Meta-Doomsday Argument: Uncertainty About the Validity of the Probabilistic Prediction of the End of the World

Link post

Ab­stract: Four main forms of Dooms­day Ar­gu­ment (DA) ex­ist—Gott’s DA, Carter’s DA, Grace’s DA and Univer­sal DA. All four forms use differ­ent prob­a­bil­is­tic logic to pre­dict that the end of the hu­man civ­i­liza­tion will hap­pen un­ex­pect­edly soon based on our early lo­ca­tion in hu­man his­tory. There are hun­dreds of pub­li­ca­tions about the val­idity of the Dooms­day ar­gu­ment. Most of the at­tempts to dis­prove the Dooms­day Ar­gu­ment have some weak points. As a re­sult, we are un­cer­tain about the val­idity of DA proofs and re­but­tals. In this ar­ti­cle, a meta-DA is in­tro­duced, which uses the idea of log­i­cal un­cer­tainty over the DA’s val­idity es­ti­mated based on a vir­tual pre­dic­tion mar­ket of the opinions of differ­ent sci­en­tists. The re­sult is around 0.4 for the val­idity of some form of DA, and even smaller for “Strong DA”, which pre­dicts the end of the world in the near term. We dis­cuss many ex­am­ples of the val­idity of the DA in real life as an in­stru­ment to prove it “ex­per­i­men­tally”. We also show that DA be­comes strongest if it is based on the idea of the “nat­u­ral refer­ence class” of ob­servers, that is, the ob­servers who know about the DA (i.e. a Self-Refer­enced DA). Such a DA pre­dicts that there is a high prob­a­bil­ity of a global catas­tro­phe with hu­man ex­tinc­tion in the 21st cen­tury, which al­igns with what we already know based on anal­y­sis of differ­ent tech­nolog­i­cal risks.

High­lights:

· There are four main types of DA: fu­ture pop­u­la­tion pre­dic­tion (Gott’s DA), Bayesian up­date of risks (Carter’s DA), the more prob­a­ble Late Filter (Grace’s DA) and the Univer­sal DA.

· Meta-DA treats log­i­cal un­cer­tainty about the pre­dic­tive power of the DA as a prob­a­bil­ity that DA will work.

· We used a vir­tual pre­dic­tion mar­ket of sci­en­tists to as­sess the log­i­cal un­cer­tainty of the DA, which pro­duced es­ti­ma­tion of around 0.4 of its val­idity.

· The strongest, and thus most im­por­tant form of DA is the Self-Refer­enced DA, and for this class “the end” may be as early as mid­dle of the 21th cen­tury, though this is not nec­es­sar­ily hu­man ex­tinc­tion.

· Knowl­edge about the DA could be used to up­date our global risks pre­ven­tion strate­gies by pay­ing more at­ten­tion to uni­ver­sal risks or by us­ing ran­dom pre­ven­tion strate­gies.