Scribd
Upload
5 views20 pages

Lecture2-Good Theories

The lecture discusses the criteria that qualify a theory as 'good', emphasizing the importance of precision, testability, and the ability to make multiple predictions with minimal parameters. It highlights the limitations of ambiguous theories and the pitfalls of polynomial fitting in scientific predictions. Additionally, Occam's Razor is introduced as a guiding principle for selecting between competing theories when evidence does not favor one over the other.

Uploaded by

chinman8964
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
5 views20 pages

Lecture2-Good Theories

The lecture discusses the criteria that qualify a theory as 'good', emphasizing the importance of precision, testability, and the ability to make multiple predictions with minimal parameters. It highlights the limitations of ambiguous theories and the pitfalls of polynomial fitting in scientific predictions. Additionally, Occam's Razor is introduced as a guiding principle for selecting between competing theories when evidence does not favor one over the other.

Uploaded by

chinman8964
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 20

SCNC1111

Scientific Method and Reasoning


Lecture 2
What Qualifies As Good Theories?

Dr. William M.Y. Cheung


1
Beyond the Scientific Method
• Scientific investigation requires guidance beyond the
Scientific Method
– How should we compare theories devised to explain the
same phenomena?
– What qualifies as a “good” theory?
• https://youtu.be/0KmimDq4cSU
(5:09 – 6:44)

2
A Criterion on Theory
• A good theory cannot be ambiguous!
– Predictions from our theory should be as concrete and
precise as possible.
– Mathematics help us to be concrete and precise!
• E.g. Fine Structure Constant of Quantum
Electrodynamics (QED)
– as of 2006, determined to be (with some experimental
input)
𝛼 −1 = 137.035 999 070 (98)
– This is a precision better than a part in a billion!
[ref.: Gabrielse et. al., New Determination of the Fine Structure Constant from the Electron g Value and QED,
Phys. Rev. Lett. 97, 030802 (2006), Erratum, Phys. Rev. Lett. 99, 039902 (2007)]

3
A Reflection about Economics

vs

utility = 70 units utility = 90 units

"utility” =
the amount of "satisfaction" I get via
consumption of the goods under
concern

economist
4
Utility
• What is wrong with this?
– Can you predict how many customers will prefer noodles
over rice this afternoon at the CYM canteen?

– Such utility function is different for different person.


 i.e. we all have different preferences
– Even for the same person, such utility function changes
over time.
 i.e. I may prefer different cuisines on different days

• What can we predict?

5
Being Falsifiable is NOT Enough!

Greek Myths can also provide Predictions!

6
Another Criterion
• http://www.ted.com/talks/david_deutsch_a_new_w
ay_to_explain_explanation.html
– (9:14 – 16:39)

• Being testable/falsifiable is not enough!


• Good explanations: hard to vary
– but still explain the phenomena
– every detail of the explanations is needed for the
explanations
• Beware of “explanationless theories”

7
How to Always Pretend that You
can Explain Everything

8
Unknown Parameters
• Theories built on assumptions, e.g.
– masses attract each other via gravity
– electrons and the nucleus attract because they carry
opposite charges
• Theories contain unknown parameters that the
theories themselves cannot explain, e.g.
– Universal Gravitational Constant G (Physics)
– Electric charge of an electron (Chemistry)
• What should we do?
– My theory cannot tell me the values of these parameters
– How can I make quantitative predictions?

9
Predicting with Parameters
• Measure these parameters via experiments
• Use these experimentally obtained values to make
MORE quantitative predictions
• Compare these predictions with OTHER experiments
to confirm/falsify your theories

• Have as few parameters (assumptions) as possible,


and make as many predictions as possible!

10
“Polynomial Fitting”
Y

𝑌 = 𝑎 + 𝑏𝑋 + 𝑐𝑋 2

𝑌 = 𝑎 + 𝑏𝑋

𝑌 = 𝑎 + 𝑏𝑋 + 𝑐𝑋 2 + 𝑑𝑋 3
X
11
“Polynomial Fitting”
• What’s wrong with this?
𝑌 = 𝑎 + 𝑏𝑋 + 𝑐𝑋 2 + 𝑑𝑋 3

unknown parameters to be determined


from our (experimental) data points

• We are using 4 data points to determine 4 unknown


parameters
– We are not predicting anything!
– We can always pass through n data points using a
polynomial with n coefficients!
12
An Obvious Fallacy
• Imagine the day before yesterday somebody put $10 on your desk.
• Yesterday somebody also put $10 on your desk.
• Today, once again, somebody puts $10 on your desk.
• How would you generalize to arrive at a hypothesis?

Everyday somebody puts $10 on


my desk!

Really? Can you


make a prediction
Sure! My
from this
hypothesis
hypothesis?
predicts that
somebody put
And then
$10 on my desk
confirm/falsify it with
yesterday.
an experimental test?
And we already
What the …
know this is true!
So confirmation!!
13
Another Fallacy
• Imagine a new experimental result does not agree with an
existing theory
– Falsification! This theory need to be modified, if not thrown away…
We can add a new term to
the theory to explain the
new experimental result!

This new term carries an


unknown parameter. How
can we determine its value?

You are using the new


experimental result as an The value of this
input to your new theory. You new parameter
Where is the are not explaining it! must be one that
prediction of reproduces the new
your new experimental result.
theory?
14
Which One to Pick

when two theories can both explain the


same phenomena?

15
Occam’s Razor
• What should we do if
– Two theories both succeed in
explaining the same phenomena
– No available experiments can
falsify one of them
– Which one should we prefer?

• Occam’s Razor
• http://www.youtube.com/watc
h?v=oAp3jT8n6Qs

16
Occam’s Razor
• “All things being equal, the simplest explanation
tends to be the right one.”
• E.g. I used my staff card this morning, but now I
cannot find it in my office.
– Explanation 1: I misplaced my staff card somewhere else.
– Explanation 2: Someone (one of you, a fairy, etc.) broke
into my office, stole my staff card but leaving no trace.
• Which one would you prefer?
• Recall
– Feynman discussed flying saucers in the video
– Have as few assumptions as possible!

17
Caution!
• Occam’s Razor is only a philosophical guiding
principle.
• Used when currently available evidences do not
favour one theory over the other
• No reason why simpler theories always have to be
correct
– Later experiments may favour the more complicated
theory and falsify the simpler one.

18
Some Criteria on Good Theory
• A good theory cannot be ambiguous!
– Predictions from our theory should be as concrete
and precise as possible.
– Mathematics help us to be concrete and precise!
• Should not be easily varied
• Be able to make many predictions with as few
parameters as possible
– Beware of “Polynomial Fitting”!
• Occam’s Razor

19
A Quote

20

You might also like