Table Of ContentF
e
This Element provides an accessible introduction to the n
contemporary philosophy of causation. It introduces the t
o
reader to central concepts and distinctions (type vs token n
causation, probabilistic vs deterministic causation, difference- -
G
making, interventions, overdetermination, pre-emption) and l
y
to key tools (structural equations, graphs, probabilistic causal n Philosophy of Science
n
models) drawn upon in the contemporary debate. The aim is
to fuel the reader’s interest in causation, and to equip them
with the resources to contribute to the debate themselves.
The discussion is historically informed and outward-looking.
‘Historically informed’ in that concise accounts of key historical
contributions to the understanding of causation set the stage
C Causation
for an examination of the latest research. ‘Outward-looking’ a
u
in that illustrations are provided of how the philosophy of s
a
t
causation relates to issues in the sciences, law, and elsewhere. io
n
The aim is to show why the study of causation is of critical
importance, besides being fascinating in its own right.
About the Series Series editors
This series of Elements in Philosophy of Jacob Stegenga
Science provides an extensive overview University of
of the themes, topics and debates which Cambridge Luke Fenton-Glynn
constitute the philosophy of science.
Distinguished specialists provide an
up-to-date summary of the results of
current research on their topics, as well
as offering their own take on those topics
and drawing original conclusions.
Cover image: BSIP / Contributor/
Universal Images Group/Getty Images ISSN 2517-7273 (online)
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press ISSN 2517-7265 (print)
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
Elements in the Philosophy of Science
editedby
Jacob Stegenga
UniversityofCambridge
CAUSATION
Luke Fenton-Glynn
University College London
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
UniversityPrintingHouse,CambridgeCB28BS,UnitedKingdom
OneLibertyPlaza,20thFloor,NewYork,NY10006,USA
477WilliamstownRoad,PortMelbourne,VIC3207,Australia
314–321,3rdFloor,Plot3,SplendorForum,JasolaDistrictCentre,
NewDelhi–110025,India
103PenangRoad,#05–06/07,VisioncrestCommercial,Singapore238467
CambridgeUniversityPressispartoftheUniversityofCambridge.
ItfurtherstheUniversity’smissionbydisseminatingknowledgeinthepursuitof
education,learning,andresearchatthehighestinternationallevelsofexcellence.
www.cambridge.org
Informationonthistitle:www.cambridge.org/9781108706636
DOI:10.1017/9781108588300
©LukeFenton-Glynn2021
Thispublicationisincopyright.Subjecttostatutoryexception
andtotheprovisionsofrelevantcollectivelicensingagreements,
noreproductionofanypartmaytakeplacewithoutthewritten
permissionofCambridgeUniversityPress.
Firstpublished2021
AcataloguerecordforthispublicationisavailablefromtheBritishLibrary.
ISBN978-1-108-70663-6Paperback
ISSN2517-7273(online)
ISSN2517-7265(print)
CambridgeUniversityPresshasnoresponsibilityforthepersistenceoraccuracyof
URLsforexternalorthird-partyinternetwebsitesreferredtointhispublication
anddoesnotguaranteethatanycontentonsuchwebsitesis,orwillremain,
accurateorappropriate.
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
Causation
ElementsinthePhilosophyofScience
DOI:10.1017/9781108588300
Firstpublishedonline:June2021
LukeFenton-Glynn
UniversityCollegeLondon
Authorforcorrespondence:LukeFenton-Glynn,[email protected]
Abstract:ThisElementprovidesanaccessibleintroductiontothe
contemporaryphilosophyofcausation.Itintroducesthereadertocentral
conceptsanddistinctions(typevstokencausation,probabilisticvs
deterministiccausation,difference-making,interventions,
overdetermination,pre-emption)andtokeytools(structuralequations,
graphs,probabilisticcausalmodels)drawnuponinthecontemporary
debate.Theaimistofuelthereader’sinterestincausation,andtoequip
themwiththeresourcestocontributetothedebatethemselves.The
discussionishistoricallyinformedandoutward-looking.‘Historically
informed’inthatconciseaccountsofkeyhistoricalcontributionstothe
understandingofcausationsetthestageforanexaminationofthelatest
research.‘Outward-looking’inthatillustrationsareprovidedofhowthe
philosophyofcausationrelatestoissuesinthesciences,law,and
elsewhere.Theaimistoshowwhythestudyofcausationisofcritical
importance,besidesbeingfascinatinginitsownright.
