Empathy and compassion toward other species decrease with evolutionary divergence time

Link post

Un­der­stand­ing how much to morally value an­i­mals of differ­ent species rel­a­tive to each other and rel­a­tive to hu­mans is highly con­se­quen­tial for EA.

Un­der­stand­ing how laypeo­ple ac­tu­ally do morally value an­i­mals in prac­tice is con­se­quen­tial for EA in a va­ri­ety of ways:

  • We may give some weight to moral judg­ments of large num­bers of in­di­vi­d­u­als, with a broader set of moral views than the EA com­mu­nity, on the grounds of moral un­cer­tainty or peer disagreement

  • We may think that iden­ti­fy­ing in­fluences on moral judge­ment can high­light po­ten­tial bi­ases in our own judg­ments, and po­ten­tially offer de­bunk­ing ar­gu­ments against our own intuitions

  • Know­ing in­di­vi­d­u­als’ views may be of prac­ti­cal rele­vance to our ac­tions e.g. show­ing which an­i­mals the gen­eral pub­lic are highly sym­pa­thetic to­wards and which they are not, to sug­gest which cam­paigns would or would not likely be successful

  • Such stud­ies can also high­light pre­dic­tors of in­di­vi­d­u­als car­ing more or less about non-hu­man an­i­mals, which can be valuable for similar rea­sons.

This re­cent study by Miralles et al (2019), elic­ited judg­ments from 3500 raters (2347 af­ter ex­clu­sions) about their rel­a­tive lev­els of em­pa­thy or com­pas­sion for an­i­mals of differ­ent species. Em­pa­thy was mea­sured by pre­sent­ing in­di­vi­d­u­als with the state­ment “I feel like I’m bet­ter able to un­der­stand the feel­ings or the emo­tions of...” and a choice of two pic­tures of an­i­mals, and com­pas­sion was mea­sured with “If these two in­di­vi­d­u­als were in dan­ger of death, I will spare the life of [choice among a pair of pic­tures] as a pri­or­ity.”

Both em­pa­thy and com­pas­sion scores de­creased sharply with in­creas­ing phy­lo­ge­netic dis­tance from hu­mans. How­ever, be­yond a cer­tain point scores on both mea­sures stopped de­creas­ing.

The re­sults also high­lighted an­i­mals for which re­spon­dents had rel­a­tively high com­pas­sion com­pared to em­pa­thy or vice versa, al­though these were fairly highly cor­re­lated. For ex­am­ple, re­spon­dents had rel­a­tively high com­pas­sion to­ward (or will­ing­ness to save) a platy­pus rel­a­tive to their lev­els of em­pa­thy for the platy­pus and rel­a­tive to other an­i­mals for which they had similar lev­els of em­pa­thy (such as lizards or al­li­ga­tors). Em­piri­cal data of this kind seems to have clear ad­van­tages over sim­ply rely­ing on our in­tu­itions about which an­i­mals in­di­vi­d­u­als will care about.


The study merely offered par­ti­ci­pants with a forced choice be­tween two differ­ent an­i­mals, and so the analy­ses are based around the prob­a­bil­ity of an­i­mals be­ing cho­sen in such a forced choice. This can­not tell us di­rectly about the ex­tent of com­pas­sion felt for differ­ent an­i­mals, which might vary by sev­eral or­ders of mag­ni­tude. Stud­ies pos­ing ex­plicit trade­offs (e.g. how many an­i­mals of a cer­tain species are worth the same morally as one an­i­mal of an­other species (see Alexan­der (a) (b)), or ask­ing will­ing­ness to pay to save an­i­mals of differ­ent species might be more pow­er­ful for this pur­pose.

The study com­pared em­pa­thy and com­pas­sion scores to evolu­tion­ary di­ver­gence time, but there are a whole host of other fac­tors vary­ing across species which are po­ten­tially rele­vant and may con­found these judg­ments). The study posits that “the phy­lo­ge­net­i­cally closer a species is to us, the more it shares com­mon traits with us. Our re­sults could be ex­plained by the fact that many of these traits may arouse sen­sory bi­ases.” How­ever, other traits, such as neu­ron count (see Alexan­der cita­tions above) likely also cor­re­late with phy­lo­ge­netic dis­tance and with the moral value as­cribed to an­i­mals (al­though Re­think Pri­ori­ties will be pub­lish­ing fur­ther work soon de­scribing how this may be mis­lead­ing). Fur­ther work would be nec­es­sary to in­ves­ti­gate how these judg­ments cor­re­late with fea­tures such as neu­ron count and per­ceived similar­ity to hu­mans. Fu­ture work ei­ther us­ing the paradigm em­ployed in this study or the paradigms em­ployed in Alexan­der’s and Re­think Pri­ori­ties’ stud­ies or some­thing similar could also em­ploy a wider va­ri­ety of species and stim­uli to ex­plore the in­fluences of differ­ent fea­tures on per­ceived moral value.

A sig­nifi­cant po­ten­tial benefit of this study was that it em­ployed pho­tos of an­i­mals of differ­ent species, rather than sim­ply ask­ing in ab­stracta about com­par­i­sons of an­i­mals of differ­ent species. Us­ing vi­sual prompts may pro­duce more ecolog­i­cally valid re­sults and may make visi­ble traits (such as similar­ity to hu­mans) more salient (which may be a pos­i­tive or a nega­tive effect over­all). Us­ing pho­tos of spe­cific an­i­mals from differ­ent species also raises tricky is­sues con­cern­ing stim­u­lus sam­pling. Though the au­thors used 4 pho­tos per species, it is not clear to me that this is an ad­e­quately rep­re­sen­ta­tive sam­ple or that they ad­e­quately ac­counted for the po­ten­tial effect of differ­ent pho­tos in their anal­y­sis.


Over­all I think many EAs would benefit from this pa­per and that more re­search in this area would likely be of value to the EA com­mu­nity.