Sat 20 November 2010

“Scientific Method and Falsification”

A discussion on .

Led by David Hewson


1. Catching Up & Introductions
2. GS Diary Reports - "Our GS Experiences"
3. David's “Scuentific Method and Falsification" presentation.
4. AGS Business
5. Close.

1. Catching-up

We “always” allow a little time for review of our lives and activities.

2. GS Diary Reports - "Our GS Experiences"

Some of us experienced events in the month past, which demonstrated the value, and perhaps some limitations, in the applications of GS principles. Sharing of these is a valuable part of our time together.

3. David's presentation “Scientific Method and Falsification"

How does scientific method work and what is falsification? 

The seminar was about scientific method, falsification and belief revision.  I.e. how we find out information about the world, what's knowable about reality with science and how to review our beliefs/knowledge in the light of new evidence.

Knowledge acquisition.

1)   Ways to make check out our beliefs.  Example of Tom, Dick and Harry who each claim to be able to run the 400m the fastest.

i)    Direct observation: Line them up on a 400 m track and watch them run the whole race. Dick comes first. Problems with this approach are that there are a lot of things we can’t directly observe and the cost of observing everything ourselves.

ii)   Indirect observation where something is inferred from what you observe.  E.g. No 400m track is available but there’s an old railway tunnel that’s 400m, so you use that.  You stand at the top of the hill above the tunnel and after you fire the starters gun you see them run into the tunnel.  You then walk over the top of the hill and look down at the tunnel’s exit and see Tom come out first.  You go down to congratulate him.  Problems with this approach: Alternative hypotheses.  E.g. The other’s say that Tom cheated as he had a bicycle just inside the tunnel and he rode that most of the way.  I.e. this approach is plagued with alternative reasons that can also explain the observation. Solution?  GS uses the formulation of converging inferences to help eliminate alternative hypotheses.

iii)  By authority.  You ask a sports doctor to check Tom, Dick and Harry and tell you which one can run the fastest.  He says that Dick will run the fastest.  Problems with this:  Which authority to choose when you have more than one.  And what if the authorities statements conflict.  E.g. Another authority, the coach, claims that Tom is faster as he has better technique.  So which one (if any) do you choose?

iv)  By assumption.  Harry looks the fittest so you assume he can run the fastest. Or Dick is your friend, so you support him and assume he will win because you like him.

v)   By logical inference.  This can extend any of the above methods.  I.e. inference based on observation forms method 2.  And Economists like inferences based on assumptions. In our running example: At last year’s event Harry won so you infer he’s the fastest athlete and will win again this year.  Or Harry is a fit 57 year old whereas Tom has sprained his ankle and is on crutches and Dick is over 90 and not doing much walking let alone running these days.  Hence you infer that Harry will win.

2)   Scientific method

i)    Definition of “truth” .  E.g. “Correct symbolism to factual observation”.  And we differentiated this from validity, which we defined as logical consistency.

ii)   Falsification: The act of disproving a proposition, hypothesis, or theory.

iii)  Falsifiability or refutability: the logical possibility that an assertion could be shown false by a particular observation or physical experiment. That something is "falsifiable" does not mean it is false; rather, it means that IF the statement were false, then its falsehood could be demonstrated.  The claim "No human lives forever" is not falsifiable since it does not seem possible to prove wrong. In theory, one would have to observe a human living forever to falsify that claim. On the other hand, "All humans live forever" is falsifiable since the presentation of just one dead human could prove the statement wrong. Moreover, a claim may be true and still be falsifiable; if "All humans live forever" were true, we would never actually find a dead human, and yet that claim would still be falsifiable because we can at least imagine the observation that would prove it wrong. 

iv)  Four truth values for statements from Operation Philosophy:

a)   True (falsifiable and supported with evidence, so it’s taken as tentatively true)

b)   False(falsifiable and disproved),

c)   Indeterminate (falsifiable in principle but one can't test it yet.  I.e. the truth value is dated) and

d)   Meaningless (not falsifiable at all).

v)   The scientific method uses induction to help develop theories and deduction to test them.  Science using induction can generate many theories from a set of data.  Which do you choose?  Occam's razor says choose the simplest.

vi)  We covered four Types of Science from a paper written by Bob Pula and Stuart Mayper.

a)   Accepted Science: Theories that are not yet refuted, after rigorous tests.

b)   Erroneous Science: Theories that are not yet refuted, but are tested by false data.

c)   Pseudoscience: Theories inconsistent with accepted science, attempts to refute them avoided or ignored: E.g. Astrology, Numerology, Tarot Cards, etc.