Keywords:causation,counterfactualtheoriesofcausation,probabilistic
causation,regularitytheoriesofcausation,causalmodelling
©LukeFenton-Glynn2021
ISBNs:9781108706636(PB),9781108588300(OC)
ISSNs:2517-7273(online),2517-7265(print)
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
Contents
1 Introduction 1
2 Regularity Theories of Causation 3
3 Interlude: Some Important Background 13
4 Counterfactual Theories of Causation 29
5 Probabilistic Causation 55
6 Conclusion 82
References 85
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
Causation 1
1 Introduction
We humans take a great interest in causation. Causal knowledge helps us to
understand,predict,andinfluencetheworldaroundus.Ababyquicklycomes
to realise that pushing a button on her toy causes it to play a song; an adult
exploitsherknowledgeoftheeffectsofchamomiletosooththebaby’steething
gums. Because of the close relation between causation, understanding, pre-
diction, and control, natural and social scientists devote significant time and
resources to investigating causal questions. Among other things, they ask or
have asked: What are the causes of cancer? Of climate change? Of anoma-
liesintheorbitofUranus?Whatcausedtheextinctionofthedinosaurs?The
FirstWorldWar?The2007-8financialcrisis?Trump’selection?TheCovid-19
outbreak?Whatcausesmentalhealthproblems?Crime?Priceinflation?
Causation is of importance to psychology: if we want to understand how
humanslearnandreason,weneedtounderstandtheircapacityforcausallearn-
ingandreasoning.It’sofimportanceinAI:ifwewantcomputersandrobots
tolearnaswellas(orbetterthan!)humans,andtomanipulatetheworldas(or
more!) effectively, we need to programme them to be able to acquire causal
knowledgeandtouseit.
Causationisalsocloselytiedtoquestionsofmoralandlegalresponsibility.
InalandmarkUKlegalcase,thewifeofthelateArthurFairchildsuccessfully
sued Glenhaven Funeral Services1 over her husband’s death from mesotheli-
oma–atypeoflungcancercausedbyasbestosexposure.Inordertoestablish
thecompany’sresponsibilityforFairchild’sfatalillness,itwasofcoursenec-
essary to establish that there was a causal link between something they’d
done – namely negligently expose Fairchild to elevated levels of asbestos –
andtheillnessitself.
Despiteitsubiquityandimportance,it’ssurprisinglydifficulttosayexactly
what causation is. Difficult questions about the fundamental nature of the
world–especiallythosethatdon’treadilyadmitofempiricalresolution–nat-
urallyattracttheattentionofphilosophers.Butcausationisn’tonlyofintrinsic
philosophicalinterest.Greatertheoreticalclarityonitsnaturehashadsignifi-
cantpayoffsinthesciencesandinlaw.And,closetohomeforphilosophers,it
haspayoffsinvirtueofthefactthatcausationplaysaroleinkeytheoriesofa
varietyofphilosophicallyinterestingphenomenaincluding(butnotlimitedto)
reference,perception,decision,knowledge,inference,action,andexplanation.
Itshouldn’tbethoughtthatworkonthetheoryofcausationistheexclusive
preserveofphilosophers.Muchimportanttheoreticalworkhasbeendoneby
1 FairchildvGlenhavenFuneralServicesLtd([2002]UKHL22;[2003]1AC32).
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
2 PhilosophyofScience
computerscientists,economists,statisticians,legalscholars,psychologistsand
others – reflecting the broad, interdisciplinary importance of a better under-
standing of causation. Fortunately, these days, there’s significant interaction
betweentheoristsinthesevariousdisciplines,whichhasenrichedourcollective
understanding of causation. A prime example of the payoffs of this cross-
disciplinary interaction is the theory of causal modelling that we’ll examine
aspartofSections4and5.
