Public Opinion about Existential Risk

Sum­mary: An MTurk study of peo­ple in the United States (N=395) found me­dian es­ti­mates of 1%, 5%, and 20% for the chance of hu­man ex­tinc­tion in 50, 100, and 500 years, re­spec­tively. Peo­ple were fairly con­fi­dent in their an­swers and tended to think the gov­ern­ment should pri­ori­tize pre­vent­ing hu­man ex­tinc­tion more than it cur­rently does.

Table of Con­tents

Background

The aims of the study

Methods

Es­ti­mated risk of extinction

Confidence

How much the gov­ern­ment does and should pri­ori­tize pre­vent­ing hu­man extinction

Par­ti­ci­pant comments

Discussion

References

Background

Peo­ple in the EA com­mu­nity are very con­cerned about ex­is­ten­tial risk, but what is the per­cep­tion among the gen­eral pub­lic? An­swer­ing this ques­tion is highly im­por­tant if you are try­ing to re­duce ex­is­ten­tial risk. Un­der­stand­ing peo­ple’s thoughts on ex­is­ten­tial risk may help ex­plain why the is­sue has been ne­glected and sug­gest ways to change this. It is cru­cial for move­ment build­ing and con­sid­er­ing how to ap­proach the is­sue with key de­ci­sion mak­ers.

Some stud­ies have been con­ducted to ad­dress the ques­tion of what peo­ple think about ex­is­ten­tial risk, but the re­sults have been in­con­clu­sive. In 2017, Spencer Green­berg con­ducted sev­eral stud­ies us­ing Ama­zon Me­chan­i­cal Turk and ob­tained me­dian es­ti­mates for the like­li­hood of hu­man ex­tinc­tion in the next 50 years rang­ing from .00001% to 1%, with means rang­ing from 11% to 18%. The per­centage of peo­ple who es­ti­mated a 50% or greater like­li­hood of ex­tinc­tion within 50 years also ranged from 11% to 18%. Ran­dle and Eck­er­sly (2015) sur­veyed 2073 peo­ple across the US, UK, Canada, and Aus­tralia in 2013 and found that 24% rated the like­li­hood of hu­man ex­tinc­tion in the next 100 years as greater than or equal to 50%.

Get­ting re­li­able an­swers for this type of ques­tion is no­to­ri­ously difficult. The word­ing of the ques­tions has differed across stud­ies, po­ten­tially af­fect­ing the re­sult. How­ever, the re­sults may vary even when the word­ing is held con­stant: Green­berg’s me­dian es­ti­mates of 1% and .00001% were both ob­tained from sam­ples of over 200 peo­ple on Me­chan­i­cal Turk, and the word­ing of the ques­tion was the same in each case. You can view a com­par­i­son of the word­ing and re­sults across his stud­ies here. The way ques­tions are pre­sented adds an­other level of vari­abil­ity to the an­swers. Green­berg pro­vided par­ti­ci­pants with a list of op­tions, and six of the eigh­teen op­tions were less than 1%. In the study by Ran­dle and Eck­er­sly (2015), on the other hand, 1% was the low­est op­tion available. What’s more, they pro­vided ver­bal de­scrip­tions with the odds, in­clud­ing “No chance, al­most no chance” for 1 in 100 and “Slight pos­si­bil­ity” for 2 in 10 (!). In the cur­rent study, par­ti­ci­pants were asked to type in a per­centage, rather than be­ing given op­tions to choose from.

I hy­poth­e­sized that the vari­abil­ity in re­sults might be due to peo­ple not hav­ing strong opinions on the sub­ject and be­ing un­cer­tain about their an­swers, so I asked peo­ple to in­di­cate how con­fi­dent they were in each of their an­swers. I also wanted to test whether phras­ing the ques­tion in terms of go­ing ex­tinct (as Green­berg did) or be­ing wiped out (as Ran­dle and Eck­er­sley did) would af­fect the re­sponses. Fi­nally, I wanted to start ex­plor­ing how much of a pri­or­ity peo­ple think pre­vent­ing hu­man ex­tinc­tion should be for gov­ern­ment and whether that an­swer is cor­re­lated with their es­ti­mate of ex­tinc­tion risk.

