Opinion: Estimating Invertebrate Sentience


Between May 2018 and June 2019 Re­think Pri­ori­ties com­pleted a large pro­ject on the sub­ject of in­ver­te­brate sen­tience.[1] We in­ves­ti­gated the best method­ol­ogy to ap­proach the ques­tion, out­lined some philo­soph­i­cal difficul­ties in­her­ent in the pro­ject, de­scribed the fea­tures most rele­vant to in­ver­te­brate sen­tience, com­piled the ex­tant sci­en­tific liter­a­ture on the topic, sum­ma­rized our re­sults, and ul­ti­mately pro­duced an in­ver­te­brate welfare cause pro­file. We are cur­rently in the pro­cess of iden­ti­fy­ing con­crete in­ter­ven­tions to im­prove in­ver­te­brate wellbe­ing, with a re­port on the welfare of man­aged honey bees due out in mid-Novem­ber and a re­port on the welfare of farmed snails near­ing com­ple­tion.

One thing we did not do was pub­lish ex­plicit nu­mer­i­cal es­ti­mates of the prob­a­bil­ity that var­i­ous groups of in­ver­te­brates are sen­tient.

Our team dis­cussed pub­lish­ing such es­ti­mates many times, but these dis­cus­sions gen­er­ated con­sid­er­able in­ter­nal dis­agree­ment. Two mem­bers of the (four per­son) team be­lieved that pub­lish­ing ex­plicit sen­tience es­ti­mates was a bad idea. The other two mem­bers felt that it was a good idea. In the end, we set­tled on the fol­low­ing com­pro­mise: sev­eral months af­ter the com­ple­tion of the in­ver­te­brate sen­tience pro­ject, we would pub­lish an un­offi­cial opinion piece in which each of us could share her/​his own rea­son­ing on the sub­ject and, if so de­sired, her/​his own es­ti­mates.[2]

This post fulfills that com­pro­mise. In it, the four mem­bers of Re­think Pri­ori­ties’ in­ver­te­brates team—Daniela R. Wald­horn, Mar­cus A. Davis, Peter Hur­ford, and Ja­son Schukraft—out­line their views on the value, fea­si­bil­ity, and dan­ger of quan­ti­ta­tive es­ti­mates of in­ver­te­brate sen­tience. Mar­cus and Peter provide nu­mer­i­cal es­ti­mates of sen­tience for each of the taxa we in­ves­ti­gated for our in­ver­te­brate sen­tience pro­ject, Daniela offers a qual­i­ta­tive rank­ing of the same taxa, and Ja­son ar­gues that we are not yet in a po­si­tion to de­liver es­ti­mates that are ac­tion­able and ro­bust enough to out­weigh the (slight but non-neg­ligible) harm that pub­lish­ing such es­ti­mates pre­ma­turely might en­gen­der.

What fol­lows are the per­sonal opinions of in­di­vi­d­ual re­searchers. Offi­cially, Re­think Pri­ori­ties does not have a po­si­tion on the ex­plicit prob­a­bil­ity that var­i­ous in­ver­te­brates are sen­tient.

Daniela R. Waldhorn

1. Vertebrates

There is an am­ple and de­tailed body of em­piri­cal data which jus­tifies be­liev­ing that non-hu­man ver­te­brates are sen­tient. In par­tic­u­lar, there is solid neuro-anatom­i­cal, phys­iolog­i­cal and be­hav­ioral ev­i­dence that ver­te­brates like cows and chick­ens are con­scious. There is also a grow­ing trend to rec­og­nize that these an­i­mals do not only ex­pe­rience phys­i­cal suffer­ing (and plea­sure) but also have emo­tional lives (see e.g. Mar­ino, 2017; Proc­tor et al., 2013).

Based on ex­ist­ing ev­i­dence and the gen­er­al­ized ac­cep­tance of the Cam­bridge Dec­la­ra­tion on Con­scious­ness (2012)[3], my over­all con­clu­sions re­gard­ing the prob­a­bil­ities of con­scious­ness for these an­i­mals are pre­sented as fol­lows:

2. Invertebrates

When we con­sider in­ver­te­brates, the de­bate about whether they are con­scious be­comes much more com­plex. First, it must be con­ceded that the nu­mer­ous in­ver­te­brate species and their di­ver­sity im­pose se­vere con­straints to jus­tifi­able gen­er­al­iza­tions about the pres­ence of con­scious­ness in this group of an­i­mals. Se­cond, the sci­en­tific liter­a­ture about sen­tience in in­ver­te­brates is not only scarce but frag­men­tary–that is to say, the ex­tent to which in­ver­te­brates have been in­ves­ti­gated varies. Thus, there are some par­tic­u­lar species about which there is a com­par­a­tively great deal of knowl­edge (e.g., fruit flies), whereas much less re­search has fo­cused on in­di­vi­d­u­als of other taxa (see our Sum­mary of find­ings Part 1 and Part 2).

The ex­ist­ing gaps in this field of re­search en­tail that we face sig­nifi­cant con­straints when as­sess­ing the prob­a­bil­ity that an in­ver­te­brate taxon is con­scious. In my opinion, the cur­rent state of knowl­edge is not ma­ture enough for any in­for­ma­tive nu­mer­i­cal es­ti­ma­tion of con­scious­ness among in­ver­te­brates. Fur­ther­more, there is a risk that such es­ti­mates lead to an over­sim­plifi­ca­tion of the prob­lem and an un­der­es­ti­ma­tion of the need for fur­ther in­ves­ti­ga­tion. As it was stated in our sum­mary of find­ings, the available ev­i­dence is not enough for de­ter­min­ing with suffi­cient cer­tainty whether most in­ver­te­brate taxa are con­scious or not. Yet it does al­low us to iden­tify cer­tain as­pects as po­ten­tially rele­vant for con­scious­ness and flag cer­tain ar­eas that should be fur­ther in­ves­ti­gated if we want to ar­rive at more ro­bust con­clu­sions.

Nev­er­the­less, our liter­a­ture re­view al­lows us to con­clude, for in­stance, that there is rel­a­tively strong ev­i­dence that oc­to­puses are con­scious. We can even as­sert that oc­to­puses are more likely to be con­scious than other in­ver­te­brates, like earth­worms. Hence, al­though, as stated, this field is not ma­ture enough to make ro­bust es­ti­ma­tions of sen­tience, we can still make some rele­vant com­par­i­sons. There­fore, I pro­pose de­vel­op­ing a frame­work for ex­am­in­ing ex­ist­ing ev­i­dence for or against con­scious­ness in par­tic­u­lar in­ver­te­brate taxa.

To con­tribute to this dis­cus­sion, here I pre­sent a pre­limi­nary and generic at­tempt to such a frame­work. It con­sists of five broad cat­e­gories of an­swers to the ques­tion of whether in­ver­te­brate or­ganisms of a par­tic­u­lar taxon are sen­tient or not. Th­ese cat­e­gories are “very prob­a­bly yes”, “prob­a­bly yes”, “pos­si­bly yes”, “pos­si­bly no”, and “prob­a­bly no”. A sixth cat­e­gory, “very prob­a­bly no” is later in­tro­duced to as­sess con­scious­ness in other non-in­ver­te­brate taxa in­cluded for com­par­i­son pur­poses. Th­ese six cat­e­gories range from high­est to low­est as­signed prob­a­bil­ities of con­scious­ness.

