Participation

Please attend class regularly and participate (10% of your grade). This is a “seminar style” course. You are expected to take an active role in the discussions. Please study the assigned materials before class and be ready to discuss them.

Assignments

You should write ten pass/fail précis, one pass/fail take-home exam, and three graded essays or a research paper (90% of your grade).

8/22

Week 1: The Collins case

Probability and statistical methods can be used in trial proceedings for evaluating the evidence presented against the defendant. Sometimes the evidence itself is presented in the form of probabilities and statistics. This numerical presentation is often useful but can also lead to confusions, exaggerations and errors. To illustrate some of the problems, we will start by discussing the infamous case People v. Collins (1968). This case started the law and probability debate in the United States back in the 70ies. Along the way, we will learn the basics of probability theory.

Guides, handouts, notes, etc.

• Reading guide week 1

• Handout week 1

Primary readings

• People v. Collins (1968)

• Finkelstein and Levin (2001), Statistics for Lawyers, Sec. 3.1-3.2

Additional readings

• Finkelstein and Fairley (1970), A Bayesian Approach to Identification Evidence

8/29

Week 2: Match evidence

Match evidence, such as DNA and fingerprint evidence, offers the clearest example of how probability and statistics can help to weigh evidence. We will discuss how match evidence works, and the debates about how it should be presented at trial. We will continue learning about probability theory.

Guides, handouts, notes, etc.

• Reading guide week 2

• Handout week 2

• Basics of Probability

• Handout Bayes’s theorem

• Handout betting interpretation

Primary readings

• People v. Rush (1995)

• Wasserman (2004), Forensic DNA Typing

Additional readings

• Haavik(2016), It’s a match. Or is it?

• Koehler (1993), Error and Exaggeration in the Presentation of DNA Evidence

• Zabell (2005), Fingerprint Evidence

9/5

Week 3: Unlikely coincidences

If a suspiciously high number of people die in a hospital when a nurse is on duty—much higher than when other nurses are on duty—is this evidence that the nurse is culpable, or was this just a coincidence? We will look at two infamous cases, Lucia de Berk and Sally Clark, in which statistics about unlikely coincidences were used—actually, abused!—to conclude that a wrongdoing occurred.

Guides, handouts, notes, etc.

• Reading guide week 3

• Handout week 3

Primary readings

• R v Clark (2003)

• Derksen (2007), The Fabrication of Fact

Additional readings

• Meester and others (2006), On the (Ab)use of Statistics in the Legal Case Against the Nurse Lucia de B

• Fienberg and Kaye (1991), Legal and Statistical Analysis of some Mysterious Clusters

• Dotto, Gill and Mortera (2022), Statistical Analyses in the Case of an Italian Nurse Accused of Murdering Patients

• Gill, Groeneboom and de Jong (2019), Elementary Statistics on Trial: the Case of Lucia de Berk

9/12

Week 4: Profiling

There is little doubt that match evidence should be used at trial as evidence of guilt, for example, when the defendant matches the crime traces. But what about evidence that people with certain traits are extremely likely to a commit a crime? For example, data show that people who committed a sexual offense are very likely to commit it again. How should this information be used at trial?

Guides, handouts, notes, etc.

• Reading guide week 4

• Notes Pundik

• Notes Redmayne

• Key topics weeks 1 to 6

• Risk assessment tools: Static-99, SORAG, OxRisk, PATTERN,

Primary readings

• Redmayne (2003), The Relevance of Bad Character

• Pundik (2020), Predictive Evidence and Unpredictable Freedom

Additional readings

• Aviv (2013), The Science of Sex Abuse

• Di Bello and O’Oneil (2020), Profile Evidence, Fairness, and the Risks of Mistaken Convictions

• United States v. Vue (1994)

9/19

Take-home exam due

Week 5: Probability of guilt?

We have seen how different types of evidence can be evaluated using probability and statistics. But what is the goal of this evaluation? Some think that the goal is to arrive at an assessment of the probability that the defendant is guilty, but there are also different probabilistic measures of the value of the evidence.

Guides, handouts, notes, etc.

