What Does Make a Difference? It’s Really Simple

This is re­ally sim­ple:
Sup­pose you want to check if some ac­tion of yours makes a differ­ence.
How to do it?
The wrong thing to do:
Think of the con­se­quences of your ac­tion and eval­u­ate them to see if they fit your pur­poses. If they do, go on and do it.

The rea­son for this be­ing wrong: If some­one else does some­thing with the same con­se­quences, and if your do­ing or not your ac­tion makes no differ­ence to the fact that THAT per­son will do it, then you are not nec­es­sary for those con­se­quences, they would hap­pen any­way.
This is also true if some­thing, not a per­son, would do an ac­tion with the same con­se­quences.

The right thing to do: Con­sider what would hap­pen if you DIDN’T do your ac­tion. Sub­tract that from what would hap­pen if you DID do your ac­tion.
This is the differ­ence it would make if you did it.

There is a rea­son it is called a ‘differ­ence’, it is the differ­ence be­tween you do­ing it and you not do­ing it.

Ex­am­ple: Sup­pose you think you will make a differ­ence by care­fully con­sid­er­ing your vote, and vot­ing.

Wrong: Well, I’m par­tially causally re­spon­si­ble for the elec­tion of X so my ac­tion would make a differ­ence.

Right: If I do vote or if I don’t vote, the same can­di­dates will be elected. There­fore my vote makes no differ­ence.
(In more than 16000 elec­tions in the USA it was NEVER the case that one vote would have made a differ­ence)

The AWFUL ar­gu­ment peo­ple usu­ally say: But what if ev­ery­one did it?

The rea­son it does not work: Every­one will NOT do it. Yes. That sim­ple.

The rea­son it is awful: Com­pare “I don’t think I should go to the movies to­day, what if ev­e­ri­one did it?”

So, when you are will­ing to make a differ­ence, not feel good, not do what ev­ery­one does, not clean your con­scious­ness. When you want to REALLY, REALLY make a differ­ence, you should con­sider the differ­ence be­tween do­ing and not do­ing it.

It is that sim­ple.

(Note to the Less Wrong en­try: I know most peo­ple here know that poli­tics is a mind kil­ler, but delv­ing deeper into the sim­ple ar­gu­ment, no­tice it works for any case of overde­ter­mi­na­tion, such as post­ing a com­ment or en­try that some­one else would. If you re­al­ize now how im­por­tant it is to figure out what is im­por­tant, don’t for­get more than one per­son have already done it, use them.)

• Sup­pose that you and some­body else shoot some­body at the ex­act same time. Be­cause they would have died any­ways, your kil­ling them hasn’t caused any harm, and so is morally OK by your the­ory. How­ever, the other kil­ler is also morally blame­less un­der your the­ory, be­cause if they hadn’t kil­led the other per­son, you would have.

I be­lieve that this re­quires a mod­ifi­ca­tion.

• This en­tire ar­gu­ment is defi­ni­tion driven. If you define “mak­ing a differ­ence” in the way stated, the au­thor is cor­rect. If you don’t, he’s not. As clev­erly pointed out el­se­where, the two-per­son shoot­ing in­di­cates that this method of eth­i­cal eval­u­a­tion is com­pletely at odds with al­most any ac­tual per­son’s eth­i­cal be­liefs. For a more real world ex­am­ple, con­sider the mur­der of Julius Cae­sar, or re­ally any con­spir­acy plot—it would have hap­pened with­out you, but your in­volve­ment was still im­proper. It may not have “made a differ­ence,” but vir­tu­ally ev­ery per­son’s con­cept of moral­ity will reg­ister it as im­moral.

In­deed, that is ex­actly why this is a use­less defi­ni­tional dis­pute: know­ing that some­thing “made a differ­ence” doesn’t tell us any­thing about that thing other than it would not have hap­pened had you not done it. You seem to be en­dors­ing this as some kind of moral frame­work, which I think it over­whelm­ingly fails as. This is par­tic­u­larly true if you hold the ideal that any moral frame­work should be uni­ver­sally prac­ti­ca­ble, which this one is ob­vi­ously not.

And a point of fo­rum eti­quette: use ital­ics rather than CAPS. CAPS tend to read like I’M SO RIGHT I’M YELLING! and, at least in my ex­pe­rience, cor­re­late in­versely with hav­ing a good point.

• Down­vot­ing for poor ex­pla­na­tion, un­nec­es­sary use of al­l­caps, and strong as­ser­tions made with lit­tle ev­i­dence or rea­son­ing to back them up.

• Se­conded. Ter­rible ex­po­si­tion; it triv­ial­izes some­thing that is non-triv­ial. Also, it would be nice if the writer used para­graphs and did not use CAPS (un­less re­ally shout­ing).

• I’m not sure if I should re­spond, but …

You also have to take into ac­count the fact that other peo­ple are largely similar to you and there­fore, when you set the out­put of the al­gorithm that defines you, you are also set­ting the out­put of ev­ery other al­gorithm if and to the ex­tent that it re­sem­bles the one that is you, so the effect is ac­tu­ally larger than your ac­tion-differ­ence taken in iso­la­tion.