Each cat­e­gory is es­tab­lished by ap­peal to spe­cific crite­ria. Hence, ex­ist­ing ev­i­dence of con­scious­ness of an in­ver­te­brate taxon is as­sessed in light of those crite­ria. The differ­ent taxa have been clas­sified de­pend­ing on the de­gree to which they meet them. Th­ese al­lows for com­par­i­sons be­tween these taxa. Ad­di­tion­ally, this frame­work can be use­ful for as­sess­ing new ev­i­dence that may up­date our views, and for study­ing other in­ver­te­brate taxa as well.

It must be stressed that differ­ences be­tween ad­ja­cent cat­e­gories do not nec­es­sar­ily rep­re­sent equal in­ter­vals in the un­der­ly­ing “scale” which gives rise to these groups. How­ever, since the cat­e­gories are or­dered from higher to lesser prob­a­bil­ity, this frame­work al­lows us to claim, for ex­am­ple, that honey bees are more likely to be con­scious in­di­vi­d­u­als than jel­lyfish. This and similar judge­ments re­flect my in­ter­pre­ta­tion of the available ev­i­dence in light of the se­lected crite­ria. New ev­i­dence may change how cur­rent knowl­edge is un­der­stood and how or­ganisms of a given taxon are clas­sified. There­fore, the fact that an or­ganism falls into a cat­e­gory should not be in­ter­preted as a defini­tive an­swer to that or­ganism’s prob­a­bil­ities of be­ing con­scious.

In what fol­lows, I try —to a limited de­gree— to de­scribe the crite­ria that define each cat­e­gory. I de­vote space to elab­o­rate on such crite­ria be­cause a prac­ti­cal frame­work for cat­e­go­riz­ing our find­ings re­quires clear and defined in­di­ca­tors. Other­wise, if a crite­rion is not cor­rectly speci­fied, what counts or not as ev­i­dence and what the ev­i­dence points to will be sub­ject to the re­searcher’s in­ter­pre­ta­tion. How­ever, please keep in mind that this is a pre­limi­nary work, and I am not an ex­pert. Pre­sum­ably, spe­cial­ized knowl­edge, fur­ther em­piri­cal re­search and the de­vel­op­ment of a fun­da­men­tal the­ory of con­scious­ness may con­tribute to im­prove our con­cep­tual clar­ity and to bet­ter weigh the ev­i­dence we pos­sess about con­scious­ness in non-hu­man in­di­vi­d­u­als.

Fi­nally, the fact that sev­eral or­ganisms fall into the same cat­e­gory does not nec­es­sar­ily en­tail that the ex­ist­ing ev­i­dence for them is similar. There are cases in which our ev­i­dence for some taxa is stronger than for oth­ers, even in the same cat­e­gory. To bet­ter un­der­stand why a spe­cific taxon falls into a given cat­e­gory, see our ‘In­ver­te­brate Sen­tience Table’ and our Sum­mary of find­ings Part 2, where it is fur­ther de­scribed which po­ten­tially con­scious­ness-in­di­cat­ing fea­tures are found in each taxon.

Very prob­a­bly yes

I in­clude in this cat­e­gory taxa that meet the fol­low­ing crite­ria:

  • Given the cur­rent state of our knowl­edge, there is di­rect ev­i­dence that in­di­vi­d­u­als of these taxa ex­hibit fea­tures which, ac­cord­ing to ex­pert agree­ment, seem to be nec­es­sary –al­though not suffi­cient– for con­scious­ness (Bate­son, 1991; Broom, 2013; EFSA, 2005; El­wood, 2011; Fiorito, 1986; Sned­don et al., 2014; Sned­don, 2017). Th­ese fea­tures are:

    • Neu­roanatom­i­cal struc­tures and phys­iolog­i­cal func­tions, such as no­ci­cep­tors or equiv­a­lent struc­tures, cen­tral­ized in­for­ma­tion pro­cess­ing, ver­te­brate mid­brain-like func­tion, and phys­iolog­i­cal re­sponses to no­ci­cep­tion or han­dling. Ad­di­tion­ally, it is ex­pected that con­scious in­di­vi­d­u­als have opi­oid-like re­cep­tors and anal­gesics re­duce their no­ci­cep­tive re­flexes and avoidant be­hav­iors;

    • Be­hav­ioral re­sponses that are po­ten­tial in­di­ca­tors of pain ex­pe­rience, such as defen­sive be­hav­ior or fight­ing back, and mov­ing away from nox­ious stim­uli. Th­ese re­ac­tions seem to take into ac­count a nox­ious stim­u­lus’ in­ten­sity and di­rec­tion. Other ob­served be­hav­iors in­clude pain re­lief learn­ing, and long-term be­hav­ior al­ter­a­tion to avoid a nox­ious stim­u­lus.

  • In sev­eral cases, di­rect ev­i­dence of those fea­tures is in­com­plete. Still, we find an im­por­tant body of in­for­ma­tion about other po­ten­tially con­scious­ness-in­di­cat­ing fea­tures that, broadly, ac­count for the or­ganism’s abil­ity to dis­play com­plex and flex­ible be­hav­iors. Ad­di­tion­ally, other be­hav­ioral in­di­ca­tors sug­gest that these or­ganisms may ex­pe­rience emo­tional states and have cog­ni­tive skills. This crite­rion in­cludes:

    • The in­di­vi­d­ual shows differ­ent re­ac­tions to nox­ious stim­uli, de­pend­ing on ex­oge­nous or en­doge­nous changes (mo­ti­va­tional trade-offs);

    • Proxy in­di­ca­tors of mem­ory are ob­served, such as ‘spa­tial mem­ory’ and ‘long-term be­hav­ior al­ter­a­tion to avoid nox­ious stim­u­lus (24+ hours)’ (about the im­por­tance of mem­ory for as­sess­ing con­scious­ness, see Baars (2003) or Stein et al., (2016));

    • There is ev­i­dence of ad­di­tional forms of learn­ing differ­ent from ha­bit­u­a­tion, sen­si­ti­za­tion and mere as­so­ci­a­tive learn­ing. That is to say, the or­ganism can learn in forms that are more likely to re­quire con­scious­ness, such as op­er­ant con­di­tion­ing with an un­fa­mil­iar ac­tion, ob­ser­va­tional learn­ing and con­tex­tual learn­ing;

    • In­di­vi­d­u­als may dis­play cer­tain bench­marks of cog­ni­tive so­phis­ti­ca­tion, such as tool use[4];

    • As con­scious pain is plau­si­bly cor­re­lated with cer­tain kinds of be­hav­iors, so, too, we ex­pect that var­i­ous emo­tional states are cor­re­lated to an ar­ray of be­hav­ioral in­di­ca­tors. Hence, be­hav­iors that sug­gest differ­ent pos­i­tive and/​or nega­tive mood states are also ob­served in these or­ganisms;

    • Var­i­ous forms of nav­i­ga­tional skills are iden­ti­fied.

