beta question will resolve to the unmodified beta the experts give (median of experts). The conjugate alpha question will resolve the median value that the experts give for alpha minus one, to reflect the fact that the experts already have additional evidence we don't because they know they survived.

As an analogy, if you are estimating the bias of a coin as a Beta distribution, given one H, you can still estimate the value you would predict without that evidence by modifying the distribution accordingly.

This means you can predict your true estimate of Beta distribution parameters for our survival, and the effect of survivorship bias on the resolution should be removed.

This question is for the alpha parameter of the expert's Beta distribution (minus one).

The question for the corresponding beta parameter is here https://manifold.markets/bleepbloop/if-humanity-survives-this-century-w

* They will be restricted to the parameter ranges in the questions