Professional Documents
Culture Documents
Trial by Probability: Bayes' Theorem in Court
Trial by Probability: Bayes' Theorem in Court
Presented By...
Dave Bucheger
Jill Thompson
Sally Danielson
Justin Koplitz
Eric Hartmann
Presentation Overview
Illinois Cases
Governor suspended death
penalty
13 inmates cleared
12 put to death
64 criminal convictions
cleared nationally from
DNA testing
PROSECUTORS FALLACY!
P (G | TD ) P (TD | G) P (G )
1
P (G )
F (T ) * P (G ' )
F (T)
Mean P0
Mean P1
Bayes P1
Goodman
.001
.29
.47
.997
Goodman
.10
.29
.34
.803
Faigmon,
Baglioni
.20
.61
.70
.876
Faigmon,
Baglioni
.40
.64
.71
.816
Results of Chart
The mean P1 and the Bayes P1 show the jury
underweighs probabilistic evidence.
Prosecutors Fallacy of P( TD | G ) = P ( G | TD )
does not arise.
Heres a case where probabilistic evidence was
severely underweighed:
What is DNA?
Double Helix Form
Has 3 billion places
that tell traits
Much code identical
Unique places are
called DNA markers
T
Phosphate
G
Deoxyribose
DNA Fingerprinting
Compares DNA samples
DNA separated by
detergent
pressure
Conditional Probability
M = DNA match of
defendants blood and blood
at crime scene
I = event defendant is
innocent
I = event defendant is guilty
Conditional Probability
P(M | I) = match of two individuals blood sample, very
low chance, 10-8 or 10-10
P(I | M) = probability of innocence given the evidence
Bayes theorem gives
P( I )
P ( I=| P(M)
M)
P(M | I )
So P(I|M) P(M|I) unless P(I)
P( M )
Conditional Probability
P( I ) P( M | I )
P( I | M )
P( I ) P( M | I ) P( I ' ) P( M | I ' )
P( M | I ) P( I )
P( M | I ) P( I )
P( I ' ) P( M | I ' )
1
P( M | I ' ) P( I ' )
P( M | I ) P( I )
P( I ) P( M | I )
P( M | I ' ) P( I ' )
P( I ' )
Compounding Evidence
M1, M2, , Mk = all evidence introduced
k
Compounding Evidence
PI () P ( | I ) for a given event I
Conditionally Independent means
PI ( M 2 | M 1) PI ( M 2 | I )
Strongly Associated measured by the ratio
P ( M 2 | M 1)
P ( M 2)
Compounding Evidence
P ( M 2 | M 1)
P ( M 2)
P ( M 1 | I ) 5.88 x10 9
M2 = blood found on
defendants sock is
consistent with victims
P ( M 2 | I ) 1.47 x10 10
P( I ) P( M 1 | I ) P( M 2 | I )
10 9 (8.65 x10 10 )
(8.65 x10 10 )
...Translation
Probability of innocence
9 people in a billion!
Metaphorical Analogy
52! 8.07 x10 67 possible card permutations
DNA permutations much larger
Not all information can be determined
depending on degree
Analogous to a card if only number, suit, or
color could be determined
Metaphorical Analogy
Obtained from Crime Scene
13 13 1
1
1
1 24
52 51 50 49 48 47 46
6.0154 x10
Conclusion
Works Cited
Dale, Mike. Evidence: The True Witness.
[http://library.thinkquest.org/17049/]. 1998.
Saunders, Sam C. Compounding Evidence from Multiple
DNA-Tests. Mathematics Magazine. 72, NO1. Feb. 1999:
39-43.
Starr, Cecie. Biology Concepts and Applications. 4th Ed.
Brooks/Cole, 2000.
Special Thanks
Steve Deckelman - Models Professor
Steve Nold - Biology Professor, DNA Guru
Eric Larson - Down right smart guy