Cause: Better political systems and policy making.

Should EAs be fight­ing for bet­ter poli­ti­cal sys­tems and bet­ter policy mak­ing? For gov­er­nance where the de­ci­sion mak­ers, at min­i­mum, are in­cen­tivised to act in the best long-term in­ter­est of the pop­u­la­tion?

At first glance this could be su­per im­por­tant.

  • If you think policy mak­ers should be putting in place poli­cies that are ac­tu­ally good for the pop­u­la­tion,

  • If you think gov­ern­ments should care about long term risks to hu­man­ity,

  • If you think more pri­ori­ti­sa­tion re­search should be done into how best to use re­sources, [1]

  • If you think that we need to see sig­nifi­cant sys­temic changes,

then try­ing to im­prove how Govern­ments work could lev­er­age huge amounts of ac­tion and re­sources go­ing to­wards these goals.

But what does cre­at­ing bet­ter poli­ti­cal sys­tems ac­tu­ally look like? Is this even fea­si­ble? Is this re­ally a high im­pact use of time?

The key fea­tures of bet­ter poli­tics and policy

Most of the ba­sic fea­tures of a good poli­ti­cal sys­tem are in place in most de­vel­oped coun­tries. We mostly have democ­racy, uni­ver­sal suffrage, peace­ful trans­fers of power, an in­de­pen­dent rule of law, term limits, free speech, free press, no ex­ces­sive ed­u­ca­tion con­trols, limit on Govern­ment’s’ pow­ers, and so on. (In fact for those of us liv­ing in coun­tries where there above ap­plies we should be pretty damn grate­ful to have those ba­sics fea­ture—it re­ally could be so much worse.)

That said, if we want to up­grade form the ba­sic pack­age there are a few other fea­tures of a well de­signed poli­ti­cal sys­tem it would be re­ally nice to have. Such as:

  1. Hon­est lead­ers and cam­paigns.

  2. Min­imise poli­ti­cal tail end risks. Prevent the rise of Hitler type figures who can turn a democ­racy into a mon­strous dic­ta­tor­ship of suffer­ing, or of in­di­vi­d­u­als who may be trig­ger happy on press­ing the big red doom the earth but­ton.

  3. Lead­ers who care about the long-term fu­ture of hu­man­ity and will deal with is­sues that need in­ter­na­tional co­op­er­a­tion (eg. cli­mate change) the other risks (eg. AI).

  4. No tyranny of the ma­jor­ity situ­a­tions (eg. ban­ning harm­less re­li­gious ac­tions).

  5. Dis­in­cen­tive poli­ti­ci­ans from rhetoric or ac­tions that will help them win elec­tions but are not in the best in­ter­est of the pop­u­la­tion (eg. a policy that will benefit vot­ers in swing lo­ca­tions but harm other vot­ers)

  6. The use of ev­i­dence in policy mak­ing and the im­ple­men­ta­tion of poli­cies that ac­tu­ally work to achieve the stated goals.

  7. Allow limited low-risk im­prove­ments to the sys­tem (such as im­prov­ing vot­ing sys­tems).

Ini­tial re­search on what should change and how to cre­ate change

But those all sound im­pos­si­ble, can we ac­tu­ally cre­ate a sys­tem that will de­liver all of the above?

Un­for­tu­nately I am go­ing to give the usual EA cop-out an­swer: we should do more re­search in this area.

We should re­search:

  • Pri­ori­ti­sa­tion ques­tions: what is most im­por­tant about a work­ing poli­ti­cal sys­tem (this could mean adding to and or­der­ing the list above)? Which coun­tries are the most im­por­tant and tractable to fo­cus time and effort on?

  • Good poli­ti­cal sys­tems: what are the prac­ti­cal things that could be im­ple­mented in a demo­cratic sys­tem that provide the bonus fea­tures listed above.

  • Im­ple­men­ta­tion tac­tics: What tac­tics are needed to cre­ate change that leads to good poli­ti­cal sys­tems.

Luck­ily for any­one in­ter­ested I’ll give you a kick­start by shar­ing my (not en­tirely un­in­formed) opinions on the above. What more could you need?

Pri­ori­ti­sa­tion ques­tions: I think the most im­por­tant thing will be to have sys­tems that limit ex­treme risks, as I am gen­uinely wor­ried we are not go­ing to be able to keep hu­man­ity al­ive that much longer, and to fo­cus on more pow­er­ful coun­tries, as these are the ones that have the most power to cre­ate change and lead the way. But I have no strong views on this.