  • In ad­di­tion, these or­ganisms also re­spond to sev­eral drugs in a man­ner similar to hu­mans;

  • There may be other be­hav­ioral ev­i­dence of con­scious­ness, such as goal-di­rected be­hav­ior, com­mu­ni­ca­tional be­hav­ior and forms of in­ter­ac­tion that prob­a­bly re­quire con­scious­ness;

  • Con­sis­tent with the above, there is some agree­ment among promi­nent sci­en­tists and philoso­phers that these an­i­mals may be con­scious.

Given the cur­rent ev­i­dence I as­sign a high cre­dence to in­di­vi­d­u­als of the fol­low­ing taxa be­ing sen­tient:

Re­cent re­search has shown that oc­to­puses are highly in­tel­li­gent and po­ten­tially ca­pa­ble of ex­pe­rienc­ing pain. This has led to the in­clu­sion of oc­to­puses in an­i­mal pro­tec­tion leg­is­la­tion in some ju­ris­dic­tions (e.g. in the Euro­pean Union, see Direc­tive 2010/​63/​EU), and their recog­ni­tion as con­scious in­di­vi­d­u­als in the Cam­bridge Dec­la­ra­tion on Con­scious­ness (2012).

Prob­a­bly yes

In these cases, the defined crite­ria are:

  • Although in­com­plete, there is di­rect ev­i­dence that in­di­vi­d­u­als of these taxa ex­hibit fea­tures which, ac­cord­ing to ex­pert agree­ment, seem to be nec­es­sary –al­though not suffi­cient– for con­scious­ness (Bate­son, 1991; Broom, 2013; EFSA, 2005; El­wood, 2011; Fiorito, 1986; Sned­don et al., 2014; Sned­don, 2017) (see the first crite­rion of the ‘very prob­a­bly yes’ cat­e­gory);

  • There is an im­por­tant body of in­for­ma­tion about other po­ten­tially con­scious­ness-in­di­cat­ing fea­tures that, broadly, ac­count for the or­ganism’s abil­ity to dis­play com­plex and flex­ible be­hav­iors. Ad­di­tion­ally, other be­hav­ioral in­di­ca­tors sug­gest that these or­ganisms may ex­pe­rience emo­tional states and have cog­ni­tive skills (see the first crite­rion of the ‘very prob­a­bly yes’ cat­e­gory). How­ever, this sort of ev­i­dence is not as abun­dant as in the pre­vi­ous case (‘very prob­a­bly yes’). Or, for some fea­tures, it is dis­cussed whether a spe­cific be­hav­ior is a con­scious re­ac­tion or an au­to­matic re­sponse (e.g., au­to­tomy or groom­ing, as forms of pro­tec­tive be­hav­ior). Alter­na­tively, it may hap­pen that available ev­i­dence is in­dica­tive of con­scious­ness but it is still rel­a­tively re­cent;

  • In ad­di­tion, these or­ganisms also re­spond to some drugs in a man­ner similar to hu­mans;

  • There may be other be­hav­ioral ev­i­dence of con­scious­ness, such as goal-di­rected be­hav­ior, com­mu­ni­ca­tional be­hav­ior and forms of so­cial in­ter­ac­tion and or­ga­ni­za­tion that prob­a­bly re­quires con­scious­ness;

  • Hence, some promi­nent sci­en­tists and philoso­phers claim that these an­i­mals may be con­scious. How­ever, there does not seem to be as wide­spread an agree­ment on this as that ob­served for the pre­vi­ous cat­e­gory (‘very prob­a­bly yes’).

Given the cur­rent ev­i­dence I con­sid­ered that it is prob­a­ble that in­di­vi­d­u­als of the fol­low­ing taxa are con­scious:

Honey bees dis­play con­sid­er­able learn­ing abil­ities, com­plex so­cial be­havi­ours and even com­mu­ni­ca­tional be­hav­iors. They are prob­a­bly con­scious but we have lit­tle ev­i­dence of the pres­ence of a pain sys­tem, if any. Re­gard­ing de­ca­pod crus­taceans (e.g., crabs, crayfish), re­cent re­search has shown that they are po­ten­tially ca­pa­ble of ex­pe­rienc­ing pain. Ac­cord­ing to the Euro­pean Food Safety Author­ity (2005), these an­i­mals ex­hibit com­plex be­hav­iors and should be legally pro­tected. A similar po­si­tion is held by the Bri­tish Ve­teri­nary As­so­ci­a­tion (2017). Fruit flies, for their part, have been widely used as biolog­i­cal mod­els. New re­search pro­vides com­pel­ling ev­i­dence sug­gest­ing that they not only feel pain, but also ex­pe­rience chronic pain that lasts long af­ter an ini­tial in­jury has healed (Khuong et al., 2019). Be­hav­iorally, how­ever, fruit flies do not seem to ex­hibit such com­plex re­ac­tions as those ob­served in honey bees or de­capods.

Pos­si­bly yes

In these cases:

  • Sev­eral po­ten­tially con­scious­ness-in­di­cat­ing fea­tures are found through these taxa, sug­gest­ing that these in­di­vi­d­u­als may be sen­tient. Most of this ev­i­dence refers to be­hav­ioral ob­ser­va­tions. Th­ese be­hav­ioral in­di­ca­tors in­clude:

    • Re­sponses such as mov­ing away, es­cap­ing and avoidance, that seem to ac­count for nox­ious stim­uli in­ten­sity and di­rec­tion;

    • Proxy in­di­ca­tors of mem­ory, such as ‘spa­tial mem­ory’ and ‘long-term be­hav­ior al­ter­a­tion to avoid nox­ious stim­u­lus (24+ hours)’ (about the im­por­tance of mem­ory for as­sess­ing con­scious­ness, see Baars (2003) or Stein et al., (2016));

    • There is ev­i­dence of ad­di­tional forms of learn­ing differ­ent from ha­bit­u­a­tion, sen­si­ti­za­tion and mere as­so­ci­a­tive learn­ing. That is to say, the or­ganism can learn in forms that are more likely to re­quire con­scious­ness, such as op­er­ant con­di­tion­ing with an un­fa­mil­iar ac­tion, ob­ser­va­tional learn­ing and con­tex­tual learn­ing:

    • The in­di­vi­d­ual shows differ­ent re­ac­tions to nox­ious stim­uli, de­pend­ing on ex­oge­nous or en­doge­nous changes (‘mo­ti­va­tional trade-offs’);

    • There may be other in­di­rect be­hav­ioral ev­i­dence of con­scious­ness, such as nav­i­ga­tional skills or forms of so­cial in­ter­ac­tion and or­ga­ni­za­tion that prob­a­bly re­quire con­scious­ness.

  • How­ever, in these cases, the neu­roanatom­i­cal struc­tures and neu­roanatom­i­cal func­tion­ing re­lated to valenced ex­pe­rience re­main un­clear. In gen­eral, the phys­iol­ogy or struc­tures that seem to cor­re­late with the pres­ence of sen­tience (e.g., no­ci­cep­tors), if they ex­ist, have not been iden­ti­fied yet;

  • Nev­er­the­less, limited ev­i­dence of the neu­ro­phys­iolog­i­cal bases of con­scious­ness should not be con­fused with their ab­sence. Fur­ther stud­ies should as­sess some open ques­tions about the neu­roanatom­i­cal and phys­iolog­i­cal bases of ex­ist­ing be­hav­ioral find­ings.

