32
CSC 599: Computational Scientific Discovery Lecture 2: Philosophy of Science Post-Logical Empiricists

CSC 599: Computational Scientific Discovery Lecture 2: Philosophy of Science Post-Logical Empiricists

Embed Size (px)

Citation preview

CSC 599: Computational Scientific Discovery

Lecture 2:

Philosophy of SciencePost-Logical Empiricists

Outline

Problems with Logical EmpiricismPhilosophy of Science Post-Logical Empiricism

Thomas Kuhn Imre Lakatos Larry Laudan Sociologies of science Model-base philosophies of Ronald Giere

Differences between CSD and Philosophy of Science

What shall we conclude from all of this?

Problems with Logical Empiricism

Problems with verification criterion Ayer, Popper

Problems with reductionism Quine

Problems with language Quine, Maxwell and Goodman

Problems with removing science from historical context

Philosophy of Science in the Anglophone world in the 1960sLogical Empiricism “This is science”/ “This isn't science” By people who aren't even scientists (a bit arrogant, isn't it?)

Other ways to characterize science: Historically

“What have scientists done in past?” Sociologically

“What are scientists doing now?”

Thomas Kuhn andthe Historical Perspective

Kuhn was a theoretical physics student assigned to teach course on ancient science

Suppose your job was to teach some ancient science, like mechanics circa 300 B.C.E.

Perhaps you might . . .1. Get into their mindset2. Study their worldview3. Solve problems the way they conceived of them

1: Get in their mindset

The Earth is the center of the Universe

2: Study their worldview

2 Realms Terrestrial: Below the moon

Linear motion Celestial: Above the moon

Circular motion

Five elements: Ether (Purely celestial) Fire Air Water Earth

(Rises)

(Sinks)

3. Solve problems the way they conceived them

Motion due to elements out of their place Fire rises Smoke (= fire + air), so it rises too Water flows down

Rain Rivers flow down to the sea

Earth Further down than the water Riverbeds, seabeds

No more reason motive force when an element is at its natural level

Kuhn's Insight

To the ancients, their science made sense! Largely self-consistent Explanatory (smoke rises) Predictive (What happens when you throw a rock

in the air? In water?)

Individual ideas should NOT be understood divorced from context Geocentrism 5 elements Natural motion is at-rest with same element

Part of a larger worldview

Kuhnian Paradigm

A self-consistent worldview, specifies which data are relevant laws are allowable experiments make sense to try theory make sense

Paradigms have: Exemplars

This is how you solve the problem E.g. Harmonic oscillator = mass + spring

Training networks that indoctrinate people into a way of thinking

Kuhn's View of Science(Normal Science)

Problem solving There are always problems

Predictions that can't be done Predictions that come out wrong Discrepancies, inconsistencies (“anomalies”)

Use the techniques culturally allowed by the current paradigm

Can NOT just do anything Must solve the problems of the paradigm with the

paradigm

What most scientist do most of the time

Kuhnian Revolution

Problems always accumulate Crisis = set of major problem(s) not solvable in old

paradigm Eventually some scientists try to invent new

paradigm in which they can be solved

All paradigms are incommensurate! Two competing paradigms have their own view of:

What the important problems are What counts as “data” What explanations look like

No way to objectively say one is better than other

What Kuhn Gives Up

Any notion of “rational progress” Would need some objective vantage point to

compare two paradigms As humans we don't have that

Old paradigms only abandoned when its old-fogey adherents die-off

Newtownian physics is just different that Relativity

What Kuhn Gains

Agreement with a lot of history Many examples of Revolutions

Phlogiston vs. Oxygen Geocentric vs. Heliocentric

Many examples of old paradigms dying only when there obstinate practitioners die off Priestley never accepted Lavoisier's oxygen Lord Kelvin never accepted electromagnetic

theory

Back to Our Example

Ancient vs. Modern PhysicsModern Physicist “Our physics is better because

we can predict the phases of Venus”Ancient Physicist “I don't believe you. All things

in heavenly orbs are perfect spheres”Modern “Our notion of elements is better than

yours too. We can predict chemistry and spectroscopy”

Ancient “So? We predict their motion and stillness!”

What do you think?

Allied Approaches

In 1960s and 1970s people wanted Kuhnian agreement with history Logical Empricist notion of “rational progress”

Allied Approahes Lakatos Laudan

Imre Lakatos

Born Imre Lipschitz in Hungary (1922) Jewish Degree in math, physics and philosophy, 1944 Changed name to Imre Molnár to avoid Nazis

Mother, grandmother, killed in Auschwitz Communist during WWII Changed name to “Lakatos” (= Locksmith)

Honors Geza Lakatos (Hungarian general) Falling out with communists in early 1950s Fled to England after Soviet Union invaded in

1956

Imre Lakatos

Research programmes: Hard core + Auxiliary hypotheses

Hard core = unshakable beliefs of practitioners Negative heuristic “Don't change the Hard Core!” E.g. “DNA turns to RNA turns to protein” To be defended against falsification

Aux hypotheses = beliefs that can be revised E.g. “Gene XYZ regulates pathway ABC” Protective belt around hard core

Lakatos' answer to Kuhn's paradigmProgressive vs. degenerate

Progressive: growing, new predictions, techniques, novel predictions!

