A tale of three centuries
UNIVERSITY OF NOTRE DAME,
THE REILLY CENTER FOR VALUES, SCIENCE AND
INDIANA UNIVERSITY - PURDUE UNIVERSITY, FORT WAYNE
CELFIS, University of Bucharest, May 2014
What I am concerned with?
the interplay between natural science and philosophy (both
metaphysics and epistemology)
In metaphysics, I argue for a form of scientism, called “scientific
metaphysics”: inspired by the practice and advancement of
How do we build modalities from a scientific theory?
This is a new naturalism for metaphysics (esp. analytic metaphysics)
In this present case I argue that scientific epistemology guides
Modality and science in general
Standard answer: science describes “reality as actuality” and not “as possibility”
Hence, a “division of labor” between science and metaphysics
Science is all about actuality, philosophy is about possible or impossible entities
The non-standard answer: natural science (at least) is involved with modalities
B. van Fraassen: the modeling process is a process of dealing with possible outcomes of
Any model is un to its neck in modalities
Quantum mechanics is up to its neck in modality esp. many worlds interpretations of QM
(see: D. Wallace (2012), Saunders, Barrett, Kent Wallace (eds.) 2011
Ditto about cosmology and multiverse (Tegmark 2013)
String dualities are “modally involved” (Muntean 2014)
Even biology adds a new type of modality
Statistical mechanics relates to modality (viz. ensembles)
The involvement of classical mechanics with modality is rarely discussed in philosophy
of science. Exceptions: Stoltzner, Butterfield, Pulte.
Minimization and philosophy
What has been done?
relate minimization (and related concepts) to the work of David
Lewis (Butterfield 2002, 2004)
Relate minimization to simplicity and unification in physics (cf. M.
Planck, D. Hilbert in the 1900s)
The anti-metaphysics of the minimization of action and logical
empiricism (esp. Ph. Frank, M Schlick)
Historical approach to the minimization of action (Pulte 1989,
Relate minimization to dispositions (Katzav 2008)
Integrate better minimization with optimality and computational
This is how science is practice nowadays.
This will relate better epistemology to modality (in this context!)
Relate the CM minimization to the standard approaches in QM and
probably to QFT
Relate it to statistical mechanics
Metaphysical voltages of
“the perfectly acting being… can be compared to a clever engineer
who obtains his effect in the simplest manner one can choose”
Nature has “the greatest simplicity in its premises and the greatest
wealth in its phenomena” (Leibniz)
of particle mechanics with wave optics (Hamilton)
A gate to modality (in conjunction with D. Lewis’ work on
Physical voltage of minimization
With PLA mechanics is freed of forces and becomes “non-
Forces are replaced with “virtual motions” and “virtual work” in statics
Universality and generality
One can infer conservation of energy of different quantities and
Can be used in conjunction with symmetries
One can infer Newton’s equations of motion from minimization
Hence a leap towards special relativity is possible through
What makes the actual special?
At least in one interpretation, minimization involves a calculation
over an infinite number of possible evolutions.
Ostwald 1893: “from an infinite number of possibilities of one
process, the one that actually happens is distinguished among the
Deflationism about minimization
It is a convenient alternative to standard mechanics, but with no
It is just a mathematical tool with no philosophical implications.
It is more or less like a clever coordinate transformation, or a Fourier
transformation or a Laplace transformation, with no metaphysical or
Mach: as a mathematical principle, “it is an economically ordered
experience of counting”
Butterfield mentions a eliminativism attitude towards modality in
The tale of the 18th century
Huygens used Fermat’s principle to infer the laws of optics
Euler and Maupertuis used the principle of least action to infer the
laws of collision of bodies
“Nature makes some quantities a minimum or a maximum”
Euler (1743): “all processes in nature obey certain min or max laws.
A system moves from one configuration to another such that the
variation of the integral between the actual and the virtual path,
co-terminus in space and time, is zero
Hamilton’s principle (in Butterfield)
For any one-parameter family, parametrized by say α, of
kinematically possible histories of the mechanical system, that may
deviate from the actual history between t0 and t1, but must match
the actual history as regards the configurations q0, q1 at times t0, t1:
the action as a function of:
with the integral taken along the history labelled by parameter-value
, has zero gradient at the value of corresponding to the
19th century and the Hamilton
The H-J equation
H q t
( , , )
Philosophical stances towards
minimization in the 20th century
Realism about minimization and action: Planck, Hilbert. Minimization
points to a deeper structure in nature:
Pure heuristics: E. Mach, Ph. Frank. It is an useful alternative to PDE
The metaphysics of minimization
Butterfield postulates three levels of involvement with modality:
1: keep the problem (i.e. the L) keep the laws of motion, but change
2: alter the problem (change the L), alter the initial condition but
keep the laws
3: alter the laws of nature.