Hence, the case for the taxa in this cat­e­gory be­ing con­scious is not as strongly sup­ported as in pre­vi­ous taxa. I there­fore as­sign a lower cre­dence to their be­ing con­scious:

So­cial ants, and to a lesser ex­tent, spi­ders (es­pe­cially, jump­ing spi­ders) and cock­roaches, show con­sid­er­able learn­ing abil­ities and com­plex so­cial be­havi­our. The small size of the brain does not nec­es­sar­ily mean poor func­tion­ing, since their nerve cells are very small. How­ever, we have lit­tle ev­i­dence of a pain sys­tem (if any). Of these three taxa, the available ev­i­dence is weaker for cock­roaches.

Pos­si­bly no

  • There is limited ev­i­dence about fea­tures which, ac­cord­ing to ex­pert agree­ment, seem to be nec­es­sary –al­though not suffi­cient– for con­scious­ness. In par­tic­u­lar, some neu­roanatom­i­cal and phys­iolog­i­cal fea­tures may be ob­served (e.g., no­ci­cep­tors). How­ever, there is too lit­tle ev­i­dence for us to pos­i­tive con­clude that these or­ganisms pos­sess a pain sys­tem;

  • Be­hav­iorally, differ­ent no­ci­cep­tive re­sponses may be ob­served, along with ex­pres­sions of as­so­ci­a­tive learn­ing. Nev­er­the­less, the ex­ist­ing ev­i­dence about mo­ti­va­tional trade­offs, mood state be­hav­iors or cog­ni­tive so­phis­ti­ca­tion is highly limited;

  • In gen­eral, there is very limited sci­en­tific ev­i­dence re­lated to the pres­ence of valenced ex­pe­riences in these in­di­vi­d­u­als. Still, there is more pos­i­tive than nega­tive ev­i­dence about the pos­si­bil­ity that these in­di­vi­d­u­als are sen­tient.

Sea hares (Aplysia) are an ex­am­ple of a taxon fal­ling into this cat­e­gory, as per the above crite­ria.

Aplysia are the most ac­tive marine gas­tro­pod mol­luscs. Much re­search has been car­ried out on the ner­vous sys­tem of sea hares and their rel­a­tives. Ev­i­dence of learn­ing and flex­i­bil­ity of be­havi­our is con­sid­er­able but there are also stud­ies show­ing very rigid re­sponses.

Prob­a­bly no

  • There is highly re­stricted ev­i­dence of fea­tures that –ac­cord­ing to ex­pert agree­ment– seem to be nec­es­sary for con­scious­ness. This re­duces in a sig­nifi­cant man­ner the like­li­hood that these an­i­mals are sen­tient;

  • Similarly, in­for­ma­tion about other anatom­i­cal, phys­iolog­i­cal and be­hav­ioral fea­tures is scarce. Other po­ten­tially in­di­rect ev­i­dence of con­scious­ness has not been not ob­served ei­ther;

  • We may not en­counter di­rect nega­tive ev­i­dence against any po­ten­tially con­scious­ness-in­di­cat­ing fea­tures in these or­ganisms, but in­di­rect re­search sug­gests that some fea­tures may not be pre­sent or that there is a low chance of them to be found in these an­i­mals;

  • Broadly, these an­i­mals’ pain sys­tem (if they have any) has not been suffi­ciently stud­ied. And in spite of cer­tain be­hav­ioral ev­i­dence, it is hy­poth­e­sized that ob­served re­ac­tions may be re­flexes.

Ac­cord­ing to these crite­ria, the fol­low­ing taxa fall into this cat­e­gory:

Re­gard­ing earth­worms, there is limited re­search about them, since they are not usu­ally con­sid­ered sen­tient or­ganisms. Most of the ex­ist­ing ev­i­dence points out that their nox­ious-stim­uli-re­lated re­sponses are highly rigid.** **How­ever, of the three taxa listed in this cat­e­gory, cur­rent ev­i­dence is slightly stronger for earth­worms, es­pe­cially be­cause of the rel­a­tive com­plex­ity of their ner­vous sys­tem.

In the case of C. el­e­gans, de­spite be­ing a widely stud­ied an­i­mal, ev­i­dence of their be­ing con­scious is weak and their be­hav­ioral re­ac­tions are sim­ple and highly stereo­typed. Fi­nally, about moon jel­lyfish, al­though they dis­play some nox­ious stim­uli re­ac­tions and have a nerve net that al­lows them to de­tect var­i­ous stim­uli, they do not pos­sess a cen­tral ner­vous sys­tem or other equiv­a­lent struc­ture that cen­tral­izes in­for­ma­tion pro­cess­ing.

3. Other organisms

As ex­plained in a pre­vi­ous post, we in­ves­ti­gated plants (king­dom Plan­tae), prokary­otes and pro­tists to give a sense of how of­ten po­ten­tially con­scious­ness-in­di­cat­ing fea­tures are found in or­ganisms that are widely be­lieved to lack con­scious­ness. Re­gard­ing these three broad taxa**:**

  • There is no solid ev­i­dence that in­di­vi­d­u­als of these taxa ex­hibit fea­tures which, ac­cord­ing to ex­pert agree­ment, seem to be nec­es­sary –al­though not suffi­cient– for con­scious­ness. In par­tic­u­lar:

    • The phys­iol­ogy and struc­tures that seem to cor­re­late in­vari­ably with the pres­ence of sen­tience (e.g., no­ci­cep­tors) are not ob­served. No equiv­a­lent struc­tures or func­tion­ing are found ei­ther;

    • Be­hav­iorally, the or­ganism may dis­play some ba­sic nox­ious stim­uli re­ac­tions (e.g., mov­ing away), but these re­sponses do not seem to ac­count for nox­ious stim­u­lus in­ten­sity and di­rec­tion.

  • Similarly, in­for­ma­tion about other anatom­i­cal, phys­iolog­i­cal and be­hav­ioral fea­tures is not found. Other po­ten­tially in­di­rect ev­i­dence of con­scious­ness is not ob­served ei­ther;

  • Con­sis­tent with the above, there is a gen­eral agree­ment among the sci­en­tific com­mu­nity that these or­ganisms are not con­scious.

Fol­low­ing these crite­ria, plants, prokary­otes and pro­tists are very prob­a­bly not con­scious:

4. Results

The fol­low­ing graph sum­ma­rizes my pre­vi­ous es­ti­ma­tions:

Mar­cus A. Davis

Es­ti­mat­ing the Prob­a­bil­ity of Sentience

Hu­man­ity cur­rently doesn’t un­der­stand what fun­da­men­tally gen­er­ates the abil­ity for hu­mans to ex­pe­rience the world. Even were we to gain such an un­der­stand­ing, un­less we also gather the abil­ity to un­der­stand ev­ery­thing that could plau­si­bly gen­er­ate such ex­pe­riences, as op­posed to merely what does in hu­mans, it will still be difficult to as­cer­tain if other crea­tures who have differ­ent neu­roar­chi­tec­ture than hu­mans are also ex­pe­rienc­ing the world. For these rea­sons and oth­ers, I’m very un­cer­tain about the value of pro­vid­ing quan­ti­ta­tive es­ti­mates for the taxa we’ve stud­ied but still think it’s worth do­ing so on bal­ance. To me, the case for do­ing so is based pri­mar­ily on rea­son­ing trans­parency as op­posed to oth­ers us­ing my es­ti­mates as place­hold­ers in their calcu­la­tions. Given the un­cer­tainty around per­cep­tions of in­ver­te­brate sen­tience, for oth­ers con­sid­er­ing our re­search and ac­tions it’s use­ful to see how this pro­ject af­fected, or did not af­fect, my be­liefs and what those be­liefs are with re­spect to in­ver­te­brates.

