On alien science

Link post

In his book The Fabric of Real­ity, David Deutsch makes the case that sci­ence is about com­ing up with good and true ex­pla­na­tions, with all other con­sid­er­a­tions be­ing sec­ondary. This clashes with the more con­ven­tional view that the goal of sci­ence is to al­low us to make ac­cu­rate pre­dic­tions—see for ex­am­ple this quote from the No­bel prize-win­ning physi­cist Steven Wein­berg:
“The im­por­tant thing is to be able to make pre­dic­tions about images on the as­tronomers’ pho­to­graphic plates, fre­quen­cies of spec­tral lines, and so on, and it sim­ply doesn’t mat­ter whether we as­cribe these pre­dic­tions to the phys­i­cal effects of grav­i­ta­tional fields on the mo­tion of planets and pho­tons [as in pre-Ein­stei­nian physics] or to a cur­va­ture of space and time.”

It’s true that a key trait of good ex­pla­na­tions is that they can be used to make ac­cu­rate pre­dic­tions, but I think that tak­ing pre­dic­tion to be the point of do­ing sci­ence is mis­guided in a few ways.

Firstly, on a his­tor­i­cal ba­sis, many of the great­est sci­en­tists were clearly aiming for ex­pla­na­tion not pre­dic­tion. Astronomers like Coper­ni­cus and Ke­pler knew what to ex­pect when they looked at the sky, but spent their lives search­ing for the rea­son why it ap­peared that way. Dar­win knew a lot about the rich di­ver­sity of life on earth, but wanted to know how it had come about. Ein­stein was try­ing to rec­on­cile Maxwell’s equa­tions, the Michel­son-Mor­ley ex­per­i­ment, and clas­si­cal me­chan­ics. Pre­dic­tions are of­ten use­ful to ver­ify ex­pla­na­tions, but they’re rarely the main mo­ti­vat­ing force for sci­en­tists. And of­ten they’re not the main rea­son why a the­ory should be ac­cepted, ei­ther. Con­sider three of the great­est the­o­ries of all time: Dar­wi­nian evolu­tion, New­to­nian me­chan­ics and Ein­stei­nian rel­a­tivity. In all three cases, the most com­pel­ling ev­i­dence for them was their abil­ity to cleanly ex­plain ex­ist­ing ob­ser­va­tions that had pre­vi­ously baf­fled sci­en­tists.

We can fur­ther clar­ify the case for ex­pla­na­tion as the end goal of sci­ence by con­sid­er­ing a thought ex­per­i­ment from Deutsch’s book. Sup­pose we had an “ex­per­i­ment or­a­cle” that could pre­dict the re­sult of any ex­per­i­ment, but couldn’t tell us why it would turn out that way. In that case, I think ex­per­i­men­tal sci­ence would prob­a­bly fade away, but the the­o­rists would flour­ish, be­cause it’d be more im­por­tant than ever to figure out what ques­tions to ask! Deutsch’s take on this:
“If we gave it the de­sign of a space­ship, and the de­tails of a pro­posed test flight, it could tell us how the space­ship would perform on such a flight. But it could not de­sign the space­ship for us in the first place. And even if it pre­dicted that the space­ship we had de­signed would ex­plode on take-off, it could not tell us how to pre­vent such an ex­plo­sion. That would still be for us to work out. And be­fore we could work it out, be­fore we could even be­gin to im­prove the de­sign in any way, we should have to un­der­stand, among other things, how the space­ship was sup­posed to work. Only then would we have any chance of dis­cov­er­ing what might cause an ex­plo­sion on take-off. Pre­dic­tion – even perfect, uni­ver­sal pre­dic­tion – is sim­ply no sub­sti­tute for ex­pla­na­tion.”

The ques­tion is now: how does this fo­cus on ex­pla­na­tions tie in to other ideas which are em­pha­sised in sci­ence, like falsifi­a­bil­ity, ex­per­i­men­tal­ism, aca­demic free­dom and peer re­view? I find it use­ful to think of these as­pects of sci­ence less as foun­da­tional episte­molog­i­cal prin­ci­ples, and more as ways to coun­ter­act var­i­ous cog­ni­tive bi­ases which hu­mans pos­sess. In par­tic­u­lar:
  1. We are bi­ased to­wards shar­ing the be­liefs of our in­group mem­bers, and forc­ing our own upon them.

  2. We’re bi­ased to­wards aes­thet­i­cally beau­tiful the­o­ries which are sim­ple and el­e­gant.

  3. Con­fir­ma­tion bias makes us look harder for ev­i­dence which sup­ports than which weighs against our own be­liefs.

  4. Our ob­ser­va­tions are by de­fault filtered through our ex­pec­ta­tions and our mem­o­ries, which makes them un­re­li­able and low-fidelity.

  5. If we dis­cover data which con­tra­dicts our ex­ist­ing the­o­ries, we find it easy to con­fab­u­late new post-hoc ex­pla­na­tions to jus­tify the dis­crep­ancy.

  6. We find it psy­cholog­i­cally very difficult to ac­tu­ally change our minds.


We can see that many key fea­tures of sci­ence coun­ter­act these bi­ases:
  1. Science has a heavy em­pha­sis on aca­demic free­dom to pur­sue one’s own in­ter­ests, which miti­gates pres­sure from other aca­demics. Nul­lius in verba, the motto of the Royal So­ciety (“take no­body’s word for it”) en­courages in­de­pen­dent ver­ifi­ca­tion of oth­ers’ ideas.