Metaphysics of Modality 1
Butterlfield thinks that we can identify a world with either:
a configuration (qi,t)
a state (qi,pi,t)
A statement defended by Butterfield: “any actually true proposition
(not only: any law of nature) should be made true by actual facts,
i.e. goings-on in the actual world. (So the threat does not depend
on the evolutions mentioned by the law being contralegal: what
matters is that they are not actual.)”
Three strategies about modality in
(Vindicate): The role of possible (indeed, contralegal) histories can
be vindicated—it is not problematic
(Eliminate): This role of possible histories can be eliminated—the laws
can be formulated without invoking it;
(Useful): The variational formulation of the laws is nevertheless useful,
or even advantageous compared with formulations that do not
mention possible histories.
What is missing?
Here I am interested in the epistemology associated with
minimization and optimization
Minimization is a search procedure
The practice of science is conducive to its metaphysics and its
Analytical mechanics is a scheme for solving problems; therefore, a
I concur with Butterfield: one can interpret Jacobi-Hamilton
formalism metaphysically: you can read off some metaphysics from
the formalism, but not that much
Problem solving and modality
Butterfield 2004: “aspects to do with problem-solving (the use of
separation of variables, leading on to action-angle variables and
Liouville’s theorem) […] are not illuminating about modality”.
I dissent: epistemology and metaphysics work hand in hand.
The forth strategy: the computational
turn of the 20th century
We need to think more pragmatically about minimization
Minimization replaced with optimization, in a computational was
Enters the computational method
We need to remember the practice of current analytic mechanics
which is deeply engaged with numerical simulations.
Minimization is replaced with optimization. The absolute minima of
functions is rarely reached by numerical methods
So we discover minima, but good enough minima
Scientific reasoning and
two types of reasoning in science:
rule-based reasoning RBR: new theories or models are inferred based on
a set of rules
Case-based reasoning CBR: exemplars used to solve problems in
science (Kuhn, Nickles)
Kuhn argued that CBR is more frequently used in science than RBR
CBR computation applied to science (Bod 2006)
CBR and scientific reasoning
“scientists explain new phenomena by maximizing derivational
similarity between the new phenomenon and previously derived
phenomena. And the shortest derivation provides a possible way to
attain this goal. The rationale behind maximizing derivational
similarity is that it favors derivation trees which maximally overlap
with previous derivation trees, such that only minimal recourse to
additional derivational steps needs to be made.”
what is followed is previous patterns of derivations, not phenomena
Reminiscent of Kitcher’s unificatory explanation.
Possibly of consilience too (computational consilience?)
Scientific discovery and
This is my contribution
A blatant counterexample to computational deflationism, and a
negative answer to A1 is evolutionary computation.
Formalist and biomimetic strategies
A. computation and reason/logic
B. computation and mind
C. computation and life
Witness that B and C are both biomimetic strategies. A is more or less
a formalist strategy
The real novelty for a philosophy of science is, I argue, analyzing C
C: Computation and life
“the oldest and the most fundamental of all questions about
simulation” (Von Neumann, Keller 2003):
Q2: How closely can a mechanical simulacrum be made to
resemble an organism?
St. Ulam (Monte Carlo method): the right question when relating
mathematics and computer science to biology is not: “What
mathematics can do for biology?”, but:
“What biology can do for mathematics” (Ulam 1972).
It’s Turing Year!
1950: “Computing Machinery and Intelligence”:
“We cannot expect to find a good child-machine at the first attempt.
One must experiment with searching one such machine and see how
well it learns. One can then try another and see if it is better or worse.
There is an obvious connection between this process and evolution, by
Structure of the child machine = Hereditary material
Changes of the child machine = Mutations
Natural selection = Judgment of the experimenter”
G. Chaitin (the father of the algorithmic information theory):
Metabiology: a field parallel to biology that studies the random
evolution of artificial software (computer programs) rather than
natural software (DNA), and that is sufficiently simple to permit
rigorous proofs or at least heuristic arguments as convincing as those
that are employed in theoretical physics. (my emphasis). (Chaitin
The biomimetic assumption
life = search for optimality
computation = search for optimality
In the 1930s S. Wright interpreted a biological species as a system
that evolves in time by exploring a multi-peaked landscape heuristic
of optimal solutions to a “fitness problem” (Wright 1932).
GA literature: (Tomassini 1995; Koza et al. 1999, 20sqq.; Koza et al.
2003; Affenzeller 2009; Olariu and Zomaya 2006).
More foundational approach (Jong 2006)
Start with a number of algorithms (individuals), mostly randomly
chosen = initial population
Let them live in an environment (let thme output solutions to a
Generate new algorithms by:
Reproduction (not perfect though!)