Well, it made much more sense when Eliezer Yud­kowsky and Wei Dai said all that stuff about similar com­pu­ta­tional pro­cesses and set­ting their log­i­cal out­put...

• What is the prob­a­bil­ity your vote will make a differ­ence?

About 1 in 60 mil­lion for the av­er­age Amer­i­can voter on the last pres­i­den­tial elec­tion day. Was the to­tal differ­ence be­tween your fa­vorite can­di­date and the other one, di­vided by 60 mil­lion, worth your time and effort? Quite pos­si­bly. At the very least it seems ab­surd to say it definitely wasn’t worth your time.

• If 16000000 seems like worth it to you, I would sug­gest spend­ing much less time and effort buy­ing lot­tery tick­ets and af­ter win­ning (here in Brazil the odds are 1 to 52 mil­lion) in­vest your money in fi­nanc­ing SIAI or what­ever you think makes a differ­ence....

• You’ve also got to con­sider the size of the out­come. Win­ning a lot­tery gives you, what, a few hun­dred thou­sand to a few mil­lion dol­lars? The ex­pected value of buy­ing a lot­tery ticket is then around 2 cents- less than the price of a lot­tery ticket, and so buy­ing is a ter­rible de­ci­sion.

Elec­tions have much larger im­pacts. The 2000 elec­tion is a clas­sic ex­am­ple: it’s been ar­gued that Gore would not have in­vaded Iraq, and that the Iraq in­va­sion has cost around a trillion dol­lars; in that case, your vote has an ex­pected value to the US of around 16,000 USD. (More if you live in a swing state, more if you be­lieve Gore would have been bet­ter in other ways. Less oth­er­wise.)

Not that I’m say­ing that this is a closed ques­tion- in fact, my point is that it isn’t.

• Note that you also have to con­sider how cer­tain it is that your preferred can­di­date re­ally is bet­ter. And since the di­vi­sion of opinion is usu­ally 5050 or so, un­less you are ex­tremely over­con­fi­dent, it is quite un­cer­tain that your can­di­date is bet­ter. So you need do mul­ti­ply the benefit not only by the prob­a­bil­ity that you will cause your can­di­date to win, but also by the prob­a­bil­ity that your can­di­date is bet­ter, and then you have to sub­tract the di­su­til­ity in the case that your can­di­date turns out to be worse.

In other words, don’t vote.

• Note that the prob­a­bil­ity in gen­eral that an elec­toral sys­tem us­ing some­thing like the elec­toral col­lege in­creases the chance for ev­ery­one that they will be the de­cid­ing vote as­sum­ing that no vot­ing area is ex­tremely in the di­rec­tion of one can­di­date. To see this with a small ex­am­ple, con­sider the toy ex­am­ple of 9 peo­ple with three states each of three peo­ple with an elec­tion for two can­di­dates.. If the elec­tion is by pop­u­lar vote then the only way your vote mat­ters is if ex­actly four of the peo­ple vote for one can­di­date and ex­actly four vote for the other. But, if you use the states then your vote breaks a tie when­ever the other two states dis­agree and your state has a 1-1 for the other two vot­ers. That’s a much larger set of cir­cum­stances. This ap­plies in gen­eral al­though the math to show it be­comes a bit uglier when one has many differ­ent states of differ­ent sizes.

• Agree with your point. But no­tice that if we con­sider the case in which you ac­tu­ally win, which would be GuySrini­vasan’s pro­posal, then those 4 mil­lions you win in the lot­tery could be in­vested in things that will have much greater pro­por­tional im­pact, for you’d be in­vest­ing in the curve’s tail....

Just a re­minder. The post is NOT about poli­tics and vot­ing. It is about overde­ter­mi­na­tion and de­ci­sion the­ory.

• A quote from the linked ar­ti­cle:

A prob­a­bil­ity of 1 in 10 mil­lion is tiny but, as dis­cussed by Edlin, Gel­man, and Ka­plan (2007), can provide a ra­tio­nal rea­son for vot­ing; in this per­spec­tive, a vote is like a lot­tery ticket with a 1 in 10 mil­lion chance of win­ning, but the pay­off is the chance to change na­tional policy and im­prove (one hopes) the lives of hun­dreds of mil­lions, com­pared to the al­ter­na­tive if the other can­di­date were to win.

(it was 1 in 10 mil­lion in New Mex­ico, Virginia, New Hamp­shire, and Colorado)

If you want to make a post about overde­ter­mi­na­tion, I’d say don’t use the vot­ing ex­am­ple, since here’s one per­son at the very least who thinks the ex­am­ple’s far from clear-cut. The movies thing is fine—the prob­a­bil­ity enough peo­ple go to ruin ev­ery­one’s ex­pe­rience times the ex­pe­rience ru­ined is still tiny, not plau­si­bly large.