• Reading guide week 5

Primary readings

• Regina v. T (2010)

• Dawid (2001), Bayes’s Theorem and Weighing Evidence by Juries

Additional readings

• Fenton (2011), Improve Statistics in Court

• Hamer (2012), The R v T Controversy: Forensic Evidence, Law and Logic

• Urbaniak and Di Bello (2021), Legal Probabilism, Sec. 1 and 2

9/26

Week 6: Probability and decisions

Once the evidence has been presented and evaluated, a decision must be made. To this end, legal standard of proof serves as decision criteria, for example, convict only if the defendant’s guilt was established beyond a reasonable doubt. Since these standards are not precise, legal probabilists have translated them in the more precise language of probability and statistics. The usefulness of this translation is debated because eliminating all reasonable doubts about someone’s guilt seems to require more than establishing a high probability of guilt.

Primary readings

• Smith (2017), When Does Evidence Suffice for Conviction?

• Gardiner (2019), The Reasonable and the Relevant: Legal Standards of Proof

Additional readings

• Kaplan (1968), Decision Theory and the Factfinding Process

• Redmayne (2008), Exploring the Proof Paradoxes

10/3

Week 7: Knowledge and legal proof (remote class)

Epistemologists are concerned with knowledge and its value. Knowledge requires more than high probability, as lottery propositions suggest. Could the notion of knowledge help us better understand the standards of proof in legal cases?

Primary readings

• Moss (2023), Knowledge and Legal Proof

Additional readings

• Papineau (2021), The Disvalue of Knowledge

10/10

First essay due

Week 8: Relative plausibility

A prominent alternative to the probabilistic approach is the theory of relative plausibility. On this theory, trial decisions are guided by comparing the relative plausibility of two competing explanations of the evidence. The explanations are put forward by the parties at trial and the more plausible should be prevail. This comparative approach tames the complexity of trial decision-making in which alternative possibilities about what could have happened are potentially infinite in number and cannot be examined in their entirety.

Q&A with Ronald Allen

The first part of the class will consist in a Q&A session with Ronald Allen, a legal scholar of evidence law, criminal procedure, and constitutional law. He was recently retained by the Tanzanian Government to assist in the reform of their evidence law. Ronald is the author of one of the readings assigned for today.

Primary readings

• Pardo and Allen (2007), Juridical Proof and Inference and the Best Explanatiopn

• Allen (2017), The Nature of Juridical Proof: Probability as a Tool in Plausible Reasoning

Additional readings

• Steele and Stefansson (2021), Belief Revision for Growing Awareness

• Dahlman (2024), What Does it Mean to Be Plausible?

10/17

Week 9: The story model

A close cousin of the theory of relative plausibility is the story model. According to this model, judges and jurors evaluate the evidence presented at trial not in a piecemeal manner, but holistically by constructing a unifying narrative or story of what happened. Can probability theory model this holistic process? Bayesian networks have been used to calculate numerically the weight of complex bodies of evidence. We will discuss the prospects of using Bayesian networks for evaluating evidence in trial proceedings.

Q&A with Mackor, Dahlman and Lagnado

The first part of the class will consist in a Q&A session with Anne Ruth Mackor, Christian Dahlman, and David Lagnado. The first two are legal scholars and philosophers, and the third is a cognitive psychologist. They are interested in understanding how judges and jurors assess evidence in conditions of uncertainty. They are the authors of some of the readings assigned for today.

Primary readings

• Pennington and Hastie (1993), ‘The Story Model for Juror Decision Making’, in Inside the Juror: The Psychology of Juror Decision Making - check canvas

• Fenton and other (2019), Analyzing the Simonshaven Case Using Bayesian Networks

Additional readings

• Dahlman and Mackor (2019), Coherence and Probability in Legal Evidence

• Urbaniak and Di Bello (2021), Legal Probabilism, Sec. 3

10/24

Week 10: The Conjunction Paradox

Legal cases often require that multiple propositions be established. For example, to establish bribery, a prosecutor needs to establish several elements: (i) a public official received the bribe; (ii) the bribe was something of value; (iii) the public official carried out an official act because of the bribe; (iv) the bribing party had intent to bribe. Can probability and statistics help to put them together, and yield a unified picture? This is where legal probabilism faces significant challenges, one of the most discussed being the difficulty about conjunction.

