What Jurors Hear When DNA Experts Testify: Results From Controlled Experiments Jonathan J. Koehler McCombs School of Business

Download Presentation

What Jurors Hear When DNA Experts Testify: Results From Controlled Experiments Jonathan J. Koehler McCombs School of Business

Loading in 2 Seconds...

- 117 Views
- Uploaded on
- Presentation posted in: General

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

What Jurors Hear When DNA Experts Testify:

Results From Controlled Experiments

Jonathan J. Koehler

McCombs School of Business

The University of Texas at Austin

koehler@mail.utexas.edu

Forensic Bioinformatics5th Annual Conference

Dayton, OH

August 13, 2006

A scientific method for testing hypothesized causal relationships between independent and dependent variables

Randomly assign subjects to groups

Manipulate independent variables

Measure dependent variables

Advantage: Control

Disadvantage: Artificial

Who Are the Subjects in Jury Experiments?

Students

“Real” jurors

“[T]he overwhelming majority of studies that have directly compared different mock juror samples have failed to find consistent differences…[There is] strong evidence that factors at trial affect students and nonstudents in the same way.”

(Bornstein, 1999, p. 80)

Simple Written Materials (1 page case summaries)

Detailed Written Materials (opening arguments, witness statements, closing arguments, judicial instructions, deliberation)

Detailed Video Materials (videotaped trials based on actual cases, deliberation)

“Studies that have directly compared presentation media [e.g., written summaries, transcripts, audiotape, videotape] – for either a whole trial or a portion of testimony – fail to offer consistent findings. . . . [R]esearch on the trial medium tends not to find many differences.”

(Bornstein, 1999, p. 82 & 84)

The results of controlled studies (i.e., experiments) with mock

jurors can provide insight into how real jurors think and respond at trial.

Theory

The weight that people attach to statistical evidence is influenced by whether or not they can easily imagine examples of the event in question.

Lottery Example

There are different ways to describe the chance of winning the daily three-digit New York Lottery Numbers game:

1. One in every 1,000 tickets out of the 500,000 tickets that are sold each day will win.

2. There is a 0.1% chance that your ticket will win.

Statistical Information Can Be “Formed” in Different Ways

Frequency Form: 1 in 1,000

Probability Form: 0.1%

Odds Form: 1:999

Statistical Information Can be “Targeted” in Different Ways

Multiple Target (examples: yes)

Single Target (examples: no)

“0.1% of the 500,000 tickets sold each day will win” [multiple]

“There is a 0.1% chance that your ticket will win” [single]

Statistical Form and Target influence whether jurors find it easy or hard to imagine examples of others who might match.

Easy to imagine examples of other matches?

Yes No

Two Groups

1. Probability Form + Single Target

“The probability that Mr. Clinton would match the semen stain if he were not its source is 0.1%”

2. Frequency Form + Multiple Target

“1 in 1,000 people in Washington D.C. who are not the source would also match the semen stain”

Facts

Hardware store owner shot and killed during robbery

Killer wore a mask

Killer bled at crime scene

Several neighborhood residents gave blood samples

One suspect matched partial PCR DNA profile

(RMP = 1 in 100,000)

No corroborating evidence

Independent Variables

Form (probability, frequency)

Target (single, multiple)

Reference Class Size (small, large)

Dependent Variables

Evidence Strength

Probability of Source

Probability of Guilt

Verdict

RMP & Reference Class Size affect whether examples of coincidental matches will be easy or hard to imagine.

RMPReference Class SizeExamples?

1 in 1,0005,000,000Yes

1 in 1,000,0005,000,000Yes

1 in 1,000,000,0005,000,000No

Implication: The way RMP is presented matters less as RMP becomes very small.

Expert: My tests cannot rule out [the suspect] as a possible source of the recovered genetic material. Approximately [X] out of [Y] people share this DNA profile, and the suspect is one of those people.

RMP = 1 in 100,000

A. 0.1 out of 10,000 people

B. 1 out of 100,000 people

C. 2 out of 200,000 people

RMP = 1 in 1,000

A. 0.1 out of 100 people

B. 1 out of 1,000 people

C. 2 out of 2,000 people

DNA evidence can be presented in odds form as the ratio of 2 conditional probabilities, when a few assumptions are made.

This is the Likelihood Ratio (LR)

The LR is a term that appears in a mathematical formula called Bayes Theorem.

77% of people verbally confuse LRs with Posterior Odds Ratios (Wolfe, 1995)

Most people reason more accurately with frequencies than with conditional probabilities (Cosmides & Tooby, 1996)

Some people equate P(H|D) with P(H&D) (Gigerenzer & Hoffrage, 1995) or with P(D|H) (Thompson, 1989)

Experts Make Errors When Explaining the meaning of a LR

LR = 14,961 [State of Texas v. Griffith, 1996]

Expert: “Given this evidence, it is 14,961 times more likely that the defendant is the father than a random man.”

Proper Statement (LR): "It is 14,941 times more likely that we would see this evidence if the defendant were the father than if the defendant were not the father.“

Improper Statement (Posterior Odds): “Given this evidence, it is 14,941 times more likely that the defendant is that father than that the defendant is not the father.“

Is it reasonable to expect that jurors will understand that these two statements are different?

Three Groups

1. Frequency: Approximately 1 person out of every 1000 would yield a DNA match with the semen and the defendant is one such person.

2. LR: It is approximately 1000 times more likely we would see this DNA match if the defendant is the source of the semen than if the defendant is not the source of the semen."

3. Posterior Odds Ratio: Given that we see this DNA match, it is approximately 1000 times more likely that the defendant is the source of the semen than that he is not the source of the semen.

Jurors are more persuaded by a LR than by a frequency

Jurors respond to a LR just as they respond to a posterior odds ratio.

Conclusion: Jurors think that a likelihood is a posterior.

Do jurors understand how to integrate error rates with RMPs?

False Positive Error Rate = .02 (also used .001)

RMP = 1 in 1,000,000,000

Three Groups

1. Error rate only

2. RMP only

3. Error Rate + RMP (separately)

The way DNA RMPs are presented matter

Easy to think of examples: Good for Defense

Hard to think of examples: Good for Prosecution

LRs are misunderstood

DNA error rates are often ignored

When not ignored, jurors are not sure how to

incorporate them.

Schklar & Diamond (1999)

Jurors usually underweight DNA evidence

Jurors overweight DNA evidence when given separate estimates

for the RMP and error rate

- overweighting persists with aggregation instruction

Nance & Morris (2005)

Aggregation instruction for RMP + error rate has no effect

LR approach yields higher conviction rate