How­ever, there are sev­eral other ad­di­tional qual­ifi­ca­tions worth at­tach­ing to my es­ti­mates in­clud­ing the fol­low­ing:

  • I’m by no means an ex­pert on con­scious­ness nor am I an ex­pert on any of the classes of in­ver­te­brates we are con­sid­er­ing.

  • In gen­eral, I don’t be­lieve my es­ti­mates are very likely to be well cal­ibrated, and my best guess isn’t nec­es­sar­ily very sta­ble in all cases. Even rel­a­tive to the difficul­ties of gen­er­ally be­ing well-cal­ibrated, I find my sub­jec­tive es­ti­mates likely aren’t very mean­ingful and I’ve ac­cord­ingly pro­vided rel­a­tively wide es­ti­mates for how prob­a­ble these crea­tures are to be sen­tient. Even with this caveat, the range is still more of a guideline for my sub­jec­tive im­pres­sion than a dec­la­ra­tion of what all agents would es­ti­mate given their en­gage­ment with the liter­a­ture.

  • For many groups of in­ver­te­brates, the amount that is cur­rently un­known sug­gests that any at­tempt to place a par­tic­u­lar num­ber is largely a fore­cast about what fu­ture stud­ies may re­veal as op­posed to re­flec­tive of what cur­rent stud­ies have re­vealed.

  • Given the un­cer­tainty, the ab­solute point es­ti­mate I provide of each taxa’s sen­tience is prob­a­bly less valuable than the rel­a­tive rank­ings of a given taxon com­pare to each other.

  • Ex­trap­o­lat­ing from the taxa we ex­am­ined to other in­ver­te­brate taxa, while tempt­ing, is not ad­vis­able. I set my es­ti­mates for the spe­cific crea­tures un­der con­sid­er­a­tion, and with­out con­sid­er­ing how oth­ers might try to ex­trap­o­late them to other crea­tures, even other or­ganisms within the same genus or fam­ily we con­sid­ered.

  • Th­ese es­ti­mates do not rep­re­sent the moral weight I be­lieve these crea­tures would pos­sess pro­vided they are in fact sen­tient. I’ve not at­tempted to es­ti­mate moral weight for these groups but were I to do so the an­swer could be af­fected by a range of fea­tures in­clud­ing in­tel­li­gence, in­ter­nal clock speed, so­cial com­plex­ity, and af­fec­tive com­plex­ity.

Fi­nally, given these con­sid­er­a­tions, I would add, stick­ing my spe­cific point es­ti­mates (or all of RP’s) di­rectly into a cost-effec­tive­ness anal­y­sis is prob­a­bly foolhardy. Ranges would serve you bet­ter, but even then you re­duce these es­ti­mates to an easy calcu­la­tion used for ex­pected value at your own haz­ard.

How I’ve up­dated since the be­gin­ning of this project

Over­all, I found some ev­i­dence of in­ver­te­brate be­hav­ior rel­a­tively sur­pris­ing, par­tic­u­larly some re­sponses to drugs, cer­tain tool use, and long-term changes due to learn­ing. Though I didn’t cre­ate point es­ti­mates for each of these taxa be­fore be­gin­ning this pro­ject, I sus­pect the com­bi­na­tion of work this area caused me to, at least, up­date pos­i­tively for ants, cock­roaches, hon­ey­bees, fruit flies, and oc­to­puses.

How­ever, per­haps my largest sur­prise wasn’t an up­date to­ward or against a par­tic­u­lar type of an­i­mal, rather it was based on the ex­tent of con­di­tioned learn­ing be­hav­ior that is more or less ex­hibited by all taxa we con­sid­ered, in­clud­ing sin­gle-cel­led or­ganisms and an­i­mal bod­ies de­tached from brain com­mu­ni­ca­tion, in­clud­ing the lower body of a mouse with a sev­ered spine. While one could take this as weak ev­i­dence of wide­spread sen­tience, this up­dated me to­ward think­ing many of these be­hav­iors aren’t very im­pres­sive and they were thus largely dis­re­garded in con­tem­plat­ing the pos­i­tive case for sen­tience.

Fur­ther notes on spe­cific taxa

For chick­ens and cows, my pri­ors about their prob­a­bil­ity of sen­tience re­mained ap­prox­i­mately where it was when we be­gan this pro­ject. This isn’t a sur­prise given they weren’t our fo­cus and I didn’t learn much new in­for­ma­tion about chicken or cow ca­pac­i­ties dur­ing our work.

For the re­main­ing non-in­ver­te­brate taxa—plants, pro­tists, prokary­otes—they pos­sess par­tic­u­larly weak ev­i­dence of sen­tience, and lack cen­tral ner­vous sys­tems and brains. My es­ti­mates re­flect this doubt in their sen­tience. How­ever, one sig­nifi­cant is­sue I en­coun­tered was while they vary in abil­ities and thus there’s a good case a pre­cise and well-cal­ibrated es­ti­mate of sen­tience would vary, gen­er­ally at­tempt­ing to es­ti­mate the prob­a­bil­ity that any of these taxa are sen­tient amounts to an at­tempt to es­ti­mate the prob­a­bil­ity that more or less ev­ery­thing I un­der­stand about sen­tience is in­cor­rect. Are the odds of such an oc­cur­rence one in ten thou­sand? One in one mil­lion? Or, are they one in one hun­dred billion? I don’t have a rele­vant refer­ence class to draw from to an­swer such a ques­tion and thus it would likely be more ac­cu­rate to sur­mise I take the prob­a­bil­ity that any of these groups is sen­tient is van­ish­ingly un­likely and, for all in­tents and pur­poses, ap­prox­i­mately zero.

Peter Hurford

Some of my col­leagues have de­clined to pro­duce es­ti­mates for the like­li­hood of sen­tience for var­i­ous in­ver­te­brates, in­stead sug­gest­ing that the in­ter­ested reader go through our ev­i­dence and cre­ate their own es­ti­mates. I’m some­what sym­pa­thetic to this view—there is a large amount of un­cer­tainty in­volved and judge­ment calls must be made on both per­sonal val­ues and weigh­ing the ev­i­dence we have. Fur­ther­more, keep in mind that yet to be re­solved ques­tions around moral weight, Pas­cal’s mug­ging, and gen­er­al­iz­abil­ity also com­pli­cate how these prob­a­bil­ities will be used in prac­tice. Lastly, I don’t want you to over­weight my opinions or think of me as some sort of ex­pert where I am not. While I may have my per­sonal prob­a­bil­ities, I re­spect that there may still be a wide va­ri­ety of dis­agree­ment on these prob­a­bil­ities, im­plied val­ues, and im­plied views on how to make de­ci­sions based on these. I in­vite peo­ple to come to their own con­clu­sions.

