attacking non-bayesian reasoning

18
Attacking Non-Bayesian Reasoning William Thompson UC Irvine February 15, 2014

Upload: irving

Post on 23-Feb-2016

54 views

Category:

Documents


0 download

DESCRIPTION

Attacking Non-Bayesian Reasoning. William Thompson UC Irvine February 15, 2014. Do Jurors Give More of Less Weight Than They Should to Forensic Science?. Importance to the law of evidence Claims of Underweighting - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Attacking Non-Bayesian Reasoning

Attacking Non-Bayesian Reasoning

William Thompson UC Irvine

February 15, 2014

Page 2: Attacking Non-Bayesian Reasoning

Do Jurors Give More of Less Weight Than They Should to Forensic Science?

• Importance to the law of evidence• Claims of Underweighting– Finkelstein & Fairley (1970) , citing Edwards (1968);

Slovic & Lichtenstein (1971)• Claims of Overweighting– Tribe (1971)

• Jury Simulation Studies– E.g., Schklar & Diamond (1999); Nance & Morris,

(2002; 2005)

Page 3: Attacking Non-Bayesian Reasoning

Figure 1: Log Scale for Estimating Chances Defendant is Guilty

___Certain to be guilty___About 9,999,999 chances in 10 million that he is guilty___About 999,999 chances in 1 million that he is guilty___About 99,999 chances in 100,000 that he is guilty___About 9,999 chances in 10,000 that he is guilty___About 999 chances in 1,000 that he is guilty___About 99 chances in 100 that he is guilty___About 9 chances in 10 that he is guilty___One chance in 2 (fifty-fifty chance) that he is guilty___About 1 chance in 10 that he is guilty___About 1 chance in 100 that he is guilty___About 1 chance in 1,000 that he is guilty___About 1 chance in 10,000 that he is guilty___About 1 chance in 100,000 that he is guilty___About 1 chance in 1 million that he is guilty___About 1 chance in 10 million that he is guilty___Impossible that he is guilty

Page 4: Attacking Non-Bayesian Reasoning

What are the chances that an innocent man in a case like this would, just by coincidence, happen to match the DNA left by another man at the crime scene?– 1 in 10– 1 in 100– 1 in 1000– 1 in 10,000– 1 in 100,000– 1 in 1 million– 1 in 1 billion– Zero chances--impossible

Page 5: Attacking Non-Bayesian Reasoning

Guilty (G) Not Guilty(NG)

Framed (F) FUP FUP

Not Framed

(NF)

1-FUP 1-FUP

Match (M) No Match(NM)

Match Reported

(R)

1 FRP

Not Reported

0 1-FRP

Framed (F) Not Framed (NF)

G NG G NG

Match (M) 1 1 1 RMPNo Match

(NM)0 0 0 1-RMP

Figure 5: Bayesian Network Model for Evaluating the Probative Value of the Forensic Evidence Based on Individual Perceptions of the RMP, FRP and FUP

Page 6: Attacking Non-Bayesian Reasoning
Page 7: Attacking Non-Bayesian Reasoning
Page 8: Attacking Non-Bayesian Reasoning
Page 9: Attacking Non-Bayesian Reasoning

A DNA Match Between Perpetrator and Suspect: Is the suspect guilty?

Page 10: Attacking Non-Bayesian Reasoning

The DNA Expert Says:

Page 11: Attacking Non-Bayesian Reasoning

The prosecutor says:

Page 12: Attacking Non-Bayesian Reasoning

But the defense lawyer says:

Page 13: Attacking Non-Bayesian Reasoning

So what are the chances he’s guilty?

Page 14: Attacking Non-Bayesian Reasoning
Page 15: Attacking Non-Bayesian Reasoning

How many of the matching people are plausible suspects?

Page 16: Attacking Non-Bayesian Reasoning

We don’t really know…

Page 17: Attacking Non-Bayesian Reasoning
Page 18: Attacking Non-Bayesian Reasoning