Der Kausalitätsbegriff zwischen Intuition und Wissenschaft

Der Kausalitätsbegriff zwischen Intuition und

Wissenschaft

Max KistlerUniversité Pierre Mendès France,

GrenobleInstitut Jean Nicod, Paris

Philosophical debate on causation

At least five main conceptions presently defended:

1. Deductive-nomological2. Probabilistic3. Counterfactual4. Manipulability/Agency5. Transference

Methodology

Criteria of adequacy?

Aim of a philosophical account of causation?

Debate structured by search for counterexamples.

Counterexamples

Counterexample against an analysis A of causation

= Causal statement S (or the situation described by S) such that:

• S is intuitively true but A predicts S to be false, or

• S is intuitively false but A predicts S to be true.

Agreement with common-sense intuition is a fundamental criterion of adequacy.

Methodology 2

Two other criteria:

1. Overall coherence.

2. Agreement with science.

Coherence

Example of a conflict between intuition and coherence:

Intuition requires that causation be anti-symmetric:

Given two particular events c and e, it is excluded thatc is the (or a) cause of e, ande is the (or a) cause of c.

Do ice cubes cool drinks?

c: a particular melting of an ice cube in a particular glass of water.

e: drop in water temperature.

Intuitively, c causes e: the melting of the ice cube causes the cooling of the water.

The event: the water’s giving off some heat to the ice cube = e, the cooling of the water.

The water’s giving off some of its heat to the ice cube causes the melting of the ice cube, e causes c.

Do ice cubes cool drinks? 2

This set of intuitions:

c causes e, and

e causes c,

said of two particular events, contradicts the intuition that causation is anti-symmetric.

Agreement with science

Imagine Jack Nicholson hitting Leonardo Di Caprio in a movie.

The impact of Nicholson’s fist may seem to cause Di Caprio’s head to swing backwards.

But we know that this is an illusion: none of the events happening on the screen (as opposed to the events they represent) causes any other event on the screen.

Epiphenomena

The images of Nicholson’s fist moving towards Di Caprio’s chin, and of Di Caprios head swinging backwards are epiphenomena (though not in a strict sense).

Epiphenomenon:Event that is caused but causes nothing in its turn.

Epiphenomena 2

The real causal processes:

• Movement of the film in the projector

• Propagation of light rays from projector to screen.

The image of Nicholson’s fist does not cause the image of Di Caprio’s head swinging back.

Two types of philosophical theories of causation

1. Conceptual analysis

• Agreement with science: irrelevant.• Most important: agreement with common

sense.• But, in case a set of convictions turns out to be

incoherent, some intuitions are rejected to restore coherence.

• Ex: one of the judgments on the ice cube, that it cools the water and that the water heats it, must be judged false although both are intuitively correct.

Two types of philosophical theories of causation2. Causation as a natural kind concept

• Causation as a natural kind of relation (or process).

• To conceive of some category as a natural kind, is to conjecture that all items belonging to the category share their essence.

• Gold, water, cats, humans can all be conceived of as natural kinds.

Natural kinds

• Most samples of gold share a set of features by which we recognize them as gold and which play a role in our use of gold: price, specific colour and brilliant aspect, malleability, no rusting, etc.

• But not all samples share all traits: white gold.• What makes it the case that something is gold, is

independent of these superficial features. • Essence of gold : be composed of atoms with atomic

number 79.• Natural kinds have a real essence (Locke) not only a

nominal essence, contrary to such kinds as « tools » or « grass ».

Causation as a natural kind concept

All three criteria equally important:

1. Agreement with common sense intuition.

2. Overall coherence.

3. Agreement with scientific knowledge.

Conflict between intuition and science

• Propagation of the summit of a water wave seems to be a causal process.

• Such a summit can propagate with a speed greater than the speed of light.

• Physics tells us that no real process propagates with speed greater than that of light.

• Therefore, the propagation of a wave summit is not a causal process.

Causation as a natural kind Summing up

• Just as theories of causation aiming at « conceptual analysis », a natural kind theory of causation must reject some types of common sense causal judgments in case of incoherence. (Ice cube case)

• Difference 1: Conceptual analysis theories have no principled reason to reject one or the other of two incompatible causal judgments.

• Agreement with science may at least sometimes force the choice: the water heats the ice cube, not the other way round.

Causation as a natural kind Summing up 2

• Difference 2: Some common sense causal judgments must be rejected although there is no conflict with other common sense judgments. Ex: propagation of wave summit.

• Scientific knowledge permeates common sense. If we know how movies work, we stop judging that Nicholson’s fist, seen in a movie, has any impact on Di Caprio’s head.

