Ex ante prioritarianism and negative-leaning utilitarianism do not override individual interests


It seems that “ex ante” views (like ex ante pri­ori­tar­i­anism) haven’t been dis­cussed much within the EA com­mu­nity. Ba­si­cally, the ap­proach is to ag­gre­gate the util­ity in each in­di­vi­d­ual first, over their life and by tak­ing the ex­pec­ta­tion, and then ap­ply what­ever so­cial welfare func­tion you like to the re­sult­ing in­di­vi­d­u­ally ag­gre­gated util­ities.

Fur­ther­more, you could take these in­di­vi­d­ual ag­gre­ga­tions/​ex­pec­ta­tions con­di­tional on ex­is­tence (past, cur­rent or fu­ture), and only in­clude the terms for ac­tual (past, cur­rent or fu­ture) in­di­vi­d­u­als; so the set of in­di­vi­d­u­als to ag­gre­gate over would be a ran­dom vari­able. You’d then take an­other ex­pec­ta­tion, this time of of the so­cial welfare func­tion ap­plied to these ag­gre­gated util­ities over the set of ex­ist­ing in­di­vi­d­u­als.

The main benefit here is to avoid ob­jec­tions of over­rid­ing in­di­vi­d­ual in­ter­ests while still be­ing pri­ori­tar­ian or nega­tive-lean­ing, since we can treat per­sonal and in­ter­per­sonal trade­offs differ­ently.

Math formalism

We define to be the ag­gre­gated util­ity of in­di­vi­d­ual over all time (or just the fu­ture), in a given de­ter­mined out­come (no ex­pec­ta­tions ap­plied yet); in the out­comes in which they haven’t ex­isted and won’t ex­ist, is left un­defined. Then we define

and we ap­ply our so­cial welfare func­tion to the set

E.g., for some func­tion which is in­creas­ing (or non-de­creas­ing) and con­cave. Some ex­am­ples here. To­tal util­i­tar­i­anism has for all , and the ex ante view ap­plied to it ac­tu­ally makes no differ­ence. A fairly strong form of nega­tive util­i­tar­i­anism could be defined by for all , i.e. if and , oth­er­wise; this means that as long as an in­di­vi­d­ual is ex­pected to have a good life (net pos­i­tive value), what hap­pens to them doesn’t mat­ter, or could be lex­i­cally dom­i­nated by con­cerns for those ex­pected to have nega­tive lives (i.e. only if we can’t im­prove any nega­tive lives, can we look to im­prov­ing pos­i­tive ones).

Fi­nally, we rank de­ci­sions based on the ex­pec­ta­tion of over :


We can be both pri­ori­tar­ian or nega­tive-lean­ing and avoid over­rid­ing in­di­vi­d­ual in­ter­ests; we don’t give greater weight to the bad over the good in any in­di­vi­d­ual’s life, but we give greater weight to bad lives over good lives. Per­sonal and in­ter­per­sonal trade­offs would be treated differ­ently. You would be per­mit­ted, un­der an ex ante pri­ori­tar­ian or nega­tive-lean­ing view, to choose great suffer­ing to­gether with great bliss or risk great suffer­ing for great bliss all within one in­di­vi­d­ual, but you can’t im­pose great suffer­ing on one in­di­vi­d­ual to give great bliss to an­other (de­pend­ing on the ex­act form of the so­cial welfare func­tion).

Let’s look at an illus­tra­tive ex­am­ple where the ex ante view dis­agrees with the usual (“ex post”) one, taken from “Pri­ori­tar­i­anism and the Separate­ness of Per­sons” by Michael Ot­suka (2012):

Two-per­son case with risk and in­versely cor­re­lated out­comes: There are two peo­ple, each of whom you know will de­velop ei­ther the very se­vere or the slight im­pair­ment and each of whom has an equal chance of de­vel­op­ing ei­ther im­pair­ment. You also know that their risks are in­versely cor­re­lated: i.e., when­ever one of them would suffer the very se­vere im­pair­ment, then the other would suffer the slight im­pair­ment. You can ei­ther sup­ply both with a treat­ment that will surely im­prove a re­cip­i­ent’s situ­a­tion if and only if he turns out to suffer the very se­vere im­pair­ment or sup­ply both with a treat­ment that will surely im­prove a re­cip­i­ent’s situ­a­tion if and only if he turns out to suffer the slight im­pair­ment. An effec­tive treat­ment for the slight im­pair­ment would provide a some­what greater in­crease in util­ity than would an effec­tive treat­ment for the very se­vere im­pair­ment.

An ex ante pri­ori­tar­ian would choose to treat the slight im­pair­ment, while the usual (ex post) pri­ori­tar­ian who does not first ag­gre­gate or take ex­pec­ta­tions over the in­di­vi­d­ual would choose to treat the very se­vere im­pair­ment. From the point of view of each in­di­vi­d­ual, treat­ing the slight im­pair­ment would be prefer­able.

EDIT: Here’s an ex­am­ple which might seem pretty weird to some but also a bit in­tu­itive to oth­ers:

Sup­pose there are two peo­ple who are equally well off, and you are con­sid­er­ing benefit­ting ex­actly one of them by a fixed given amount (the amount of benefit would be the same re­gard­less of who re­ceives it).

Then, if you are an ex ante pri­ori­tar­ian, it would be bet­ter to choose one to benefit at ran­dom than to use a de­ter­minis­tic rule to choose one. How­ever, the ac­tual out­come will be the same, up to swap­ping the two in­di­vi­d­u­als’ util­ities.

For what it’s worth, un­der empty in­di­vi­d­u­al­ism (the view that one phys­i­cal per­son over time should re­ally be treated as a se­quence of dis­tinct in­di­vi­d­u­als from mo­ment to mo­ment, per­son-mo­ments), ap­ply­ing this ex ante mod­ifi­ca­tion ac­tu­ally doesn’t make any differ­ence. It’ll look like we’re over­rid­ing prefer­ences, but un­der empty in­di­vi­d­u­al­ism, there are only in­ter­per­sonal trade­offs, no per­sonal trade­offs. See also “Does Nega­tive Utili­tar­i­anism Over­ride In­di­vi­d­ual Prefer­ences?” by Brian To­masik.

Refer­ences and other reading

“Pri­ori­tar­i­anism and the Separate­ness of Per­sons” by Michael Ot­suka (2012) de­scribes this ap­proach, and gives ex­am­ples to raise ob­jec­tions to pri­ori­tar­i­anism and ex ante pri­ori­tar­i­anism.

That is­sue of Utili­tas is fo­cused on pri­ori­tar­i­anism, with a pa­per by Parfit which also dis­cusses ex ante views (I have yet to read it).

“De­cide As You Would With Full In­for­ma­tion! An Ar­gu­ment Against Ex Ante Pareto” by Marc Fleur­baey and Alex Voorho­eve (2013) cited in “Pri­ori­tairanism and the Separate­ness of Per­sons”, has more crit­i­cism of ex ante views.

Toby Ord’s ob­jec­tions to pri­ori­tar­i­anism and nega­tive util­i­tar­i­anism which do not ap­ply to the ex ante view:

“Why I’m Not a Nega­tive Utili­tar­ian”

“A New Coun­terex­am­ple to Pri­ori­tar­i­anism” (2015)