Location

Courtyard by Marriott Washington, DC/Foggy Bottom 515 20th Street Northwest Washington, DC 20006 United States

“It is our contention that a Bayesian network (BN), which is a graphical model of uncertainty, is especially well-suited to legal arguments. A BN enables us to visualise the relationship between different hypotheses and pieces of evidence in a complex legal argument. But, in addition to its powerful visual appeal, it has an underlying calculus that determines the revised probability beliefs about all uncertain variables when any piece of new evidence is presented.” (Fenton, Neil, & Lagnado, 2013)

While Bayesian networks have been recognized as a powerful research framework in many fields of study for decades, the practical application of Bayesian networks in the field of law remains in its infancy. This new seminar program introduces law practitioners to the fundamental advantages of employing Bayesian networks and BayesiaLab for probabilistic reasoning and decision support.

We present a unified, normative framework for judicial reasoning by implementing the Bayesian network formalism with the BayesiaLab 10 software. It provides a visual and explainable approach to Artificial Intelligence to support core reasoning tasks related to the practice of law.

The seminar will focus on (1) Probabilistic Evidential Reasoning and (2) Causal Inference. If time permits, we will provide an outlook on topics related to (3) Decision Support & Optimization:

1. Probabilistic Evidential Reasoning

• Probabilistic reasoning with conflicting evidence
• Overcoming the Prosecutor’s Fallacy (Fallacy of the Transposed Conditional)
• Intercausal reasoning (“explaining away”)
• Quantifying the importance of observations and their sensitivity (Mutual Information)
• Analyzing evidence consistency (Bayes Factor)
• Using joint probability as a measure of plausibility
• Reconstructing foreseeability
• Eliciting knowledge from stakeholders using the Bayesia Expert Knowledge Elicitation Environment (BEKEE)

2. Causal Inference

• Formal and intuitive treatment of causation, prevention, omission, and omission of prevention
• Evaluating bias claims and discrimination (dealing with Simpson’s Paradox)
• Causal inference for estimating effects from observational data and expert knowledge
• Counterfactual causal analysis (“Had it not been for…”)
• Computing the “most relevant explanation” of an observed outcome
• Contribution/attribution analysis for allocating damages between multiple defendants

3. Decision Support & Optimization

• Developing adversarial reasoning strategies
• Quantifying outcome uncertainty (Entropy)
• Modeling jury perception
• Estimating overall case risk