A Historical Introduction to the
Philosophy of Science

The following is a summary of the fifteenth chapter of John Losee's book, A Historical Introduction to the Philosophy of Science (fourth edition), with some ancillary notes.

Philosophers discussed in this chapter: Wesley Salmon (1925—); Peter Railton (1950—); Philip Kitcher (1947—)

Salmon's Causal Model

(p. 210) Because the Covering-Law Model and the Deductive-Nomological Pattern (DN) ignore real causal relations, they fail to explain how the correlations in the flagpole and barometer cases are not causally related (see pp. 180–2). Similarly, the Inductive-Statistical Pattern (IS) cannot explain rare cases of leukemia.

Salmon argued that scientific explanations specify causal mechanisms (e.g., effect of gamma rays on cells).

(p. 211) For Salmon, a 'cause' triggers the production and propagation of a structure and is analyzed with reference to 'process', 'intersection' and 'probability'.

A 'process' is the persistence of an entity, quality, or structure (e.g., motion of bodies) and is of two types:

1. 'causal processes' that transmit modifications ('marks')
2. 'pseudo-processes' that do not transmit modifications (e.g., search light swept through red filter)

For Salmon, new structure is produced whenever causal processes intersect such that their modifications persist in time.

This intersection is of two types:

1. 'conjunctive fork' arising from special background conditions
   – a particular effect of the cause does not change the probability of another particular effect from the same cause (e.g., contracting leukemia from atomic bomb blast)
2. 'interactive fork' from direct physical interactions
   – a particular effect of the cause does change the probability of another particular effect from the same cause (e.g., colliding billiard balls) (p. 212)

(pp. 211–12) In a conjunctive fork of two effects from a single cause:

1. the probability (P) of both effects occurring given the cause equals (=) the product (X) of the probabilities (P) of each of the effects occurring given the cause
2. the probability (P) of both effects occurring given the absence of the cause (=) the product (X) of the probabilities (P) of each of the effects occurring given the absence of the cause
3. the probability (P) of the first effect occurring given the cause is greater than (>) the probability (P) of the first effect occurring given the absence of the cause
4. the probability (P) of the second effect occurring given the cause is greater than (>) the probability (P) of the second effect occurring given the absence of the cause

[LA: For a primer on probability theory, see https://www.britannica.com/science/probability-theory]

(p. 212) Reichenbach demonstrated that these four conjunctive fork conditions imply that:

1. 5.
   the probability (P) of both effects occurring is greater than (>) the product (X) of the probabilities (P) of each of the effects occurring

Formula 5. grounds Reichenbach's 'Principle of the Common Cause':
Where the probability of the joint occurrence of effects is greater than would be expected if the two effects are statistically independent, posit a common cause.

Salmon supports this 'Principle of the Common Cause' as a directive principle for scientists looking for conjunctive forks.

In an interactive fork of two effects from a single cause:

1. the probability (P) of both effects occurring given the cause is greater than (>) the product (X) of the probabilities (P) of each of the effects occurring given the cause

An example of 1. is a cue ball colliding with an eight ball with the effect of rebounding at 45 degrees.

Salmon successfully reconciles two distinct approaches to causal relations:

1. singularity view of a process as propagating structure
2. regularity view of individual processes generating modifications of structure

(p. 213) The reconciliation occurs in that conjunctive forks and interactive forks can only be seen as statistical regularities.

Kitcher provided a counterexample to Salmon's analysis of causation (projectile moving between a vehicle and its shadow).

Salmon abandoned his modification/mark-transmission criterion for causal relatedness in the face of Cartwright's counterexample that demonstrated Salmon needed to include counterfactuals in his analysis.

Salmon then adopted Dowe's 'conserved quantity' theory of causation that identified:

• causal process = world line of an object with conserved quantity
• causal interaction = intersection of world lines with exchange of conserved quantity

Conserved quantities = quantities remaining constant within closed systems (e.g., mass-energy, electric charge)

On Dowe's 'conserved quantity' view, the nitrogen/α-particle reaction and the decay of Radium are causal interactions as charge is conserved through the world lines.

(pp. 213–15) However, Salmon accepting Dowe's 'conserved quantity' theory leaves unexplained how the emission of a particular α-particle is 'caused' by a decaying Radium atom and not just the probability of its emission.

Railton's Deductive–Nomological–Probabilistic Model

(p. 215) Railton's probabilistic explanatory model includes three factors:

1. Deductive–Nomological argument for probability of α-particle emission (see p. 214)
2. causal account of underlying mechanism for this probability
3. information about specific emission

Even though the explanation includes that which is to be explained, Railton insisted the account is still explanatory as it explains a highly improbable, non-causal and indeterministic event in terms of:

1. low but finite probability
2. actual atomic decay during that time
3. quantum-mechanical tunnelling

Several philosophers extend non-causal type explanations to non-quantum physics as well.

(p. 216) Such non-causal, non-temporal explanations occur in cases of:

1. static equilibrium and equation of state (e.g., thermodynamics of gases)
2. classification (e.g., why Fido is a dog)
3. evolution (e.g., sexual equilibrium)

Kitcher and Maxwell on Explanatory Unification

Kitcher tipped the question of 'causal relatedness' on its head: We accept 'x causes y' only after we accept an explanation of y in terms of x.

Thus, Kitcher turned our attention to comparing the adequacy of rival explanatory theories. He thought the better explanation was the one that unified previous explanations by:

Thus, Kitcher turned our attention to comparing the adequacy of rival explanatory theories. He thought the better explanation was the one that unified previous explanations by:

1. minimizing the number of patterns of derivation of earlier laws
2. maximizing the number of conclusions generated

(p. 217) For Kitcher, sometimes a principled trade-off needs to be made between satisfying conditions 1. and 2. where they clash.

Maxwell argued that the aim of unification presupposes the universe is comprehensible in terms of invariant and universal laws that cohere.

Einstein's Special Theory of Relativity embodied this unification aim by marrying the seemingly discordant Newtonian Mechanics (dynamics) and Electromagnetic Theory (electrodynamics).

(p. 218) For Maxwell, Einstein's 'aim-oriented empiricism' is more adequate than the 'standard empiricism' that only judged theories by their degree of agreement with observations. That is the lesson from Goodman's paradox (see pp. 184–6) resulting from the generation of ad hoc theories that equally marry with observational data.

Salmon sought to combine Kitcher's unification model with causal models of explanation. The unification model aims at systematizing empirical knowledge while causal models complementarily uncover nature's hidden mechanisms.

Questions to Consider:

1. Is 'causation' about propagation and modification of structure, as Salmon maintains?
2. Is Reichenbach's 'Principle of the Common Cause' a useful directive principle for scientists looking for a cause of statistical events?
3. Do you think Salmon successfully reconciled the two different approaches to analyzing causation?
4. Does Dowe's 'conserved quantity' theory of causation rescue Salmon's approach to indeterminate processes?
5. Must all scientific explanations be causal? What other non-causal explanations can you think of?
6. Is 'aim-oriented empiricism' a substantive improvement on 'standard empiricism'?

Copyright © 2022–3 Leslie Allan