ThisElementexaminessomeoftheprogressthat’sbeenmadeinunderstand-
ingthenatureofcausationaswellassomeoftheunresolvedchallenges.Since
the literature on causation is large, this Element is, of necessity, a selective
introduction. It focuses on three broad traditions within the theory of causa-
tion: the regularity, counterfactual, and probabilistic approaches. Perhaps the
mostcontentiousomissionsaretheprocessapproach–whichseekstoanalyse
causationintermsofcausalprocesseswiththelatterunderstood,onthemost
promising such account, as the world-lines of objects that possess conserved
quantities(Dowe1992,2000;Salmon1994,1997)–andtheNewMechanist
approach–exemplifiedbyMachameretal.(2000)andGlennan(2017),among
manyothers.I’vemadethischoice,notbecauseIdon’tthinkthatunderstand-
ingcausalprocessesandmechanismsisofvitalimportance(Ido!),butbecause
Idon’tthinkthatthesearecausalbedrock:Ithinktherearerelationsofcausa-
tionthataremorefundamentalthanthenotionofacausalprocessormechanism
andthatanadequateunderstandingofprocessesandmechanismswillrequire
anadequateunderstandingofthesemorefundamentalcausalrelations.2Regu-
larity,counterfactual,andprobabilisticapproachesareattemptstounderstand
thesefundamentalcausalrelations.
This view is contentious as process theorists and some New Mechanists
think that the fundamental causal relation(s) can be understood in terms of
processes or mechanisms. For example, Dowe (2000, 90) seeks to define
causalinteractionsintermsofprocesses,whileGlennan(1996)suggeststhat
causation – at least outside the domain of microphysics – might be analysed
intermsofmechanisms.However,processtheorieshavebeenplaguedbythe
problemofdistinguishinggenuinecausalprocessesfromworld-linesthatdon’t
correspondtoprocesseswithoutfallingbackonanappealtosomemorebasic
causalrelation,perhapsunderstoodintermsofcounterfactualdependence(see,
e.g.,Hitchcock1995,2009;Choi2002).
As regards mechanistic approaches, there’s certainly no consensus among
NewMechaniststhatthenotionofmechanismispriortothatofcausation.As
2 ThenotionthattheremightbemorethanonefundamentalcausalrelationistakenupinSection
3.4.2.
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
Causation 3
CraverandTabery(2019)note,‘Mechanistshavedisagreedwithoneanother
about how to understand the cause in causal mechanism. … Four ways of
unpacking[it]havebeendiscussed:conservedquantityaccounts,mechanistic
accounts,activitiesaccounts,andcounterfactualaccounts’.
I’ve already said that it’s doubtful that the conserved quantity approach
can yield an understanding of the fundamental causal relation(s). The activ-
ities approach, on the other hand, is a primitivist approach (see Craver and
Tabery2019),whereaswe’llbeexaminingaccountsthatseekadeeperunder-
standing of causation. Meanwhile, the counterfactual approach is one that
we’ll be exploring in Section 4. Finally, the mechanistic account – as advo-
cated by Glennan (1996) – is regressive. The proposal is that causal con-
nections that may seem basic at (say) the biological level can be understood
in terms of mechanisms at the chemical level, and those at the chemical
level in terms of mechanisms at the physical level. There’s thus a hierarchy
of mechanisms. The concern, though, is that this hierarchy bottoms out at
the level of fundamental physics at which level we have causings that can’t
be mechanistically understood. Again, this favours the view that there are
fundamental causal relations in terms of which mechanisms can ultimately
be understood. The regularity, counterfactual, and probabilistic approaches
seem the most promising approaches to understanding these basic causal
relations.
2 Regularity Theories of Causation
2.1 Hume
Though Western theorising about causation dates back at least to Aristotle
(Physics 195 a 4–14; Metaphysics V.2), David Hume (1739, 1748) is rightly
considered the father of the modern tradition of attempts to understand that
relation. Hume is standardly interpreted as advocating a regularity theory of
causation.
