An update on the Global Priorities Project

The pur­pose of this post is to let peo­ple know a bit more about what we’ve been do­ing and what we hope to do. Please let us know what ques­tions you have that we don’t an­swer be­low!

Activities

The Global Pri­ori­ties Pro­ject (GPP) was cre­ated in or­der to pur­sue a wide range of low-hang­ing fruit. This has meant that our work has cov­ered a fairly broad range across two ar­eas—pri­ori­ti­sa­tion and policy work.

Our pri­ori­ti­sa­tion work has been about iden­ti­fy­ing con­sid­er­a­tions and meth­ods for choos­ing be­tween differ­ent ways of helping the world. Our policy work has been ex­plor­ing for high-value ways for the effec­tive al­tru­ism com­mu­nity to in­ter­act with policy pro­cesses. A lit­tle over one-third of our work has been policy, and the rest pri­ori­ti­sa­tion.

In the next sec­tions I’ll give an overview of what we’ve been do­ing in these ar­eas. I’ll cover most of our ac­tivi­ties, but a few are ex­cluded be­cause (i) they were a small time in­vest­ment and we do not re­gard them as that im­por­tant; (ii) we are still ex­plor­ing them and may not com­plete them; or (iii) the pro­jects in­volve other peo­ple, and we are be­ing cau­tious about not jeop­ar­dis­ing re­la­tion­ships.

Prioritisation

We have three ar­eas of pri­ori­ti­sa­tion work:

  1. Cause prioritisation

  2. Pri­ori­ti­sa­tion within catas­trophic risk reduction

  3. Aca­demic pri­ori­ti­sa­tion methodology

Cause prioritisation

This work is aimed prin­ci­pally at peo­ple who are in­ter­ested in mak­ing strate­gic choices about what ar­eas to fo­cus on in helping the world—in par­tic­u­lar the effec­tive al­tru­ism move­ment. We fo­cus on fa­cil­i­tat­ing com­par­i­sons be­tween di­verse ar­eas, rather than on mak­ing choices within ar­eas. We have pro­duced a flowchart as an overview and com­mu­ni­ca­tion tool, and also have some posts look­ing in depth at more spe­cific is­sues, for ex­am­ple try­ing to com­pare move­ment growth to more di­rect work. We have drawn on some of these tools (and the more method­olog­i­cal ones dis­cussed be­low) to ad­vise Effec­tive Altru­ism Ven­tures on how to make com­par­i­sons be­tween pro­jects in very differ­ent ar­eas. We are in­ter­ested in de­vel­op­ing the flowchart fur­ther into more de­tailed and in­ter­ac­tive ver­sions.

Pri­ori­ti­sa­tion within catas­trophic risk reduction

This is a strand of work which grew out of cause pri­ori­ti­sa­tion. We wanted to be able to provide prin­ci­pled com­par­i­sons be­tween work on catas­trophic risk re­duc­tion and other ar­eas. We still re­gard this as a very high-value ques­tion we have not prop­erly an­swered. How­ever in in­ves­ti­gat­ing it we de­vel­oped meth­ods for com­par­ing the value of differ­ent ar­eas of work within catas­trophic risk re­duc­tion. This helped us to se­cure fund­ing from the Fu­ture of Life In­sti­tute to de­velop and ex­tend the method­ol­ogy for pri­ori­ti­sa­tion within AI safety. Other work in this cat­e­gory looks at the timing of work on ex­is­ten­tial risk, and at how we can pro­duce crude es­ti­mates of the value of safety work in AI.

Aca­demic pri­ori­ti­sa­tion methodology

This work aims to iden­tify im­por­tant con­sid­er­a­tions in pri­ori­ti­sa­tion, and write them up rigor­ously so that they can be em­bed­ded in the aca­demic dis­course. We see this as an im­por­tant long-term route to im­pact. Our largest work in this space is a pa­per we are just ready­ing for jour­nal sub­mis­sion on es­ti­mat­ing the ex ante value of marginal re­search. This de­vel­ops a pro­ject (within cause pri­ori­ti­sa­tion) from last year in which we looked at es­ti­mat­ing the re­turns from work­ing on pro­jects of un­known difficulty. Also in the cat­e­gory of aca­demic pri­ori­ti­sa­tion, Owen has pre­pared a book chap­ter on dis­count­ing for un­cer­tainty in health, and con­tributed to an­other on in­te­grat­ing the val­u­a­tion of health and con­sump­tion effects.

Im­pact of pri­ori­ti­sa­tion work

It is hard to es­ti­mate ex­actly what the im­pact of differ­ent bits of this re­search is. Par­tic­u­larly the aca­demic work has quite a long feed­back loop, and for some of the other re­search we see the main value in its ca­pac­ity to in­form fu­ture more de­vel­oped re­search. Nonethe­less we are in­ter­ested in track­ing which parts of the work seem to be par­tic­u­larly valuable, and are worth pur­su­ing fur­ther or look­ing for op­por­tu­ni­ties to dis­sem­i­nate more widely. To date, we have mostly been eval­u­at­ing this by (i) ask­ing peo­ple whose judge­ment we trust which work is im­por­tant, (ii) see­ing which work is cited by oth­ers, and (iii) look­ing at which work gets most views on our web­site.

Policy

Our work in policy so far has largely been driven by high-value op­por­tu­ni­ties. We have seen two kinds of op­por­tu­nity: where we have a par­tic­u­larly good con­nec­tion to spe­cific peo­ple, or where we have an idea that seems im­por­tant.