That be­ing said, I still wish to offer my own prob­a­bil­ities. For one, I think we owe the in­ter­ested reader the clear­est pos­si­ble take­aways to do as much work for them and save them time—as long as we provide the ap­pro­pri­ate caveats and warn­ings (you have been warned). Ad­di­tion­ally, I think stat­ing clear prob­a­bil­ities can be a clear way to ex­plain and ex­press my per­sonal wor­ld­view and help make dis­agree­ments more pre­cise, similar to what was done by Muehlhauser (2017), Sec­tion 4.2. I hope that the way we have struc­tured this ex­er­cise—by pre­sent­ing in­di­vi­d­ual opinions that show­case dis­agree­ment among our own team and that go above and be­yond to state the many caveats and un­cer­tain­ties—will be enough to dis­suade you from over­weight­ing our opinions. I be­lieve you de­serve to know where I am per­son­ally com­ing from—more pre­cisely and quan­ti­ta­tively.

Over­all, I be­lieve that many kinds of in­ver­te­brates are a lot more likely to be ca­pa­ble of phe­nom­e­nal con­scious­ness than I thought, given that these taxa are a lot more be­hav­iorally so­phis­ti­cated than I thought, and it be­comes difficult to point to cer­tain so­phis­ti­ca­tions that cows and chick­ens dis­play that hon­ey­bees do not. My gen­eral “case against” pri­mar­ily rests upon what I think is a sub­stan­tial chance for sys­tem­atic bias in the pa­pers we cite to­ward find­ing startling pub­lish­able con­clu­sions, plus think­ing that maybe a fun­da­men­tal the­ory of con­scious­ness will even­tu­ally emerge that leads me to put much more re­duced epistemic rate on be­hav­ioral ev­i­dence, plus some re­main­ing cre­dence to my un­in­formed prior in­tu­ition for in­ver­te­brates are not all that so­phis­ti­cated (e.g., in­tu­itions from phy­lo­ge­netic dis­tance, in­tu­itions that com­plex be­hav­ior may not ac­tu­ally be that com­plex, difficul­ties figur­ing out the com­plex­ity of be­hav­iors).

I should also en­deavor to re­peat that I don’t have any par­tic­u­lar rea­son to think these prob­a­bil­ities are well-cal­ibrated or even all that use­ful, ex­cept as a more de­tailed ex­pres­sion of my cur­rent views and how they changed (in some cases, sig­nifi­cantly). I’m definitely not an ex­pert in think­ing prob­a­bil­is­ti­cally and I’m un­sure of the ex­tent to which these prob­a­bil­ities do cap­ture my views—though I do think they cap­ture my views more clearly than I would ex­press qual­i­ta­tively in words, hence my rea­son­ing for giv­ing them.

While I would like to share my prob­a­bil­ities both be­fore and af­ter start­ing this re­search, I un­for­tu­nately did not ac­tu­ally record prob­a­bil­ities prior to start­ing the re­search. I tried my best to re­con­struct them now as I would’ve thought about them then, to at least provide a rough guide as to how my views have changed.

I think my ex­pla­na­tion for chang­ing my views comes from two sources. First is the large amount of re­search that went into our table—I found my­self par­tic­u­larly weigh­ing cen­tral­ized in­for­ma­tion pro­cess­ing, con­tex­tual learn­ing abil­ities, so­cial learn­ing, long-term be­hav­ior al­ter­a­tion, flex­ible tool use, and play be­hav­ior when at­tribut­ing phe­nom­e­nal con­scious­ness to some or­ganisms and not oth­ers. Also, while I weigh neu­ron count now much less as a fac­tor than I pre­vi­ously did, I still think it is plau­si­ble that the ca­pac­ity for phe­nom­e­nal con­scious­ness may scale to some de­gree with higher neu­ron count.

Se­cond was just gen­er­ally think­ing about the role and func­tion of con­scious­ness for an­other year lead me to be­come more con­fi­dent in cer­tain ar­eas. For ex­am­ple, while I am now more skep­ti­cal of a clear line be­ing es­tab­lished be­tween chick­ens and hon­ey­bees, I am more cer­tain we can es­tab­lish a clear line be­tween hon­ey­bees and plants that ex­cludes the plau­si­bil­ity of plant sen­tience. While look­ing into re­search on plant sen­tience did make me more im­pressed at the ca­pa­bil­ities of plants, it also helped me un­der­stand how much differ­ent they are than the ca­pa­bil­ities of other in­ver­te­brates and how seem­ingly com­plex be­hav­ior can likely re­sult with­out con­scious­ness.

*I added chim­panzees here be­cause they were in Muehlhauser (2017), Sec­tion 4.2 and seem use­ful as a point of com­par­i­son, but they weren’t ac­tu­ally stud­ied in our work.

Ja­son Schukraft

The Issue

What are the odds an oc­to­pus has the ca­pac­ity to ex­pe­rience plea­sure and pain? How much like­lier is it that flies are sen­tient com­pared to earth­worms? Given the available ev­i­dence, what is the ra­tio­nal cre­dence in the propo­si­tion crayfish are phe­nom­e­nally con­scious?

An­swers to ques­tions like these could be ex­tremely use­ful. In­ver­te­brates out­num­ber ver­te­brates by a wide mar­gin, but pop­u­la­tion counts alone aren’t the whole story. If in­ver­te­brates lack the ca­pac­ity for valenced ex­pe­rience—that is, their ex­pe­ri­en­tial states never take on a pos­i­tive or nega­tive af­fect (ei­ther be­cause they don’t have ex­pe­ri­en­tial states or those states are always neu­tral)—then in­ver­te­brates aren’t moral pa­tients and don’t pos­sess in­trin­sic moral worth.[5] If we knew the ra­tio­nal prob­a­bil­ity, con­di­tional on the available ev­i­dence, that differ­ent groups of in­ver­te­brates had the ca­pac­ity for valenced ex­pe­rience, we could mul­ti­ply that prob­a­bil­ity by the (es­ti­mated) num­ber of an­i­mals in each group. That would yield the ex­pected num­ber of an­i­mals for each taxon, and this figure could form the ba­sis for eval­u­at­ing how to di­vide scarce re­sources among differ­ent groups of an­i­mals.[6]

Nonethe­less, I think we shouldn’t (yet) try to an­swer these types of ques­tions.

More pre­cisely, I con­tend that pub­lish­ing spe­cific es­ti­mates of in­ver­te­brate sen­tience (e.g., as­sign­ing each taxon a ‘sen­tience score’) would be, at this stage of in­ves­ti­ga­tion, at best un­helpful and prob­a­bly ac­tively coun­ter­pro­duc­tive. The benefits are slim, and the risks are non-neg­ligible. To be clear: I don’t be­lieve it’s a bad idea to think about prob­a­bil­ities of sen­tience. In fact, any­one di­rectly work­ing on in­ver­te­brate sen­tience ought to be pe­ri­od­i­cally record­ing their own es­ti­mates for var­i­ous groups of an­i­mals so that they can see how their cre­dences change over time. I also be­lieve that we ought to as­pire, as a field, to be able to provide ro­bust, ac­tion­able es­ti­mates of in­ver­te­brate sen­tience. I just don’t think we’re there yet.