Specifically,accordingtoHume(1739I.iii.2),causesoccurtemporallyprior
totheireffects,andareeithercontiguouswiththeminspaceandtimeorelse
connectedtothembyacontiguous‘chain’ofcausation.Forexample,aperson
may break a window by throwing a rock at it despite the fact that the throw
occursashortintervaloftimebeforeandafewmetresawayfromthebreaking
because,oncetherockisthrown,ittracesacontinuoustrajectoryuntilithits
and shatters the window. The window’s breaking is caused by the throw via
this‘chain’(there’sno‘actionatadistance’),withthepositionandmomentum
oftherockateachstageonitstrajectorybeingcausedbyitspreviouspositions
andmomentaandbythethrowitself,andwiththewindow’sshatteringbeing
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
4 PhilosophyofScience
causedbythepriorstatesoftherockallthewaybackuntilwegettothethrow
itself.
But we don’t have a case of causation just any time an event occurs prior
to and contiguously with another. Towards the end of the movie Saving Pri-
vateRyan,inadefiantlaststand,CaptainMillerrepeatedlyfireshispistolata
GermanTigertank.Thebulletsareofcoursecompletelyincapableofpiercing
thetank’sarmour.Downtohislastbullet,Millerpointshisgunandshootsat
thetankattheverymomentthetankisblowntopiecesbyabombdroppedby
aUSP-51aircraft.TheimpactofMiller’sbulletisimmediatelypriorto,and
contiguouswith,theexplosionofthetank.Yetit’sthebombandnotthebullet
thatcausesthetanktoexplode.
Fortunately, Hume’s account doesn’t imply that the bullet impact was a
cause. That’s because, in addition to priority and contiguity, Hume adds a
third requirement: constant conjunction. For Hume, for an event c to be a
cause of an event e, it must be the case that events like c are always fol-
lowed contiguously by events like e.3 This criterion excludes the impact of
Miller’sbulletfromcountingasacauseofthetank’sexplosion.That’sbecause
eventsliketheformeraren’talwaysfollowedbyeventslikethelatter.Indeed,
Miller had already fired his gun at the tank five times prior to firing his last
bullet:theimpactofnoneofthesepreviousfivebulletswasfollowedcontig-
uously by an explosion. So Hume’s analysis yields the correct verdict about
thiscase.
AlthoughHumedoesn’tsaythis,it’stemptingtothinkthatnotanyoldcon-
stant conjunction can ground a causal relation, but rather one might wish to
require that the constant conjunction be entailed by the laws of nature. This
avoids problems such as the following. Suppose there exists an extremely
rare isotope, call it ‘unobtanium-352’. Only one atom of this isotope ever
exists. Suppose this atom happened to decay on the afternoon of Novem-
ber 19, 1863, immediately before Lincoln delivered the Gettysburg Address
and contiguously with it. Now, for any type T of event of which the Gettys-
burg Address is a member (‘famous speeches’, say), it’s true that all cases
of unobtanium-352 decay are followed by events of type T. Nevertheless, it
clearlydoesn’tfollowthatthedecayoftheunobtanium-352atomwasacause
of the Gettysburg Address.4 A sophisticated regularity theory can avoid this
3 Wecould,onHume’sbehalf,distinguishdirectfromindirectcausation,withccountingasa
directcauseofeiffcispriorto,andcontiguouswith,eandeventslikecarealwaysfollowed
contiguouslybyeventslikee.Indirectcausationwouldthenbeunderstoodintermsofchains
of(i.e.orderedsequencesofeventsthatstandinrelationsof)directcausation.
4 Notethatevenifwetake‘constantconjunction’torequireamultipleinstances,we’restill
liabletoget‘accidental’constantconjunctionsthataren’tapttounderwritecausalrelations
(seeArmstrong1983,15–17).
https://doi.org/10.1017/9781108588300 Published online by Cambridge University Press