Define a fitness function
Decide a termination condition
The genetic algorithm
produce an initial population of individuals
while `termination condition’ not met do
evaluate the fitness of all individuals
select fitter individuals for reproduction
produce new individuals
generate a new population (by inserting some new ‘good’ individuals and by discarding some
‘bad’ individuals) (6)
mutate some individuals
Call the individual(s) who satisfy the `termination condition’ the “best-fit-so-far” (9)
Humans are here!
Pick the initial population
Pick the reproduction mechanisms: the mutation factor, the
Pick the termination condition
Pick the resources available
Eureqa=GA in science
Concrete results by Schmidt and Lipson who “re-discovered” not
only analytical functions from empirical data, but structures which
are highly relevant to physical sciences:
Hamiltonians, Lagrangians, laws of conservation, symmetries, and
other invariants (Schmidt and Lipson 2009)
“Optimal forms” and meaningful invariants for the chaotic double
Schmidt&Lipson’s Termination condition: the decomposability
(similar to mechanisms)
An epistemic claim
“These terms may make up an ‘emergent alphabet’ for describing a
range of systems, which could accelerate their modeling and simplify
their conceptual understanding. […] The concise analytical expressions
that we found are amenable to human interpretation and help to
reveal the physics underlying the observed phenomenon. Many
applications exist for this approach, in fields ranging from systems
biology to cosmology, where theoretical gaps exist despite
abundance in data. Might this process diminish the role of future
scientists? Quite the contrary: Scientists may use processes such as this
to help focus on interesting phenomena more rapidly and to interpret
their meaning” (Schmidt and Lipson 2009, 82)
Epistemology of Evolutionary
Rationality and Evolution
A new solution to the Meno’s Paradox?
Genetic algorithms and Numerical
They are not Numerical Simulations per se.
But they simulate the search process in science
They are a combination of:
They are epistemic enhancers with an upward epistemology
Rationality and Evolution
GA are random
GA are inscrutable
Do they provide justification?
Are they irrational?
Can truth be reached by a set of stochastic procedures?
In principle, it can.
See the connection between rationality and evolution
(a) evolution produces individuals which are good approximations to an
optimally well designed system and
(b) optimally well-designed systems are rational agents
Philosophical finale: Meno’s
Meno: How will you look for it, Socrates, when you do not know at all
what it is? How will you aim to search for something you do not
know at all? If you should meet with it, how will you know that this is
the thing you did not know? Socrates: […] Do you realize what a
debater’s argument you are bringing up, that a man cannot search
either for what he knows or for what he does not know? He cannot
search for what he knows since he knows it, there is no need to
search—nor for what he does not know, for he does not know what
to look for. Meno, 80d4-e5, G.M.A. Grube in (Plato 1997).
1. How do we inquire into things we you know?
2. How do we inquire into things of which we know nothing?
Keyword: “searching” (ζητέω).
The Evolutionary Computation and
Problem in PhilSci: “given the vast and noisy field of possible options,
how do scientists identify which problems, techniques, and other
resources are more likely to be fruitful to pursue” Nickles 2003
The EC answers:
There are no rules of searching
There are meta-rules of searching, i.e. the rules of evolution.
Search1 can combine with search2
Combine successful search with unsuccessful search too.
Science does inspire new ways of thinking about modality
But the practice of science, i.e. in its numerical age, inspires different
kind of modalities.
The old debate about minimization of action, PLA, etc. finds new
inspiration from the concrete application of optimization through
When exact analytical methods are not available (and they usually
are not) numerical solutions optimized at least locally and find
See the concrete case of Schmidt and Lipson and GA
Butterfield, Jeremy. “David Lewis Meets Hamilton and Jacobi.” Philosophy of
Science 71, no. 5 (December 2004): 1095–1106.
———. “Some Aspects of Modality in Analytical Mechanics.”
arXiv:physics/0210081, October 19, 2002. http://arxiv.org/abs/physics/0210081.
Ellis, Brian. “Katzav on the Limitations of Dispositionalism.” Analysis 65, no. 285
(January 1, 2005): 90–92.
Katzav, J. “Dispositions and the Principle of Least Action.” Analysis 64, no. 3 (July
1, 2004): 206–14. doi:10.1093/analys/64.3.206.
Pulte, Helmut. Das Prinzip der kleinsten Wirkung und die Kraftkonzeptionen der
rationalen Mechanik. F. Steiner Verlag, 1989.
Stöltzner, Michael. “The Principle of Least Action as the Logical Empiricist’s
Shibboleth.” Studies in History and Philosophy of Modern Physics 34, pt. B, no. 2
(June 2003): 285–318.
Yourgrau, Wolfgang., and Stanley. Mandelstam. Variational Principles in
Dynamics and Quantum Theory. New York: Pitman, 1960.