Good poli­ti­cal sys­tems: I have slightly more nu­anced views on this. In brief some of the key fea­tures of a good poli­ti­cal and policy sys­tem would be:

  1. An un­elected part to the sys­tem that can veto or de­lay very bad de­ci­sions, such as the UK House of Lords.

  2. A con­sti­tu­tion /​ in­ter­na­tion­ally agreed hu­man rights bill /​ etc that gives ad­di­tional power to the in­de­pen­dent ju­di­ciary to limit ex­treme ac­tions by the state.

  3. Trans­parency. Not nec­es­sar­ily 100% trans­parency but more trans­parency about how poli­ti­cal de­ci­sions are made. (Eg. ear­lier re­lease of pub­lic records)

  4. Bet­ter le­gi­t­imised whistle­blow­ing sys­tems for civil ser­vants and oth­ers who might see bad poli­ti­cal behaviours

  5. Safe­guards to al­low party elites to push out strongly dis­agree­able lead­er­ship. See the top­i­cal ex­am­ple be­low.

  6. Ev­i­dence en­forc­ing bu­reau­cracy. For ex­am­ple mak­ing civil ser­vants have to fill out a form ex­plain­ing the ev­i­dence for and against a policy (like the UK’s bet­ter reg­u­la­tion frame­work).

  7. A code of con­duct and train­ing for civil ser­vants that en­courages good de­ci­sion mak­ing and hon­esty.

  8. All new poli­cies are re­viewed a set amount of time af­ter they are in­tro­duced

  9. Public con­sul­ta­tion for new policies

  10. Avoid­ing sys­tems like pri­maries where a poli­ti­cian may have to take a more ex­treme view to win their party lead­er­ship than they would take oth­er­wise.

  11. Limit the things the pub­lic can vote on. Democ­racy works best when the pop­u­la­tion votes on re­fined choices

  12. A sys­tem whereby party lead­ers have to have held office pre­vi­ously. Eg in the UK you have to be an MP be­fore you can be made party leader.

  13. More power moved out of the hands of poli­ti­ci­ans to tech­nocrats in cer­tain ar­eas. For ex­am­ple, the Bank of England play­ing an in­de­pen­dent role to en­sure fi­nan­cial sta­bil­ity.

  14. Bet­ter and more rep­re­sen­ta­tive vot­ing sys­tems.

Happy to elab­o­rate on any of these if peo­ple want. Worth bear­ing in mind that these kind of things will vary from coun­try to country

Im­ple­men­ta­tion tac­tics: Highly spec­u­la­tive and I am not an ex­pert in lob­by­ing (es­pe­cially out­side the UK). I think I sug­gest be­gin­ning by build­ing cred­i­bil­ity with high qual­ity re­search in this area, per­haps set­ting up (or work­ing with) a re­spected think tank. From there I would just sit back and wait for op­por­tu­ni­ties to arise. Wait un­til such top­ics are be­ing dis­cussed or on the poli­ti­cal agenda and then start in­fluenc­ing at that point. Alter­na­tively it could be worth look­ing for any policy changes that whilst seem­ingly small or un­con­tro­ver­sial but may have a sig­nifi­cant im­pact if im­ple­mented. How to have in­fluence would de­pend on the situ­a­tion and could be grass­roots, poli­ti­cal net­work­ing, le­gal and so on.

A top­i­cal ex­am­ple of how re­cent poli­ti­cal tragedy could have been avoided.

Without mean­ing to offend any­one who has differ­ing poli­ti­cal views than me, I think it is fair to say we have all re­cently seen and been shocked by a tremen­dous poli­ti­cal catas­tro­phe. Yes I am of course talk­ing about the dis­in­te­gra­tion of the UK Labour Party.

The pre­vi­ous Labour leader, Ed Miliband, changed the way by which labour lead­er­ship is de­cided. The lead­er­ship was changed to a pub­lic vote from a vote split 13 unions, 13 MPs and 13 pub­lic. In or­der to make sure they did not end up with a leader that all the MPs hated or could not work with they in­tro­duced a safe­guard that a leader would have to have the sup­port of 50 MPs to run. They did not how­ever think to put in any safe­guards to give MPs power to kick out a leader, even if ev­ery al­most ev­ery MP found them difficult to work with. And the cur­rent sham­bles is a re­sult of that. [2]

Per­haps there is with hind­sight bias and a lot of hubris lead­ing me to thank that this should have been no­tice­able. But it cer­tainly feels like this is the kind of thing that a smart per­son, with a good knowl­edge of how poli­ti­cal sys­tems worked and a long term mind­set, could have spot­ted. Espe­cially if that per­son had some time to do some re­search when­ever a ma­jor party in a ma­jor coun­try was chang­ing how they de­cided on their leader. This is a mis­take other par­ties have made and cor­rected (such as the con­ser­va­tive party’s 1922 com­mit­tee). Fur­ther­more it should not have been hard in this case to make the case to the rele­vant de­ci­sion mak­ers.