Degenerate: not growing, or no new facts in protective belt, ad hoc changes!

Larry LaudanAn American

Less colorful pass than Central EuropeansResearch traditions

Emphasizes problems solved over the theories of Lakatos' research programmes

All research traditions face 2 kinds of problems Empirical (Kuhnian anomalies) Conceptual (consistency with self or allied fields)

Rational acceptance/rejection of res. traditions Accept res. tradition that solved most problems Pursue res. tradition that's solving them at

fastest rate No notion of:

Cumulative progress (solved problems -> unsolved) Novel prediction = good, ad hoc = bad

A Sociological Perspective

Sociologists have studied scientists Toronto high-energy physics Dept. Conclusion “Run as a benign dictatorship” The “Dictatorship”

Older (read tenured) profs ran the committees of the department, were the bosses

The “benign” Younger (read untenured) profs knew the older guys

ran it and wanted them to! Younger profs wanted to concentrate on research

An extreme Sociological Perspective

Do the things talked about by these physicists really exist? Epistemological constructivism

Maybe yes, maybe no. Scientists acting in own interest (More papers!)

Ontological constructivism They are just socially-constructed things, like poems

or judicial system laws Hey is there a problem here with sociology too?

Philosophy and Self-consistency:

Philosophers make a living looking for inconsistencies in the other guys thinking Logical empiricists

Is language that devoid of metaphysics? Is reduction possible?

Kuhn Is there really no notion of progress?

Lakatos Is the “research programme” view itself a progressive

view of science? Laudan

What exactly is a research tradition and how do you know how many problems it has solved?

Sociologists Isn't your work just socially-constructed too?

Model-base philosophies of Ronald Giere

Wants to get away from “laws” Are not necessarily true Newton's Laws + Gravity assume no other

gravitational fields aroundScientific model:

Restrict generalization to some portion of “reality”

Only some objects Only some of their attributes

Compare behavior of abstract system (“model”) to allowed aspects of “reality”

Model only intended to model aspect of subset of reality

Example: Harmonic oscillator

Reality: Universe has many

things, including this spring

Model Equation: F = -kx Limited generaliztn:

k akin to spring's stiffness

x akin to mass's offset from equilibrium

Differences Between CSD and Philosophy of Science

1. Since Logical Empiricism, Philosophers of Science are interested in

What knowledge looks like If progress is being made How “real” the terms used are

2. Computer Scientists, however, are interested in writing systems that can manipulate

What shall we conclude from all of this?

1.Write systems! Do science! If scientists like our programs they will follow After the scientists follow, the philosophers

begrudingly and belatedly will This has happened once before in mathematics Proof for 4 color mapping problem done by

computer, people at first refused to call this “mathematics”

2.Philosophers of science will call what we do

“Logical Empiricism” because they are too arrogant to listen to the differences

Philosophical Issues (1)Realism vs. Instrumentalism

As real or instrumental as the scientist(s) wantReal:

Can point to simulated things and enumerate interactions

Examples: Simulation of planets

Instrumental: Imperfect simulations for lack of:

Memory Time Knowledge of initial state

Examples: protein folding in a vacuum Simulations of US/World economy

Philosophical Issues (2)

Language Use the language the scientist(s) give you Construct new terms from it

Probably definitional “New term” = “These relationships among old terms” Constructive abduction

Might have added expressive power if given “primitives” from which can express set of increasingly more powerful ideas

More have even more ideas if can define one's own set of primitives

HOW!?

Philosophical Issues (3)

Reductionism If its what the scientists want Often task is constructive

Build up how world looks from these primitives Subset of larger, reductionist task:

Try different sets of primitives to build up how world looks

Sometimes no “reductionism” in physics sense How would you tell a computer about evolutionary

biology?

Philosophical Issues (4)

Progress Of course want progress in the normal science

conceptualization Which do scientists want more?

Empirically justifying as many claims as possible (Logical Empiricists)

Making model that falsifiable, but not yet falsified (Popper)

Increasing body of theory (Lakatos) Increasing solved number of problems (Laudan)

Philosophical Issues (5)

Revolutions Better solve the “re-assemble the world from

your own primitives” problem first Don't worry too much about?

Most scientists, most of time, do Normal Science By working within given search space (primitives) can

reduce search space for answer User would want to steal credit away from program

anyway Ability to do thought experiments?

How would you even tell a computer about thought experiments?

Next time

Deterministic Representations for science Rules Equations Differential Equations QSIM (?)