The Problem

Un­doubt­edly, the work of Re­think Pri­ori­ties would be more di­gestible if we as­signed a sen­tience score (ei­ther as a straight­for­ward prob­a­bil­ity or as a po­si­tion on an ar­bi­trary scale so that such scores could be com­pared across taxa) for each of the groups of an­i­mals we in­ves­ti­gated. Un­for­tu­nately, as­sign­ing ob­jec­tively good sen­tience scores would be ex­traor­di­nar­ily difficult. The 53 fea­tures we in­ves­ti­gated are not equally im­por­tant, and the con­text in which they are dis­played of­ten makes a sub­stan­tial differ­ence to their ev­i­den­tial weight. One would have to have an ex­pert grasp on biol­ogy, philos­o­phy, and neu­ro­science (as well as lots of time on their hands) to even jus­tifi­ably be­gin such a scor­ing pro­ject. But with­out a clear method­ol­ogy, it’s un­clear how effec­tively the scores could be crit­i­cized or im­proved in the fu­ture. And be­cause sub­jec­tive ex­pe­rience is, well, sub­jec­tive, strict cal­ibra­tion in this do­main is nec­es­sar­ily im­pos­si­ble.

Of course, hav­ing stud­ied the topic for some time now, I ex­pect that my es­ti­mates would be bet­ter than the es­ti­mates of the av­er­age mem­ber of the EA com­mu­nity. If that’s true, then it’s tempt­ing to con­clude that mak­ing my es­ti­mates pub­lic would im­prove the com­mu­nity’s over­all po­si­tion on this topic. How­ever, I think there are at least three rea­sons to be skep­ti­cal of this view.

(1) Sen­tience scores for spe­cific taxa aren’t that useful

Re­think Pri­ori­ties in­ves­ti­gated the ev­i­dence of sen­tience for 18 groups of or­ganisms.[7] After de­cid­ing which types of or­ganisms we wanted to in­ves­ti­gate, we next needed to de­cide the tax­o­nomic level at which we would in­ves­ti­gate those or­ganisms. We tried to drill down to a fairly nar­row taxon (species, genus, or fam­ily) be­cause, in gen­eral, the higher up the tax­o­nomic hi­er­ar­chy one goes, the more di­verse a taxon be­comes. If a taxon be­comes too large, then say­ing that the taxon pos­sesses some fea­ture ceases to be in­for­ma­tive. If 50 differ­ent arthro­pods each pos­sess one (and only one) of the fea­tures we in­ves­ti­gated, and each species pos­sesses a differ­ent fea­ture, a database with the cat­e­gory “arthro­pod” would give the mis­lead­ing im­pres­sion that arthro­pods definitely have the ca­pac­ity for valenced ex­pe­rience.

But when it comes to ac­tual in­ter­ven­tions, some de­gree of gen­er­al­iza­tion is go­ing to be in­evitable. It’s ex­tremely un­likely that a cost-effec­tive in­ter­ven­tion is go­ing to speci­fi­cally tar­get Drosophila melanogaster or Caenorhab­di­tis el­e­gans. It’s much more likely that the in­ter­ven­tion would tar­get some much larger group, like in­sects or ne­ma­todes. So a sen­tience score for Drosophila melanogaster or _Caenorhab­di­tis el­e­gans _wouldn’t ac­tu­ally provide much ac­tion­able in­for­ma­tion with­out some idea whether and to what ex­tent that sen­tience score gen­er­al­izes. But de­vel­op­ing a jus­tified sen­tience score for Drosophila melanogaster or Caenorhab­di­tis el­e­gans is hard enough; do­ing so for groups as large and di­verse as in­sects or ne­ma­todes would re­quire far more data than we gath­ered.[8]

More­over, even well-jus­tified gen­er­al­ized sen­tience scores aren’t su­per use­ful un­less we have at least a rough han­dle on how differ­ent groups of an­i­mals com­pare in moral weight. Know­ing that there is a 25% chance eu­so­cial in­sects are sen­tient and a 95% chance birds are sen­tient doesn’t help one judge which in­ter­ven­tions are most effec­tive un­less one also know ap­prox­i­mately how many in­sect-life-days a bird-life-day is worth.

(2) Sen­tience scores would in­evitably be over-emphasized

Another con­cern is that in­ter­ested par­ties might skip straight to our (un­cal­ibrated, some­what un­jus­tified, ex­tremely spec­u­la­tive) nu­mer­i­cal sen­tience es­ti­mates with­out tak­ing the time to un­der­stand the nu­ance and in­tri­cacy of the is­sue. Num­bers are se­duc­tive be­cause they are so easy to ma­nipu­late (e.g., just stick them into a cost-effec­tive­ness anal­y­sis!). But caveats don’t fit into spread­sheets. Num­bers, even ranges of num­bers, can’t con­vey sub­tlety or com­plex­ity or nu­ance. With­hold­ing ex­plicit sen­tience es­ti­mates forces the reader to take a more holis­tic ap­proach to our work. For an enor­mously com­pli­cated sub­ject like sen­tience, the de­tails re­ally mat­ter. I would much pre­fer a reader to come away from one of our posts with a re­newed ap­pre­ci­a­tion for the com­plex­ity and im­por­tance of the is­sue rather than sen­tience cre­dences that ex­actly match mine.

It’s difficult to pre­sent ex­plicit es­ti­mates of in­ver­te­brate sen­tience in a way in which those es­ti­mates don’t steal the show. It’s hard to imag­ine a third party sum­ma­riz­ing our work (ei­ther to her­self or to oth­ers) with­out men­tion­ing lines like ‘Re­think Pri­ori­ties think there is an X% chance ants have the ca­pac­ity for valenced ex­pe­rience.’ There are very few se­ri­ous es­ti­mates of in­ver­te­brate sen­tience available, so mem­bers of the com­mu­nity might re­ally fas­ten onto ours. But if we did pub­lish our es­ti­mates, I would want them to be viewed as hy­pothe­ses to be fur­ther re­fined (or per­haps com­pletely aban­doned) as more ev­i­dence comes in rather than hard con­clu­sions that our work defini­tively sup­ports. Per­son­ally, I worry that as­sign­ing sen­tience scores sac­ri­fices too much in the name of di­gestibil­ity.

(3) Sen­tience scores might re­duce our cred­i­bil­ity with po­ten­tial collaborators

The fu­ture of the in­ver­te­brate welfare cause area will de­pend in large part on our abil­ity to col­lab­o­rate with biol­o­gists, ethol­o­gists, and neu­ro­scien­tists. There is a great deal of em­piri­cal in­for­ma­tion, from bet­ter pop­u­la­tion ap­praisals to stud­ies on the self-ad­minis­tra­tion of anx­iolytic drugs, that needs to be gath­ered to gen­er­ate an in­formed view of how to best help var­i­ous in­ver­te­brates. But sci­ence, es­pe­cially peer-re­viewed sci­ence, is an in­her­ently con­ser­va­tive en­ter­prise. Scien­tists sim­ply don’t pub­lish things like prob­a­bil­ities of sen­tience. For a long time, even the topic of non­hu­man sen­tience was taboo be­cause it was seen as un­ver­ifi­able. Without a clear, em­piri­cally-val­i­dated method­ol­ogy be­hind them, such es­ti­mates would prob­a­bly not make it into a rep­utable jour­nal. In­tu­itions, even in­tu­itions con­di­tioned by care­ful re­flec­tion, are rarely ad­mit­ted in the court of sci­en­tific opinion.