Fur­ther thoughts on tractability

How tractable this cause area is, is in­cred­ibly hard to guess. The ex­am­ple above makes it seem like op­por­tu­ni­ties to make changes do arise. It might be worth look­ing to see if there are many other past ex­am­ples of situ­a­tions where it feels like changes could have been made. If any EAs were to work on this it could be worth mak­ing pre­dic­tions about how many fu­ture such cases ac­tu­ally come up, where EA in­fluence could po­ten­tially help.

It is also my ex­pe­rience from my time in Govern­ment that there are many very small changes that could be im­ple­mented that could im­prove de­ci­sion mak­ing. I have seen bet­ter and worse de­ci­sions made and a va­ri­ety of checks and bal­ances in place that worked and did not work to vary­ing de­grees. To give one in­nocu­ous ex­am­ple from an area I worked in: the UK Trea­sury pub­lishes a Tax Im­pact and In­for­ma­tion Note (TIIN) for all tax policy changes, ex­cept for changes to lo­cal taxes. Ask­ing for this TIIN sys­tem to be ex­tended to lo­cal taxes seems sim­ple and should pos­i­tively im­pact de­ci­sion mak­ers in­cen­tives.

One counter con­sid­er­a­tion is that it may be very difficult to know what the ac­tual im­pact of changes will be. For ex­am­ple you could push for more trans­parency of doc­u­men­ta­tion and it may lead to less in­for­ma­tion be­ing writ­ten down and worse de­ci­sions be­ing made.

Is it worth EAs time to fo­cus on this?

Some things to con­sider are:

Scale. HUGE. Govern­ments have a lot of re­sources that they can put to­wards ad­dress­ing global prob­lems. Im­prov­ing how they use those re­sources would im­pact the flow of trillions of dol­lars, and could im­pact ev­ery area of what a gov­ern­ment does from so­cial care to cli­mate policy. The flow through effects seem very large.

Ne­glected. Neg­elect­ed­ness seems to per­haps be the wrong thing to con­sider with poli­ti­cal changes is­sues. If an is­sue is too ne­glected it may be difficult to cre­ate the pres­sure needed for change. The ideal would ei­ther be that there is pub­lic sen­ti­ment that change is needed but no clear leader push­ing for change, or that the re­search in this area is not be­ing pushed hard enough or is not be­ing car­ried out with an EA mind-set. There are already a few or­gani­sa­tions work­ing on ev­i­dence based policy mak­ing, vot­ing sys­tems and trans­parency. Could do with some­one look­ing more into what ex­ists in this area and what the gaps are.

Tractabil­ity. Fairly low. Dis­cussed above.

Mea­sura­bil­ity. This is re­ally re­ally hard to mea­sure the suc­cess of. In gen­eral work­ing out the im­pact of policy in­fluenc­ing is difficult: it is hard to know if your ac­tions will cre­ate change, if change hap­pens it is difficult to know if your con­tri­bu­tion was a causal fac­tor. You may need some­one com­pe­tent work­ing on this for years to even be­gin to make any wins. Fur­ther­more if we cre­ate changes that lead to bet­ter changes at some in­visi­ble point in the fu­ture or miti­gate risks it may be im­pos­si­ble to see the end im­pact of this.

Why would EA think­ing be uniquely placed to con­sider this? This uses both long term think­ing about ex­treme con­se­quences (not putting in the right safe­guards into poli­ti­cal de­ci­sions will likely have no im­me­di­ate effect but may lead to catas­tro­phe a few hun­dred years down the line) and meta-think­ing (it may make more sense to try to im­prove gov­er­nance sys­tems than panic when a poor poli­ti­cian is stand­ing or a poor de­ci­sion is made and won­der how it reached this point) which are ar­eas EAs fo­cus on.

My bi­ases and thoughts: I have had an in­ter­est in this topic since shortly af­ter I be­gun work­ing in policy. I guess re­cent events have trig­gered me to ac­tu­ally writ­ing some views on this.