d)   Fringe Science: Theories inconsistent with accepted science, not yet refuted, but attempts to do so invited.

vii) Asymmetry of knowledge about generalisations.

a)   Asymmetry of knowledge.  When testing generalities, we treat the statement as  tentatively true when supported but false with a single false result. e.g. the group tested y=2 + sin(PI*x) which gave y=2 for all the integers the group tested it with, until after about 5 minutes Gavan mentioned x=.1 where y=1.69, so the supported theory of y=2 for all x was disproved.

b)   Albert Einstein quote: "No amount of experimentation can ever prove me right; a single experiment can prove me wrong."

c)   For example, Aristotelian mechanics explained observations of everyday situations (he had an earth centred universe because everything seemed to revolve around the earth.), but these ideas were falsified by Galileo’s experiments (Galileo observed that the phases of Venus matched up with a sun centred rather than earth centred model.  He also saw the moons of Jupiter circling Jupiter rather than the Earth.  So Aristotelian mechanics were replaced by Newtonian mechanics, which accounted for the phenomena noted by Galileo (and others like Kepler). Newtonian theory covered the observed motion of all the planets. However, astronomers then noticed the size of the precession of the orbit of Mercury, which was not predicted by Newtonian mechanics, but was by Einstein's general relativity.

viii) So science tries to disprove theories.  An example of disproof in real life:

a)   Stress cause ulcers.

b)   Disproved when someone found a bacterium that caused it. 

c)   So the stress theory was tossed out.

d)   But more research showed that stress reduced the immune system’s ability to fix the stomach wall/lining ulcer that the ulcer bacteria caused.  I.e. both bacteria and stress are related to the ulcer.  One promotes it and the other reduces the body’s ability to counter it.

ix)  Despite popular impressions of science, scientific method does not answer all questions.  Scientific method answers only those that pertain to perceived reality.  To those theories that we can categorise as scientifically meaningless, general semantics suggests using the formulation of “agreeing to disagree” where appropriate.

3)   Falsifiability

i)    How falsifiability relates to: Knowledge, Information, Clarity and Precision. Generally speaking the more falsifiable a theory is, the better a theory it is. This is because the more a theory claims, the more opportunities exist to find observations that are inconsistent with it. Theories that make wide-ranging claims are considered to be epistemologically more desirable than those that do not (assuming they have not been falsified). So scientists are best to produce theories with large information content. Example:

a)   Theory 1: Mercury moves in an ellipse around the sun.

b)   Theory 2: All planets move in ellipses around their star.

ii)   Note: Theory 2 would appear to have higher status as a piece of scientific knowledge because it explains a generality in celestial mechanics. Consequently the second law is more falsifiable. Further any falsification of the first law will be a falsification of the second, however the reverse is not true.

iii)  Precision.  Which is better:

a)   If I let an object go, it will fall to the table by lunchtime.

b)   If I let an object go, it will fall to the table within one second.

iv)  And another example of precision. Which is better?

a)   If I rain dance for over 30 seconds, it will rain at least 5mm by midnight.

b)   If I rain dance hard enough, for long enough, it will rain eventually.

v)   Examples of hypotheses that are falsifiable.

a)   A heavier than air object, that is let go in still air, in Earth's atmosphere, will drop.

b)   It is never rains in Alice Springs.

c)   All substances expand when heated.

vi)  Examples of hypotheses that are not falsifiable

a)   Either it is raining or not raining (this covers all options).

b)   All points on a Euclidean circle are equidistant from the centre (this is covered by validity and not truth).

c)   The people I work with all have an inner subjective experience (no way to directly observe the subjective experience).

d)   Laurie will succeed at one of his future GS promotional projects. (no criteria to judge by. It needs a date and observed level.  E.g. Laurie will recruit 100 new members by Xmas 2011.)  I.e. the original is too vague.  Also, if Laurie’s expectations are too high and if he values it too highly as well, then he runs the risk of getting IFD.

e)   We are all just part of a huge giant’s dream.  And when the giant wakes up we will all vanish.

4)   Belief Revision

We reviewed a paper by Gabriella Pigozzi -University of Luxembourg (with modifications made by the presenter.)

i)    Introduction to belief revision

a)   Knowledge and belief

(1) Belief is a doxastic notion (pertaining to or depending on opinion; e.g. speculative or theoretical); knowledge is an epistemic notion (about what is true or false, and what constitutes valid information for making such evaluations).