The aims of the study were to:

  1. Repli­cate the re­sults of pre­vi­ous stud­ies on pub­lic per­cep­tion of how likely it is that hu­man­ity will go ex­tinct in the near fu­ture and broaden the range of time scales asked about

  2. Be­gin to ex­plore how con­fi­dent peo­ple are of these beliefs

  3. Check whether word­ing the ques­tion as “go ex­tinct” ver­sus “be wiped out” af­fects peo­ple’s es­ti­mates

  4. Gauge pub­lic opinion on the gov­ern­ment’s role in pre­vent­ing hu­man extinction

Meth­ods

The sur­vey was ad­ministered via Positly, draw­ing from Me­chan­i­cal Turk work­ers in the USA. Me­chan­i­cal Turk was cho­sen for the ease and speed of get­ting re­sponses, and be­cause pub­lished work in psy­chol­ogy sug­gests that re­sults ob­tained through Me­chan­i­cal Turk are gen­er­ally re­li­able (Pao­lacci et al. 2010, Buhrmester et al. 2011, Ma­son & Suri 2012). The sur­vey was live on Positly on July 24, 2018, with ti­tle “Risk to hu­man­ity” and de­scrip­tion “This short ques­tion­naire is part of a study I am con­duct­ing as an in­tern at the Cen­tre for Effec­tive Altru­ism. Par­ti­ci­pa­tion is vol­un­tary, and you can quit at any time. The anonymized re­sults will be pub­lished on­line.”

After ex­clud­ing data from peo­ple who en­tered a num­ber greater than 100 for any of the ques­tions ask­ing for a per­cent, there were 395 par­ti­ci­pants (247 F, 158 M), whose data is an­a­lyzed be­low. Par­ti­ci­pants ranged in age from 19 to 79, with a me­dian of 36 and mean of 39. 55% of par­ti­ci­pants had com­pleted a bach­e­lor’s de­gree or higher, and the me­dian self-re­ported in­come was $40,000 to $59,999.

The sur­vey was cre­ated us­ing Guided Track. Par­ti­ci­pants were asked about the like­li­hood of hu­mans go­ing ex­tinct in 50, 100, and 500 years (pre­sented in a ran­dom or­der). After each of these ques­tions they were asked how con­fi­dent they were of their es­ti­mate. Then they were asked (in a ran­dom or­der) how much the gov­ern­ment pri­ori­tizes pre­vent­ing hu­man ex­tinc­tion and how much they should pri­ori­tize it. More de­tails about each ques­tion and its re­sults are pre­sented be­low.

Es­ti­mated risk of extinction

Par­ti­ci­pants en­tered their nu­meric an­swer into a text box em­bed­ded in a sen­tence. There was no set range, and if a par­ti­ci­pant en­tered a num­ber greater than 100 for any of the ques­tions, their re­sponses to all of the ques­tions were ex­cluded from the anal­y­sis.

Ex­am­ple ques­tion:

How likely is it that hu­mans will be wiped out in the next 50 years?

There is a ___ per­cent chance that hu­mans will be wiped out in the next 50 years.

The es­ti­mated risk of ex­tinc­tion in­creased as the time scale in­creased, as can be seen in the table and box plot be­low. The var­i­ance of the re­sponses was also much greater for the 500-year es­ti­mate. Re­sponses were not sig­nifi­cantly differ­ent be­tween the groups re­ceiv­ing the “go ex­tinct” and “be wiped out” ver­sions of the ques­tions, so I an­a­lyze the data from those groups to­gether.

Es­ti­mated risk of hu­man extinction

50 years

100 years

500 years

Median

1%

5%

20%

Mean

8%

16%

29%

Mode

0 (134 out of 395)

0 (79 out of 395)

10 (48 out of 395)

Stan­dard deviation

16

23

30

Pro­por­tion who es­ti­mated 50%

6%

13%

30%

In the box-and-whisker plots above, the thick lines rep­re­sent the me­di­ans, and the boxes show the mid­dle 50% of the data. The whiskers are drawn at a dis­tance of 1.5 times the in­terquar­tile range away from the me­dian or at the min­i­mum and max­i­mum, whichever is less ex­treme, and any more ex­treme val­ues are plot­ted as cir­cles.

Con­fi­dence

After each of the ques­tions about the like­li­hood of ex­tinc­tion, par­ti­ci­pants were asked to rate their con­fi­dence us­ing a slider with val­ues from 0 to 100. The 0 end of the scale was la­beled “I have no idea” and the 100 end was la­beled “I am to­tally sure”, but no ex­plicit in­struc­tions were pro­vided about how to in­ter­pret the scale. The ques­tion pre­sented the es­ti­mate they had just given and used “go ex­tinct”/​”be wiped out” as be­fore.

Ex­am­ple ques­tion:

How con­fi­dent are you that there is a 1% chance that hu­mans will go ex­tinct in the next 100 years?