• Taking the aim of a theory of causation to be the construction of a natural kind concept makes causation independent of how much science common sense has absorbed.

Four traditional philosophical conceptions of causation

1. Transference account (Fair, Dowe)

Event a is a cause of event b if and only if something is transferred between events a and b.

Transference of a particular amount of energy, or of some other conserved quantity.

Transference account - counterexamples

• Omission (“negative fact” as cause)The father’s not holding back the child made her

have an accident.

• InterruptionThe turning of the switch causes the light’s going

out.

• Release/triggeringReleasing a catch causes a spring to pull a weight.

2. Deductive-nomological analysis (Hempel, Popper)

Causal relation equivalent to explanatory relation.

Scientific explanation = deductive-nomological (D-N) argument.

Premises contain at least one law of nature.

Premises contain « initial conditions » that describe the background situation.

Cause = initial conditions of D-N argument.

Effect = conclusion of D-N argument.

Counterexamples against deductive-nomological analysis

In many D-N explanations, premises and conclusion do not refer to spatio-temporally separated particular events.

1. Explanation of general laws, such as Kepler’s laws from Newton’s laws.

2. Logical and mathematical explanations.

3. Probabilistic analysis (Eells)

A is cause of B

if and only if

the occurrence of A raises the probability of the occurrence of B.

P(B|A) > P(B|A)

Probabilistic analysis - counterexamples

Higher probability of B given A,

at the level of types of events,

is not sufficient for causation of b by a,

at the level of particular events.

John’s smoking may not cause his getting cancer.

Rather than his smoking John’s inhalation of asbestos dust may cause his getting cancer.

Probabilistic analysis – counterexamples 2

Higher probability of B given A,

at the level of types of events,

is not necessary for a’s causing b either.

My firing the rifle may kill the deer although it lowers the probability of the deer’s getting killed.

(in a situation in which I am accompanied by a good hunter who shoots if (and only if) I don’t)

4. Counterfactual analysis (David Lewis

Causation is analysed in terms of causal dependence between events.

Causal dependence is analysed in terms of counterfactual dependence.

“Event e depends causally on event c” Analysed as:If c had not occurred, e would not have occurred.

Counterfactual analysis - counterexamples

1. Two effects of a common cause depend counterfactually on each other.

Smith and Jones watch all (and only) soccer matches broadcast by TV XY. Both have seen the match on day D.

Counterfactual analysis wrongly predicts that Smith’s watching the match causes Jones’ watching it.

If Smith had not watched it (because XY didn’t show it), Jones would not have watched it either.

2. Back-up cause

Hunting situation: it is not true that, if I hadn’t shot, the deer would not have got shot – it would have got shot by the hunter.

Hence, the counterfactual analysis denies, wrongly, that my shooting causes the deer’s getting shot.

Manipulability (Woodward)

• “A very central part of the commonsense notion of cause is precisely that causes are potential handles or devices for bringing about effects…. This idea … is also widely endorsed by social scientists and statisticians…” Woodward 2003, p. 12)

• “X is a total cause of Y if and only if there is a possible intervention on X that will change Y or the proability distribution of Y.” (Woodward 2003, p. 51)

• Problems: anthropocentric, circular.• Manipulability not necessary: some causal processes

outside the reach of human intervention (nuclear fusion in the sun).

• Manipulability not sufficient: one can manipulate the whole by manipulating the parts. This shows synchronic, non-causal dependence.

Two kinds of intuition about causation

• Many counterexamples to the D-N, probabilistic and counterfactual accounts rely on the intuition that transference of energy (or other conserved quantities) is necessary or sufficient for causation.

• Ex: common cause and back-up cause.

• Transference analysis can account for these intuitions.

Two kinds of intuition about causation 2

• Many counterexamples to the transference account rely on the intuition that nomic dependency is sufficient for causation.

• Theories that reduce causation to nomic dependency (D-N analysis) or to relations closely related to nomic dependency, can account for these intuitions.

Two kinds of causal statementsCommon sense example

Two kinds of statements about Mary’s causing my surprise.

1. Eventive statements:The performing of the song surprised me.The performance of the song surprised me.

2. Factive statements:That she performed the song surprised me.The fact that she performed the song surprised me.

Causal statements of two types

1. Eventive causal statements express causation between events.

The argument places of the causal verb are filled by expressions designating events

2. Factive causal statements express causal responsibility of one fact for another fact.

The argument places of the causal verb are filled by expressions designating facts.