Q&A with Emily Spottswood

The first part of the class will consist in a Q&A session with Emily Spottswood, a legal scholar interested in how reforms to the rules of procedure and evidence can make civil and criminal litigation more accurate and cost-effective. Emily is the author of one of the readings assigned for today.

Primary readings

• Cohen (1977), ‘The Difficulty about Conjunction’, in The Probable and The Provable - check Canvas

• Spottswood (2016), Unraveling the Conjunction Paradox

Additional readings

• Dawid (1987), The Difficulty About Conjunction

• Haack (2014), Legal Probabilism: An Epistemological Dissent

10/31

Second essay due

Week 11: Beyond the witness?

When evidence is presented at trial, it will be scrutinized and tested against objections by both parties. This cross-examination is directed at witnesses, so their live testimonies play a central role. But new developments in technology have challenged this testimony-centered approach. How to handle machine evidence, such as risk assessments, AI diagnostic tools, or softwares that interpret DNA evidence? Should the testimony-centered approach be abandoned?

Q&A with Ed Cheng and Alex Nun

The first part of the class will consist in a Q&A session with Ed Cheng, a legal scholar at interested in expert evidence, law and statistics, and remedies, and Alex Nun, a legal scholars interested in how evidentiary rules and the allocation of decisionmaking authority in the courtroom affect verdict accuracy, efficiency, and legitimacy. Ed and Alex are the authors of one of the readings assigned for today.

Primary readings

• Cheng and Nunn (2019), Beyond the Witness

• Roth (2017), Machine Testimony

Additional readings

• Helmus and others (2022), Static-99R: Strengths, Limitations, Predictive Accuracy Meta-Analysis, and Legal Admissibility Review

• Thompson (2023), Uncertainty in probabilistic genotyping of low template DNA: A case study comparing STRMix and TrueAllele

11/07

Week 12: Eyewitness identifications

Eyewitness testimony is often the central piece of a prosecutor’s case against the defendant. Eyewitness evidence is not numerical on its face—it consists of qualitative statements “I saw this or that”. However, a growing body of research in psychology and law has deployed the tools of probability and statistics to assess the value of eyewitness evidence. While some studies emphasize its many pitfalls, more recent studies offer a more nuanced picture.

Primary readings

• Loftus and Palmer (1974), Reconstruction of Auto-mobile Destruction: An example of the Interaction Between Language and Memory

• Wixted and Wells (2017), The Relationship Between Eyewitness Confidence and Identification Accuracy: A New Synthesis

Additional readings

• Smith et al (2020), Evaluating the Claim that High Confidence Implies High Accuracy in Eyewitness Identification

• Trenary (2013), State v. Henderson: A Model for Admitting Eyewitness Identification Testimony

11/14

Week 13: Wrongful convictions

We have examined how probability and statistics help to evaluate evidence, piecemeal or in the aggregate, and how they can guide trial decisions. But they can also be used as tools for assessing how the trial system performs as a whole, for example, how accurate it is and how often innocent people are convicted. Some estimate that the rate of wrongful convictions is about 3-5%.

Primary readings

• Risinger (2007), Innocents Convicted

Additional readings

• Simon (2011), Limited Diagnosticity of Criminal Trials

• Sangero and Halpert (2007), Why a Conviction Should Not Be Based on a Single Piece of Evidence

11/21

Week 14: Workshop

Students will workshop their research papers by giving short presentations accompanied by either a handout or slides and followed by Q&A. Each presenter will have 30 minutes to be split between the presentation itself and Q&A at the discretion of the presenter. The list of presenters is TBA.

11/28

Week 15: Thanksgiving break

12/5

Week 16: Conclusion

What have we learned? What questions are still left open and what lines of inquiry seemed more promising? Is legal probabilism a good idea or a big mistake?

Primary readings

• Tribe, Trial by Methmatics

• Saks and Neufeld (2011), Convergent Evolution in Law and Science: the Structure of Decision-making under Uncertainty

Additional readings

• Harcourt (2018), The Systems Fallacy: A Genealogy and Critique of Public Policy and Cost-Benefit Analysis

• Hedden and Colyvan (2019), Legal Probabilism: A Qualified Defence

Third essay or research paper due December 5th