  2. Even the most beau­tiful the­o­ries can­not over­rule con­flict­ing em­piri­cal ev­i­dence.

  3. Scien­tists are meant to at­tempt to ex­per­i­men­tally falsify their own the­o­ries, and their at­tempts to do so are judged by their peers. Dou­ble-blind peer re­view al­lows sci­en­tists to feel com­fortable giv­ing harsher crit­i­cisms with­out per­sonal reper­cus­sions.

  4. Scien­tists should aim to col­lect pre­cise and com­plete data about ex­per­i­ments.

  5. Scien­tists should pre-reg­ister their pre­dic­tions about ex­per­i­ments, so that it’s easy to tell when the out­come weighs against a the­ory.

  6. Science has a cul­ture of vi­gor­ous de­bate and crit­i­cism to per­suade peo­ple to change their minds, and norms of ad­mira­tion for those who do so in re­sponse to new ev­i­dence.


But imag­ine an alien species with the op­po­site bi­ases:
  1. They tend to trust the global con­sen­sus, rather than the con­sen­sus of those di­rectly around them.

  2. Their aes­thetic views are bi­ased to­wards the­o­ries which are very data-heavy and ac­count for lots of edge cases.*

  3. When their views di­verge from the global con­sen­sus, they look harder for ev­i­dence to bring them­selves back into line than for ev­i­dence which sup­ports their cur­rent views.

  4. Their nat­u­ral senses and mem­o­ries are pre­cise, un­bi­ased and high-re­s­olu­tion.

  5. When they dis­cover data which con­tra­dicts their the­o­ries, they find it eas­iest to dis­card those the­o­ries rather than re­for­mu­lat­ing them.

  6. They change their minds a lot.


In this alien species, brave icon­o­clasts who pick an un­pop­u­lar view and re­search it ex­ten­sively are much less com­mon than they are amongst hu­mans. Those who try to do so end up fo­cus­ing on mod­els with (metaphor­i­cal or literal) epicy­cles stacked on epicy­cles, rather than the clean math­e­mat­i­cal laws which have ac­tu­ally turned out to be more use­ful for con­cep­tual progress in many do­mains. In for­mu­lat­ing their de­tailed, pedan­tic mod­els, they pay too much at­ten­tion to ex­haus­tively re­play­ing their mem­o­ries of ex­per­i­ments, and not enough to what con­cepts might un­der­lie them. And even if some of them start head­ing in the right di­rec­tion, a few con­trary pieces of ev­i­dence would be enough to turn them back from it—for ex­am­ple, their he­lio­cen­trists might be thrown off track by their in­abil­ity to ob­serve stel­lar par­al­lax. Ac­tu­ally, if you’re not yet per­suaded that this alien world would see lit­tle sci­en­tific progress, you should read my sum­mary of The Sleep­walk­ers. In that ac­count of the early sci­en­tific rev­olu­tion, any of the alien char­ac­ter­is­tics above would have se­ri­ously im­peded key sci­en­tists like Ke­pler, Gal­ileo and oth­ers (ex­cept per­haps the ei­de­tic mem­o­ries).

And so the in­sti­tu­tions which ac­tu­ally end up push­ing for­ward sci­en­tific progress on their world would likely look very differ­ent from the ones which did so on ours. Their Alien Royal So­ciety would en­courage them to form many small groups which ac­tively re­in­forced each other’s idiosyn­cratic views and were re­sis­tant to out­side feed­back. They should train them­selves to seek the­o­ret­i­cal beauty rather than em­piri­cal val­i­da­tion—and ac­tu­ally, they should pay much less at­ten­tion to con­tra­dic­tory ev­i­dence than mem­bers of their species usu­ally do. Even when they’re tempted to change their minds and dis­card a the­ory, they should in­stead re­mind them­selves of how well it post-hoc ex­plains pre­vi­ous data, and put effort into ad­just­ing it to fit the new data, de­spite how un­nat­u­ral do­ing so seems to them. Those who change their minds too of­ten when con­fronted with new ev­i­dence should be de­rided as wishy-washy and un­scien­tific.

Th­ese sci­en­tific norms wouldn’t be enough to to­tally re­verse their bi­ases, any more than our sci­en­tific norms make us re­joice when our pet the­ory is falsified. But in both cases, they serve as nudges to­wards a cen­tral po­si­tion which is less bur­dened by species-con­tin­gent psy­cholog­i­cal is­sues, and bet­ter at dis­cov­er­ing good ex­pla­na­tions.


* Note that this might mean the aliens have differ­ent stan­dards for what qual­ifies as a good ex­pla­na­tion than we do. But I don’t think this makes a big differ­ence. Sup­pose that the el­e­gant and beau­tiful the­ory we are striv­ing for is a small set of sim­ple equa­tions which gov­erns all mo­tion in the so­lar sys­tem, and the el­e­gant and beau­tiful the­ory they are striv­ing for is a de­tailed chart which traces out the cur­rent and fu­ture po­si­tions of all ob­jects in the so­lar sys­tem. It seems un­likely that they could get any­where near the lat­ter with­out us­ing New­to­nian grav­i­ta­tion. So a cir­cu­lar-epicy­cle model of the so­lar sys­tem would be a dead end even by the aliens’ own stan­dards.