Decision Trees for Lawyers

An INTRODUCTION To

DECISION TREES FOR LAWYERS

Michael D. Freeborn

What are we talking about?

Let's begin this tutorial on "decision trees for lawyers" with some laboratory research by our class. Take out a pencil and piece of paper. Write down what you think each of these phrases mean, in percentage likelihood that an event will occur:

	"There is a distinct possibility"
	"It may well happen"
	"It is probable"
	"It is reasonably possible"
	"The likelihood is remote"
	"It is very likely"
	"It is very likely"

Finished? Pencils down? Let's compare answers.

If you said 10% for "distinct possibility", you might be surprised to discover that many other people would say 30% or more -- three times the likelihood which you understood the term to suggest.

Every time I have done this exercise, the result has been just as dramatic. Put ten people in a room and ask them what these words mean -- you will get a startling range of numbers. Yet these are the very words we use all the time in communicating with each other the information needed to make decisions, sometimes very important decisions.

We don't need this confusion.

Making good decisions is hard enough already, without building in such additional opportunities for error. For example, we are already dealing with:

	Complexity-- Most decisions involve a variety of issues. In litigation, for example, there are legal and factual issues before liability is determined, and then there are issues regarding damages. We need a method of organizing a complex problem into a structure that can then be analyzed.
	Uncertainty-- The future course of events is hard to predict. In litigation, we cannot be certain that a judge will rule our way on a motion for summary judgment, or that a particular witness will be credible at the time of trial. We need a technique for identifying the material sources of uncertainty and quantifying them.
	Multiple Objectives-- We want several things at once, but progress toward one may restrict progress toward another. In litigation, for example, we want to keep attorney fees low, but we want to be well prepared. We need to trade off benefits in one area against costs in another.
	Different Perceptions-- In any analytical exercise, slight changes in inputs can lead to quite different results. In litigation, different lawyers will have different beliefs regarding the likelihood of various outcomes, even when expressed in percentages. We need a tool to help the decisionmaker reconcile these differing perceptions.

What are "decision trees for lawyers?"

Briefly stated, this is a method to organize and evaluate all available information to enhance the quality of our decisions regarding litigation. It borrows from other fields some well-established techniques of decision analysis, including so-called "decision trees." (For an excellent text on decision analysis generally, see Clemen, Making Hard Decisions -- An Introduction to Decision Analysis (1991).)

We use a rudimentary form of decision analysis all the time, even if we do not realize it. For example, in crossing the street we know that there is a chance we will be hit by a bus. But we decide to cross the street anyway, because we know that the chances of being hit are very small and so the "expected cost" of that decision will likewise be small. We also know that the "expected value" of successfully crossing the street -- the reason we want to do it in the first place -- is both valuable to us and more likely to occur.

This example is trivial, of course, when compared to more complex decisions we make. But the analysis is the same.

In decision analysis, the methods we use bring structure to chaos, add some objectivity to mere intuition, improve the quality of decisionmaking, and reduce the cost of resolving disputes.

The process involves four steps:

	Structure the problem
	Elicit probabilities
	Ascertain values
	Integrate all the information

Two visual tools...

Two visual tools are of key importance -- (1) Influence Diagrams and (2) Decison Trees.

Influence Diagrams define the decisions which are to be made, and the so-called "chance" events which are uncertain. We make judgments regarding these probabilities and the values of various outcomes. It is important to remember that in an Influence Diagram, we need not show the sequence of events. Rather, we are merely defining the factors which will influence the outcomes of our decisions.

Decision Trees specify the sequence of decisions and chance events -- not necessarily the chronological sequence but the logical sequence. They also specify the realization of "values" -- the rewards or penalties -- resulting from each path of outcomes. As we travel from one end of a decision tree to another, we can visualize what may happen from decisions we are about to make and assess the desirability of each path.