Semantic difference between factive and eventive causal statements

• Imagine that Mary uses to perform the song at home but that she is afraid of doing it before a large audience. Now I learn that she performed the song on stage.

• In this situation, factive statements (1) are false, for it is not the fact that she performed the song, but the particular circumstances under which she did it, that are responsible for my surprise.

• Statements (2) are true, for there is something about her performance that causes my surprise.

Factive statements often entail eventive statements.

Example:(1) That she performed the song surprised me.

entails :

(2) The performing of the song surprised me.

But the opposite is not true. (4) may be true because of some unusual aspect of the performance, while the mere fact that she performs would not have been surprising.

Explanation why factual statements may entail eventive causal statements

Simplest case: F1 attributes property G to event c; F2 attributes property H to event e.

Factive statement CR (Gc, He): The fact that c is G is causally responsible for the fact that e is H.

Analysis of the meaning of CR (Gc, He) : a conjunction that contains C(c,e) as one conjunct.

CR (Gc, He) means: C(c,e) Gc He L(G,H).

Thus, from CR (Gc, He) on can infer:

c causes e.

But eventive statements never entail factive statements.

An eventive causal statement C(c,e) never contains information about which fact about c is causally responsible for which fact about e.

Two kinds of causal statementsScientific example

• The fact that pressure reached critical value Pc caused the explosion.

• The event at which pressure reached critical value Pc caused the explosion.

entails does not entail

• The event at which P= Pc is also the event at which T= Tc and there are situations in which the fact that T= Tc, not the fact that P= Pc is causally responsible for the explosion.

Importance of factive causal statements: Reply to objection from “lack of ambition”

Objection: Transference account “suffers from lack of ambition, twice over:First, however successful they are at limning the features of causation, as it relates events in the purely microphysical domain, there seems little hope that they can succeed in doing so in the messy macroscopic realm. Suzy kisses Billy, causing him to flush; are we to suppose that the causal relation between these two events is to be mapped out by looking at how energy, or some other fundamental physical quantity, is transferred? Well, it might be romantic to say so. But it is not, we think, particularly enlightening.” (Collins et al. 2004, p. 14)

Second objection by Collins et al.

“Second, assuming, as seems reasonable, that it is a contingent matter what the fundamental laws are, physical connection accounts cannot (because they are not designed to) tell us anything about causation as it might have been – in particular, as it is in worlds with laws very different from our own. That limitation seems not merely unfortunate but deeply misguided; for it seems that although the fundamental laws play an essential role in fixing the causal facts, they do not do so in so specialized a manner. Put another way, it seems both reasonable and and worthwile to abstract away from the gory details of those laws – thereby to produce an account that has a hope of proving to be not merely true but necessarily so.” (Collins et al., p. 14)

Importance of factive causal statements: Argument against Davidson

• Davidson (1980): eventive causal statements make reference to causal relations between events, but factive causal statements do not make reference to causal relations. They are “rudimentary causal explanations”, and “explanations typically relate statements, not events”.

• Our hypothesis that factive statements make reference to a complex relation that contains the causal relation between events as a part, accounts for the fact that some factive causal statements entail an eventive causal statement.

• Causal explanations normally work by pointing to relations of causal responsibility (expressed by factive causal statements).

Two types of causal statement and two kinds of theories of causation:

1: Truth-maker of eventive causal statements

• Eventive causal statements express the mechanistic (transference) aspect of causation. Their truth-maker is a transmission relation between events.

• The eventive statement does not say which fact about the cause event is responsible for which fact about the effect event.

2. Truth-maker of factive causal statements

Simplest case: relation of causal responsibility of a fact bearing on the cause event for a fact bearing on the effect event.

Counterexamples against transference account rely on factive causal statements that

• either do not mention the event that is the cause in the transmission sense

• or do not mention any causally relevant property. Important type: statements with a negative

expression for the cause.

Objection (Horgan, Mellor, Zucchi):Statements of omission (among others) show that causes

and effects are facts, not events.

(1) “John did not die because he did not fall.”Mellor: There is no such event as John’s non-dying

for it would be both quick and slow.

If there was such an event, Davidson’s analysis of “Don does not die quickly” would entail that the event of Don’s non-dying was quick.

But the analysis of “Don does not die slowly” would entail that the event of Don’s non-dying was slow.

Reply: First category of omission statements

The facts referred to as cause and effect bear on events causally related by transmission.

However, the factive expressions referring to cause and effect do not mention the properties taking part in causal responsibility.

If interpreted as a causal statement, the meaning of (1) is:Is causally responsible for (Not fall (John at t), Not die (John at t+dt)).

