Terminal versus instrumental goals / values / preferences

‘In­stru­men­tal goals’ or ‘in­stru­men­tal val­ues’ are things that an agent wants for the sake of achiev­ing other things. For ex­am­ple, we might want to get into a car, not be­cause we en­joy the act of open­ing car doors for their own sake, but be­cause we want to drive some­where else.

‘Ter­mi­nal’ goals, val­ues, or prefer­ences are those where the prefer­ence is de­rived lo­cally rather than by look­ing at fur­ther or dis­tant con­se­quences. If you en­joy eat­ing choco­late (and oth­er­wise ap­prove of this en­joy­ment, etcetera) then you aren’t de­riv­ing your prefer­ence based on what you be­lieve to be the fur­ther con­se­quences of eat­ing choco­late.

Imag­ine re­al­ity as an enor­mous web of events, linked by cause and effect. “Ter­mi­nal value” is usu­ally lo­cal and be eval­u­ated at a sin­gle event in­side the graph; even if it’s a non­lo­cal good thing, we’d eval­u­ate it by eval­u­at­ing the his­tory up to some point, and then we’d have a chunk of definite good­ness that would stand on its own no mat­ter what hap­pened later.

“In­stru­men­tal value” is a non­lo­cal prop­erty of an event, de­pend­ing on its real or ex­pected fu­ture, and con­tin­gent on that fu­ture; if you add up all the in­stru­men­tal val­ues on the graph, you don’t get a mean­ingful sum be­cause you may be dou­ble-count­ing some value.

On a moral or eth­i­cal level, in­stru­men­tal val­ues are jus­tified by ap­peal­ing to their con­se­quences, while ter­mi­nal val­ues are jus­tified with­out ap­peal to their con­se­quences.

Fur­ther read­ing:


  • Metaethics

    Me­taethics asks “What kind of stuff is good­ness made of?” (or “How would we com­pute good­ness?”) rather than “Which par­tic­u­lar poli­cies or out­comes are good or not-good?”