Re­spond­ing to re­cent poli­ti­cal situations

Many EAs are con­cerned about cur­rent poli­tics. A re­sponse to this could be by try­ing to change the cur­rent situ­a­tion in some way or by try­ing to change the sys­temic is­sues that led to the cur­rent situ­a­tion. Try­ing to tackle the sys­temic is­sues, to pre­vent the rise of po­ten­tial fu­ture dan­ger­ous dem­a­gogues or prob­le­matic par­ties seems more ne­glected and plau­si­bly a bet­ter way to have an im­pact than fight­ing the cur­rent situ­a­tion (un­less you think the cur­rent situ­a­tion is su­per dan­ger­ous). The ques­tion then would be which things about the mod­ern world should we try to change. We could look to im­prove: the val­ues of the pub­lic, the poli­ti­cal sys­tems, the me­dia, big busi­ness, or other fac­tors. I have no strong views on this; I have fo­cused in this piece on im­prov­ing poli­tics but would be in­ter­ested in how you can im­prove each of the above. [3]

Fur­ther­more re­cent poli­ti­cal changes that have left vast swathes of the pop­u­la­tion up­set and shocked. A ques­tion to ask would be do these changes pre­sent an op­por­tu­nity for do­ing good. Can we cap­i­tal­ise on and di­rect peo­ple’s pas­sion and dis­ap­point­ment, for these or for the po­ten­tial fu­ture poli­ti­cal is­sues? [4] Can (and should we) we do this in a non-par­ti­san way that builds sup­port from across the poli­ti­cal spec­trum to im­prove poli­ti­cal and other sys­tems? [edit:] How do we di­rect peo­ple’s pas­sion to­wards the most use­ful cause, and is this the cause to di­rect them to­wards?


Some con­clu­sions (sec­tion ed­ited [5]) :
  1. This topic should be a key fo­cus area for EAs. There should be more re­search by in­di­vi­d­ual EAs, by CEA, by Open Philan­thropy, and so on.

  2. The EA com­mu­nity’s re­sponses to re­cent poli­ti­cal events have been poor: frac­tured and slow and un­fo­cused. We should as­sume there will be an­other event that will rile peo­ple up (French elec­tions? Some­thing else?) and work fast now to be more ready di­rect peo­ple’s pas­sions in the op­ti­mal di­rec­tion (such as to­wards bet­ter poli­ti­cal sys­tems). More gen­er­ally we are wast­ing too much time dis­cussing non-ne­glected poli­ti­cal is­sues.

  3. This is clearly more im­pact­ful than other policy ar­eas (ex­cept where there is an im­me­di­ate risk to counter, such as ar­guably bio-se­cu­rity policy). It is clearly higher im­pact to change and im­prove how policy is made than to change the kind of ar­eas of policy that the Open Philan­thropy Pro­ject are look­ing into, (such as eco­nomic sta­bil­ity policy or prison re­form or an­i­mal welfare).

Tell me why I am wrong on points 1 to 3 above. Tell me if you strongly agree. Your feed­back would be su­per-helpful.

If any­one wants to work on this or fund this then get in touch:


With thanks to Alexan­der Gor­don Brown and Robert Col­lins for in­put.

[1] Bear in mind that states are by far the biggest spender by a long way on what can loosely be con­sid­ered pri­ori­ti­sa­tion re­search and on risk re­search. Sates con­stantly have to think about how to spend money and re­sources to cre­ate the biggest im­pact al­though per­haps with a slightly skewiff view of what is im­pact and they may not nec­es­sar­ily im­ple­ment the most high im­pact op­tion for a range of rea­sons.

[2] I have not looked at this in de­tail and per­haps over sim­plify the situ­a­tion. For more on this see: http://​​­states­​​poli­tics/​​elec­tions/​​2015/​​09/​​why-did-labour-use-sys­tem-elect-its-leader and http://​​­states­​​poli­tics/​​uk/​​2015/​​05/​​how-will-labour-lead­er­ship-elec­tion-work

[3] Some re­cent dis­cus­sion on how to im­prove the me­dia can be found at: https://​www.face­​groups/​effec­­tru­ists/​perma­l­ink/​1213547848701570/​

[4] Past think­ing on this is here.

[5] I ini­tially gave this ar­ti­cle the mild man­nered (and more hon­est) con­clu­sion that: “this could be a pow­er­ful thing for EAs to work on, but the case for do­ing so is still weak. At the min­i­mum more re­search could be use­ful” but de­cided to in­stead to try to spark more de­bate. I per­son­ally do not fully agree with points 1 to 3 and think I am over­stat­ing the case.