Cause event: time slice of John at t.Effect event: time slice of John at t+dt.

Guess of properties taking part in causal responsibility:

Is causally responsible for (Hangs on (John at t), Lives (John at t+dt)).

Scientific example

Operon theory of gene expression

(2) The fact that the repressor is not fixed on the operator causes the transcription of the gene.

Although not explicitly mentioned by (2), there is a “positive” property of the event implying the operator at the relevant moment before the beginning of transcription, which plays a role in causal responsibility: The conformation of the operator molecule when it is not fixed to the repressor.

The fact that the operator has that conformation is causally responsible for the fact that the protein is synthesized.

Second category of omission statements

(3) The father’s not holding back the child made her have an accident.

• In statements of this category, the expression designating the cause contains neither information about the cause event in the transmission sense nor about causally responsible facts.

• Such statements are not causal in a narrow sense.

Scientific example

(4) The fact that a mouse lacks the fosB gene causes her offspring to die.

• Mice are genetically modified by “knock-out” technique.

• Mice lacking fosB gene neglect their offspring.

Role of negative causal statements

The cause can usually be identified by knowledge of the context.Mentioning a negative fact as the cause suggests that the situation is

abnormal and contrasts it with normal situations.

1. Abnormal with respect to moral obligation (case of the father who does not hold back his child).

2. Abnormal in a statistical sense and/or in the sense of adaptation or biological fitness (case of absent fosB gene in mice).

• Such statements provide information about counterfactual causal processes.

• If gene fosB had been present and operational, the mouse would have taken care of her offspring.

• The negative causal statement presents the real situation as exceptional with respect to a normal situation where the expression of the fosB gene leads to normal maternal care.

When “cause” does not refer to a causal relation: Interruption

(5) The turning of the switch causes the light’s going out.

c: turning of the switche: opening of the electrical circuiti : initial state : light onf : final state : light out

c causes e

i f

nomic dependencet

(no causal link)

When “cause” does not refer to a causal relation: Releasing/triggering

(6) Releasing a catch causes a spring to pull a weight.

c : action causing the releasing of the catche : the catch getting releasedi : initial state : spring compressedf : final state (process) : spring expanding

c causes e

i f

nomic dependencet

(no causal link)

Conclusion

Different types of causal statements express different aspects of causation.

A philosophical analysis of causation can account for statements of all these kinds and for scientific knowledge, by conceiving of causation as a natural kind of relation or process.

Eventive statements express causality between events. What makes true eventive causal statements (and the eventive part of factive causal statements) is transmission.

Conclusion 2Several types of factive causal statements:1. Express causal responsibility of a fact about the cause event for a

fact about the effect event. Causal responsibility is a complex relation that includes transmission and nomic dependency between properties.

2. The factive cause expression bears on the cause event but does not mention the causally relevant property.

3. Other factive causal statements are causal only in a derivative sense. The factive cause expression does not bear on any event causing the effect, but provides other useful information. It may inform about counterfactual causal relations (“normal” situations with which the actual situation is contrasted), though the facts explicitly referred to are not causally related.

4. Still other factive causal statements (triggering, interruption, prevention) are made true by more complex patterns of states of affairs, in which causation and nomic dependency play different roles.

References• Collins, John, Hall, Ned, and Paul, L.A. (2004), Counterfactuals and Causation:

History, Problems, and Prospects, in Causation and Counterfactuals, Cambridge (Mass.), MIT Press, p. 1-57.

• Kistler, Max (1998), Reducing Causality to Transmission, Erkenntnis 48 (1998), p. 1-24.

• Kistler, Max (1999), Causes as events and facts, Dialectica 53, p. 25-46.• Kistler, Max (2001), Causation as transference and responsibility, in Wolfgang Spohn,

Marion Ledwig and Michael Esfeld (eds.), Current Issues in Causation, Paderborn, Mentis, p. 115-133.

• Kistler, Max (2002a), Erklärung und Kausalität, philosophia naturalis, 39, p. 89-109.• Kistler, Max (2002b), Causation in Contemporary Analytical Philosophy, Quaestio-

Annuario di storia della metafisica, vol. 2, C. Esposito and P. Porro (eds.), Turnhout (Belgium), Brepols, 2002, p. 635-668.

• Kistler, Max (2006), Causation and Laws of Nature, London, Routledge, 2006 (English version of Causalité et lois de la nature, Paris, Vrin, 1999).

• Kistler, Max (2006a), La causalité comme transfert et dépendance nomique, Philosophie 89, p. 53-77.

• Woodward, James (2003), Making Things Happen, Oxford, Oxford University Press.