(2) Knowledge is justified true belief.  I.e. a subset of someone’s beliefs, as knowledge requires some evidence.  So knowledge is more likely to be true than belief without evidence.  E.g. Eric could believe that man never landed on the moon.

(3) If beliefs can be false, it is usually assumed that a person’s belief set is at least consistent.  Although in real life this is not always the case. We had a challenge, to try to find some inconsistency in GS.  If you are not able to, then try to find something false.  Or at least find something where two or more high profile GS people disagree. The group could not find an inconsistency but they did mention “colloids related to life” as an example of something false.  They also gave Korzybski’s disclaimer about this claim (“The reader should not ascribe any uniqueness of the 'cause-effect' character to the statements which follow[about colloids], as they may not be true when generalized.”).  The group also mentioned the conflict between Hayakawa and Korzybski over the structural differential.

b)  What is belief revision?  Belief revision studies how a person should change her beliefs in face of a new information.  Suppose a person believes in a set of beliefs: B1, . . . ,Bn.  She then learns (or observes) event E, which contradicts one of the beliefs Bi . What should she do?

An example:  Suppose that a person believes this information:

(a) All European swans are white

(b) The bird caught in the trap is a swan

(c) The bird caught in the trap was trapped in Sweden.

(d) Sweden is part of Europe

(e) From a-d above you can derive: The bird caught in the trap is white.

(f) Now suppose that the bird caught is black. What should you do?  We agreed that its best to eliminate belief a)

(g) It is not obvious how to restore consistency in the example above as logic alone just says its inconsistent but not which ones to get rid of.  An additional complication is that more beliefs can be derived from the ones explicitly stated.  What do we do with the derived beliefs, do we keep them or do they go?

c)   We noted that observed facts are more reliable than inferences.  E.g. a pile of money on a book one can see is more reliable information than a pile under a handkerchief.  It turned out in our example that there where blocks under the handkerchief.

d)   So to restore consistency, an idea is that the information lost when giving up beliefs should be kept to a minimum. Another is that some beliefs are considered more important or well-established than others and the beliefs that should be retracted are the least important ones. So this leads to: The principle of minimal change. When someone revises her beliefs in light of learning/observing facts, only the minimal change necessary to incorporate the new information should be made.  The idea is that for a scientific theory change, a theory tends to accommodate new observations by making the smallest necessary changes. The author gave the following rationality criteria on performing a belief change:

(1) The set of beliefs should be kept as consistent as possible.

(2) The belief set should include logically inferred knowledge.

(3) The amount of information lost in a belief change should be kept minimal.

(4) If some beliefs are considered more important or well-established (with evidence) than others, one should retract the least important ones. (I gave the analogy of removing blocks at the top of a block tower which does not affect the overall structure versus removing the ones at the bottom which makes the whole tower collapse.)

e)   Three attitudes an agent can have with respect to a sentence and its negation

(1) Believe that is true (acceptance).

(2) Believe that is false (rejection).

(3) Neither believe that is true nor that it is false (suspension of belief).  From operational philosophy, this can be subdivided into:

(a) Currently indeterminate.

(b) Meaningless, therefore never determined.  So you forever suspend your belief.

ii)   Three kinds of belief changes.  Expansion, contraction, and revision of a belief set K (i.e. a knowledge based belief set.):

a)   Expansion:  New evidence E is added to K, together with the logical consequences of the evidence E.

b)   Contraction: Some belief Ki in a belief set K is retracted without adding any new facts. This is done for the set to be logically consistent, so some beliefs from K must be given up. (And it may also occur so as to simplify the belief set. I.e. an application of Occam’s Razor.)

c)   Revision: New information E that is inconsistent with some beliefs in K is added and to maintain consistency, some of the old beliefs in K are deleted.

iii)  Exercise

a)   All Korzybski’s statements are true.  Hence you can trust GS formulations.

b)   Korzybski stated in S&S p 111-122 that colloidal chemistry is possibly the basis for life.

c)   Five years after Korzybski’s death, Crick and Watson discovered DNA.

d)   DNA => RNA => proteins, are found to be the basis of life. Not colloids.

e)   Hence AK was wrong here.  A contradiction. What should you do?  Should one toss out all of GS?

f)      The group decided to modify statement a) getting rid of the “All” and replacing it with “most”.  I.e. “Most of Korzybski’s statements are true (i.e. supported by current day evidence.)”


4. ~ AGS Business ~

Membership, 2011 Programme etc.

5. ~ Close ~

Next Meeting: December Christmas Party - Call for details!

(Updated 22 November 2010)