Con­fi­dence in es­ti­mate of ex­tinc­tion risk (max 100, min 0)

50 years

100 years

500 years

Median

80

65

57

Mean

67

58

54

Mode

100

100

100

How much the gov­ern­ment does and should pri­ori­tize pre­vent­ing hu­man extinction

Par­ti­ci­pants were asked “How much does the gov­ern­ment pri­ori­tize pre­vent­ing hu­man ex­tinc­tion?” and “How much should the gov­ern­ment pri­ori­tize pre­vent­ing hu­man ex­tinc­tion?” in a ran­dom­ized or­der. An­swers were pro­vided us­ing a slider that went from 0 to 10, with pre­ci­sion of .1. The 0 of the slider was la­beled “The gov­ern­ment does not de­vote any re­sources to pre­vent­ing hu­man ex­tinc­tion” and the 10 end was la­beled “Prevent­ing hu­man ex­tinc­tion is the gov­ern­ment’s num­ber one pri­or­ity” (or “should not” in­stead of “does not” and “should be” in­stead of “is”). No ex­plicit in­struc­tions were pro­vided about how to in­ter­pret the scale. This rather vague for­mat was cho­sen to ab­stract away from par­tic­u­lar is­sues (e.g. whether it be pri­ori­tized more or less than ed­u­ca­tion). The par­tic­u­lar score on the ten-point scale does not mean much by it­self and can­not tell us about sup­port or op­po­si­tion for par­tic­u­lar poli­cies or bud­get de­ci­sions, but the idea was to get at peo­ple’s per­cep­tion of how much the gov­ern­ment pri­ori­tizes the is­sue ver­sus how much it should.

Does

Should

Median

2.5

6.2

Mean

3.1

5.8

Mode

0 (52 out of 395, but note that 9 said 10)

10 (36 out of 395, but note that 20 said 0)

There were weak cor­re­la­tions be­tween par­ti­ci­pants’ es­ti­mates of the like­li­hood of ex­tinc­tion over each time frame and how much of a pri­or­ity they said pre­vent­ing hu­man ex­tinc­tion should be for the gov­ern­ment (r = .19, .18, and .18 for 50, 100, and 500 years re­spec­tively; p < .001 for all three).

Par­ti­ci­pant Comments

Positly asks par­ti­ci­pants for feed­back af­ter they com­plete the sur­vey. In ad­di­tion to choos­ing a generic cat­e­gory such as “in­ter­est­ing and en­joy­able” or “an­noy­ing or ag­gra­vat­ing” (only 7 picked the lat­ter op­tion), par­ti­ci­pants can en­ter free text re­sponses. Ap­prox­i­mately 94 peo­ple said that the sur­vey was thought-pro­vok­ing (or some­thing similar), 70 com­mented that it was differ­ent from other sur­veys/​MTurk work, and 49 said that they rarely or never think about the is­sue of hu­man ex­tinc­tion.

Discussion

Re­turn­ing to the aims of the study:

  1. Repli­cate the re­sults of pre­vi­ous stud­ies on pub­lic per­cep­tion of how likely it is that hu­man­ity will go ex­tinct in the near fu­ture and broaden the range of time scales asked about

The me­dian es­ti­mate of the risk of ex­tinc­tion within 50 years (1%) was within the range Green­berg found in his Me­chan­i­cal Turk stud­ies (.0001% to 1%). It is worth point­ing out the differ­ence in for­mat once again: Green­berg had par­ti­ci­pants choose from a list of op­tions, whereas I asked them to type in a num­ber. Few peo­ple gave an­swers be­tween zero one per­cent (5%, 4%, and 3% of peo­ple did so for the 50, 100, and 500 year ques­tions, re­spec­tively). Many more peo­ple said there was a 0% chance of ex­tinc­tion over the speci­fied time range (34%, 20%, and 7% for the 50, 100, and 500 year ques­tions, re­spec­tively). The re­sponses of 0% might rep­re­sent a be­lief that there is no way we will go ex­tinct in the speci­fied time range, but it is likely that many of them are be­cause of peo­ple round­ing their an­swer to the near­est whole per­cent--1% was also a pop­u­lar choice (19%, 12%, and 9% of re­spon­dents for the 50, 100, and 500 year ques­tions, re­spec­tively).

The pro­por­tion of peo­ple who put the chance of ex­tinc­tion in 50 or 100 years at greater than or equal to 50% was smaller than in Green­berg’s or Ran­dle and Eck­er­sley’s stud­ies (6% ver­sus 11 to 18% for 50 years, and 13% ver­sus 24% for 500 years). Nonethe­less, it is worth not­ing the wide spread of the re­sponses and the tail of peo­ple as­sign­ing large prob­a­bil­ities to the hu­man ex­tinc­tion in the near fu­ture.

  1. Be­gin to ex­plore how con­fi­dent peo­ple are of these beliefs

Peo­ple were more con­fi­dent about their 50 year es­ti­mates than their 100 or 500 year es­ti­mates, as we would ex­pect.