Catas­trophic risk and policy

Much of our policy work has been helping to steer policy con­ver­sa­tions around catas­trophic and ex­is­ten­tial risk. We co­or­di­nated and wrote a policy re­port on Un­prece­dented Tech­nolog­i­cal Risks, which was shared within the UK gov­ern­ment, and led to two of the au­thors be­ing in­vited to write a chap­ter in the Govern­ment Chief Scien­tific Ad­vi­sor’s an­nual re­port. More re­cently, we have been ex­plor­ing more de­tailed cat­e­gori­sa­tions of sources of risk (fur­ther work forth­com­ing). Based on this ex­ist­ing work, we have se­cured fund­ing from the Fin­nish For­eign Ministry to write a re­port on ex­is­ten­tial risk and gov­er­nance, and we have be­gun prepa­ra­tion work for this.

The UK runs a na­tional as­sess­ment of pos­si­ble catas­trophic risks, as­sess­ing them on like­li­hood and im­pact. We no­ticed that parts of the method­ol­ogy they used for as­sess­ing im­pact was not well-grounded. We sent a short doc­u­ment ex­plain­ing how we thought this might be im­proved to the team work­ing on it, had a meet­ing with them, and sent a sec­ond doc­u­ment. They are in the pro­cess of con­clud­ing a re­view of their method­ol­ogy, and our un­der­stand­ing is that they have made some changes in the di­rec­tion of our recom­men­da­tions, and that we rep­re­sented a good frac­tion of the voices push­ing for these changes (al­though at least some were also re­quested by oth­ers). We think get­ting the method­ol­ogy cor­rect on this as­sess­ment could lead to switch­ing the or­der of con­cern about some risks and lead to more ap­pro­pri­ate re­search al­lo­ca­tion. We hope that it may be taken up in other large-scale risk as­sess­ments.

A spe­cific area we are in­ter­ested in is biosafety policy. This seems an im­por­tant and timely area. We pub­lished a pre­limi­nary policy idea last year which was well-re­ceived by some peo­ple in the biosafety com­mu­nity. Since then we have con­tinued con­ver­sa­tions, and looked at the broader aca­demic liter­a­ture on in­cen­tives and risks. More re­cently, this sum­mer we had a post­doc in life sci­ences join us as an in­tern and pre­pare neu­tral overviews of the risks, benefits and pos­si­ble re­sponses to “gain-of-func­tion” work in in­fluenza.

Other high-value policy opportunities

We’re in­ter­ested in other places where an effec­tive al­tru­ist per­spec­tive may be use­ful in in­form­ing policy. We think this is im­por­tant, be­cause most peo­ple in policy are try­ing to help, and iden­ti­fy­ing good policy can be difficult. So far much of our work here has been strate­gic. We are de­vel­op­ing one policy sug­ges­tion on im­prov­ing the clar­ity of com­mu­ni­ca­tion of statis­tics by gov­ern­ments.

People

The Global Pri­ori­ties Pro­ject was started in Jan­uary 2014. At that time it had 75% of Owen Cot­ton-Bar­ratt’s time fo­cus­ing on re­search, and a small amount of time from other in­di­vi­d­u­als at the Cen­tre for Effec­tive Altru­ism and Fu­ture of Hu­man­ity In­sti­tute, no­tably Niel Bow­er­man.

In Jan­uary 2015, Seb Far­quhar joined full time in a re­search and pro­ject man­age­ment ca­pac­ity. In July, Seb took on a role as Ex­ec­u­tive Direc­tor of the Cen­tre for Effec­tive Altru­ism, and re­duced his time on GPP to 50%.

We have also been for­tu­nate enough to have some very tal­ented in­terns join us over the sum­mers, typ­i­cally for 6-8 weeks. In 2014, Max Dal­ton worked with us. In 2015, Hong-Sheng Lim, Nick Phillips, Daniel Koko­ta­jlo, and Ni­co­las Moes worked with us.

The to­tal time (full time equiv­a­lent) that has been de­voted to GPP to date is there­fore roughly:

Owen Cot­ton-Bar­ratt: 16 months

Seb Far­quhar: 7 months

Niel Bow­er­man: 2 months

In­terns: 8 months

Others: 1 month

To­tal: 34 months (14 months in 2014, 20 months in 2015)

Fu­ture directions

His­tor­i­cally we have pur­sued both pri­ori­ti­sa­tion and policy work. We see them both as high-value ac­tivi­ties, and the op­por­tu­ni­ties were good enough that we wanted to take them. How­ever, it is not nec­es­sary that they are com­bined in one or­gani­sa­tion, and we are look­ing at sep­a­rat­ing out these work­streams more ex­plic­itly.

We’re also in­ter­ested in in­creas­ing our at­ten­tion to ar­eas other than ex­is­ten­tial risk. A year ago, we thought this was sig­nifi­cantly ne­glected for both policy and pri­ori­ti­sa­tion. How­ever it is get­ting in­creas­ing amounts of at­ten­tion, and this means that go­ing for­ward it may not be the most ne­glected area. We are heav­ily en­gaged in the area, and our cur­rent fund­ing is fo­cused on this, but we are in­ter­ested in push­ing more on strat­egy/​pri­ori­ti­sa­tion more broadly, and on other ar­eas of policy.

Also note that we are cur­rently hiring. We are look­ing to bring on more re­searchers and to build our policy team.

[Cross-posted from our web­site. Thanks to Michael Dick­ens for feed­back which prompted this post.]