@Sylvain Can we try prevent wrong resolutions on numeric questions? I have seen multiple questions being resolved with error of 1/1M times.
I don't know how resolution dialog works, but I guess it should be easy to add a blocking warning if the result is predicted to be less then 10% likely:
"Metaculus community predicts that value this small is less then 0.037% likely. The predicted value is 47M which is 1M times larger then proposed resolution. Please double check it. [DISMISS]"
What happens in case of the schism? What will be resolution if church splits into two with opposite positions? It could be messy to decide which church is the legitimate one.
@casens It's OWID data linked in the question. Life expectancy at age 10. And difference to the previous row in the spreadsheet. Note that none of these countries continued with escape velocity in the next period.
@Linch Are you taking into account the difficulty of obtaining evidence? My hunch is that even if the virus leaked from the lab it will most likely never be convincingly shown or even discovered by investogators.
The resolution conditions are not carefully drafted. Let's say some country has an extreme disaster that kills huge portion of the population in 2050. This would reduce life expectancy for this year by 5 years. Then it would regress back, growing by 5 years in a year. If the following years see any growth, then it would be enough to qualify the period of 2051-2055 as having average growth of longevity >1y/y.
I guess one could lose points here if Metaculus is down on New Year 2030, but comes back later. But I think it might be useful to have disabled scoring on some questions. It can be emulated by saying "This question will always resolve ambiguous at Metaculus, but following resolution criteria will be used at the Judgment Day".
This is actually more of a bug report. @admins links to FAQ to #admins-tag, but there is no element with this ID there. Same applies to ()moderators.
Is American parliament really so slow, that there is 43% chance that uncontested bill will not be passed for a whole year?
Resolution date is so close that I would select narrower interval if the scale allowed.
@akarlin Conditional questions are implemented as questions that resolve ambiguous if certain condition is not fulfilled.
@D0TheMath The resolution cutoff is ten times smaller: 11,100. The point of the question is to anticipate the potential legal change, which is not certain to happen at all.
@Sylvain It was just an idea for a partially unstable question.
@johnnycaffeine Metaculus should have enough data to fit parameters for how much extremization is appropriate for questions asked on the site.
"no preference" qualifies as not "accepting" Everett? Should it be changed to "preferred/favorite/best/closest to truth" interpretation, instead of "accepted" one?
@SimonM Thanks, it might be enough to mitigate the effect. I think there is still bias coming from the fact that the question time-frame varies, but this is probably not very strong.
@SamtheSamurai It seems rather silly to include years when US spanned much smaller area and so could not record many events, not to mention suffer property damage.
@BlueThing Seems like another 6 orders of magnitude error, no biggie.
Can we do something more specific then "news describe as annexed"? For example, "claimed as part of territory and administered by Russian Federation"? So we predict the geopolitical event and not the media framing.
Usually temporary occupation is not considered annexation, but it is possible that some of the sources would use this word.
Do we like this kind of questions? If Metaculus prediction for this question is x, what will be fractional part of x + 0.4?