I also ex­pected that peo­ple’s con­fi­dence in their an­swers would be quite low, since I had con­ducted a few in­ter­views in which I asked about the like­li­hood of hu­man ex­tinc­tion in the next 100 years, and the in­ter­vie­wees typ­i­cally ex­pressed a great deal of un­cer­tainty about their es­ti­mates. Con­trary to my ex­pec­ta­tions, how­ever, the sur­vey re­spon­dents did not gen­er­ally re­port great un­cer­tainty about their an­swers, and many peo­ple re­ported com­plete con­fi­dence in their pre­dic­tions. This is some ev­i­dence against my ex­pla­na­tion that the vari­abil­ity in sur­vey re­sults might be due to peo­ple hav­ing no idea of the an­swer and just pick­ing some­thing out of the air. How­ever, the large num­ber of peo­ple who, with­out prompt­ing, wrote in the com­ment sec­tion that they had rarely or never thought about the topic sug­gests that we should not dis­count that ex­pla­na­tion too strongly.

Method­olog­i­cal is­sues may ac­count for the find­ing—a slider with such pre­ci­sion may not have been an in­tu­itive or in­for­ma­tive way to ask about con­fi­dence/​cer­tainty, etc. More so­phis­ti­cated ways of ask­ing about con­fi­dence and/​or test­ing how the be­liefs change when pre­sented with new ev­i­dence, could pro­duce more con­clu­sive ev­i­dence on the strength of these be­liefs.

  1. Check whether word­ing the ques­tion as “go ex­tinct” ver­sus “be wiped out” af­fects peo­ple’s es­ti­mates

No, this word­ing did not make a differ­ence.

  1. Gauge pub­lic opinion on the gov­ern­ment’s role in pre­vent­ing hu­man extinction

On the whole, peo­ple ex­pressed that the gov­ern­ment should pri­ori­tize pre­vent­ing hu­man ex­tinc­tion more than it cur­rently does. Par­ti­ci­pants’ es­ti­mates of the like­li­hood of ex­tinc­tion were cor­re­lated with how much of a pri­or­ity they think it should be for the gov­ern­ment, but only weakly.

More re­search would be needed to de­ter­mine what poli­cies peo­ple would sup­port and how they would rate pre­vent­ing hu­man ex­tinc­tion rel­a­tive to other pri­ori­ties. Fu­ture re­search could also at­tempt to dis­en­tan­gle the is­sues of bud­get and pri­or­ity, which were con­founded in this study be­cause the lower end of the scale men­tioned “re­sources” and the up­per end said “pri­ori­tize”. Peo­ple might think that it should be a strate­gic pri­or­ity but should not be given a large share of the bud­get; this came up in the in­ter­views I men­tioned above.

It is worth bring­ing up the com­ments sec­tion again: many peo­ple said the topic of the sur­vey was thought-pro­vok­ing, un­usual, and/​or some­thing they don’t tend to think about. Com­bin­ing these com­ments with the find­ing that peo­ple think the gov­ern­ment should be do­ing more to pre­vent hu­man ex­tinc­tion sug­gests that the is­sue might be ne­glected ont be­cause peo­ple think it’s unim­por­tant but be­cause they just don’t tend to think about it at all, or at least not se­ri­ously.

References

Buhrmester M, Kwang T, Gosling SD. (2011). Ama­zon’s Me­chan­i­cal Turk: A new source of

in­ex­pen­sive, yet high-qual­ity, data? Per­spec­tives on Psy­cholog­i­cal Science. 6(1),

3‐5.

Green­berg, Spencer. (2017). So­cial Science as Lens on Effec­tive Char­ity: re­sults from four

new stud­ies—Spencer Green­berg. https://​​www.youtube.com/​​watch?v=tOSpj19eows

Ma­son W, Suri S. (2012). Con­duct­ing Be­hav­ioral Re­search on Ama­zon’s Me­chan­i­cal Turk.

Be­hav­ior Re­search Meth­ods. 44(1), 1‐23.

Pao­lacci, G., Chan­dler, J., & Ipeiro­tis, P. G. (2010). Run­ning ex­per­i­ments on Amazon

Me­chan­i­cal Turk. Judg­ment and De­ci­sion Mak­ing, 5, 411–419.

Ran­dle, M., & Eck­er­sley, R. (2015). Public per­cep­tions of fu­ture threats to hu­man­ity and

so­cietal re­sponses: A cross-na­tional study. Fu­tures, 72, 4–16.

http://​​dx.doi.org/​​10.1016/​​j.fu­tures.2015.06.004

Notes:

1. I did this pro­ject as part of a re­search in­tern­ship at the Cen­tre for Effec­tive Altru­ism.

2. I will be with­out in­ter­net for sev­eral weeks af­ter to­mor­row and will re­spond to com­ments when I re­turn.