Re­think Pri­ori­ties is a new, non-aca­demic or­ga­ni­za­tion, and it is part of a move­ment that is—frankly—sort of weird. To col­lab­o­rate with sci­en­tists, we first need to con­vince them that we are a le­gi­t­i­mate re­search out­fit. I don’t want to make that task more challeng­ing by pub­lish­ing es­ti­mates that in­tro­duce the per­cep­tion that our re­search isn’t rigor­ous. And I don’t think that per­cep­tion would be en­tirely un­war­ranted. When­ever I read a post and en­counter an overly pre­cise pre­dic­tion for a com­plex event (e.g., ‘there is a 16% chance Latin Amer­ica will dom­i­nate the plant-based seafood mar­ket by 2025’), I come away with the im­pres­sion that the au­thor doesn’t suffi­ciently ap­pre­ci­ate the com­plex­ity of the forces at play. There may be no sin­gle sub­ject more com­pli­cated than con­scious­ness. I don’t want to re­duce that com­plex­ity to a num­ber.

The Solution

The un­for­tu­nate re­al­ity is that pro­duc­ing use­ful es­ti­mates of in­ver­te­brate sen­tience is go­ing to be in­cred­ibly hard. Nev­er­the­less, the cause is too im­por­tant to give up on. There are just too many in­ver­te­brates to jus­tifi­ably ig­nore. The best and only solu­tion is just to keep work­ing as dili­gently as we can on the is­sue un­til we are in a po­si­tion to de­liver in­formed, ac­tion­able es­ti­mates. And let’s not kid our­selves: we’re not there yet.

But it’s not as if our work makes no progress on the ques­tion. We’ve ar­gued that coleoid cephalopods (i.e., squid, cut­tlefish, and oc­to­puses) are clearly the best can­di­dates for phe­nom­e­nal con­scious­ness in the in­ver­te­brate world. There is also very in­trigu­ing ev­i­dence of sen­tience for de­ca­pod crus­taceans (i.e., prawns, shrimp, crayfish, crabs, and lob­sters) and eu­so­cial in­sects (i.e., ants, ter­mites, bees, and wasps). The ev­i­dence of sen­tience is much greater for these groups than it is for jel­lyfish, earth­worms, or round­worms. There is ba­si­cally no rea­son to think plants, pro­tists, or prokary­otes are sen­tient. It’s still pos­si­ble that they are, but if so, our best philo­soph­i­cal and neu­ro­scien­tific the­o­ries of con­scious­ness are rad­i­cally mis­taken. Th­ese ini­tial claims are well-jus­tified and sup­ported by the available data. They give us a foothold from which to ex­plore other ques­tions, and they offer at least rough guidance on where to fo­cus fur­ther re­search. For now, that’s enough.


This es­say is a pro­ject of Re­think Pri­ori­ties. It was writ­ten by Ja­son Schukraft, Peter Hur­ford, Mar­cus A. Davis, and Daniela R. Wald­horn. Thanks to David Moss and Adrià Voltes for helpful feed­back. If you like our work, please con­sider sub­scribing to our newslet­ter. You can see all our work to date here.


  1. Be­fore Novem­ber 2018 there were only about 10-15 per­son-hours de­voted to this pro­ject per week. Begin­ning in Novem­ber 2018, there were closer to 40-50 per­son-hours per week de­voted to this pro­ject. ↩︎

  2. For ex­am­ples of opinion pieces pub­lished in a high-qual­ity sci­ence jour­nal, see, in­ter alia, “What does AI’s suc­cess play­ing com­plex board games tell brain sci­en­tists?”, “Govern­ing the recre­ational di­men­sion of global fish­eries,” and “Why sci­ence needs philos­o­phy,” all pub­lished in PNAS. ↩︎

  3. The Cam­bridge Dec­la­ra­tion on Con­scious­ness claims that “con­ver­gent ev­i­dence in­di­cates that non-hu­man an­i­mals have the neu­roanatom­i­cal, neu­ro­chem­i­cal, and neu­ro­phys­iolog­i­cal sub­strates of con­scious states along with the ca­pac­ity to ex­hibit in­ten­tional be­hav­iors. Con­se­quently, the weight of ev­i­dence in­di­cates that hu­mans are not unique in pos­sess­ing the neu­rolog­i­cal sub­strates that gen­er­ate con­scious­ness. Non-hu­man an­i­mals, in­clud­ing all mam­mals and birds, and many other crea­tures, in­clud­ing oc­to­puses, also pos­sess these neu­rolog­i­cal sub­strates.” ↩︎

  4. It should be noted that this crite­rion may, in it­self, be the sub­ject of fur­ther dis­cus­sion. Which be­hav­iors ac­count for “cog­ni­tive so­phis­ti­ca­tion” or as­sign­ing a “level of so­phis­ti­ca­tion” to a given ac­tion is a com­plex mat­ter that is be­yond the scope of this ar­ti­cle. Gen­er­ally, it is as­sumed that the use of tools by oc­to­puses is a dis­play of cog­ni­tive so­phis­ti­ca­tion. How­ever, some sci­en­tists re­sist such an im­pli­ca­tion. In par­tic­u­lar, Sch­nell & Clay­ton (2019) ar­gue that, al­though oc­to­puses do use tools (e.g., car­ry­ing co­conut shells around as mo­bile dens), in the ab­sence of con­trol­led ex­per­i­ments, sim­pler ex­pla­na­tions for this be­hav­ior can­not be ruled out. Fur­ther re­search is needed based on other than anec­do­tal ob­ser­va­tions. Hence, the con­clu­sions here pre­sented should be taken with cau­tion. I thank Adrià Voltes for these points. ↩︎

  5. Strictly speak­ing, this ar­gu­men­ta­tive move is too fast, but the de­tails need not de­tain us. ↩︎

  6. Of course, fi­nal de­ci­sions would fac­tor in lots more in­for­ma­tion, such as our best guesses about rel­a­tive moral weight, the tractabil­ity of var­i­ous in­ter­ven­tions, flow-through effects of those ac­tions, and so on. ↩︎

  7. I can’t just say ‘an­i­mals’ be­cause we also in­ves­ti­gated plants, pro­tists, and prokary­otes. ↩︎

  8. More­over, even well-jus­tified gen­er­al­ized sen­tience scores aren’t su­per use­ful un­less we have at least a rough han­dle on how differ­ent groups of an­i­mals com­pare in moral weight. Know­ing that there is a 25% chance eu­so­cial in­sects are sen­tient and a 95% chance birds are sen­tient doesn’t help me judge which in­ter­ven­tions are most effec­tive un­less I also know ap­prox­i­mately how many in­sect-lives a bird-life is worth. ↩︎