Vincent Aleven and Kevin D. Ashley developed CATO to help teach law
legal reasoning; They based it on factors.
This Thoughtful Thursday explores the work of Dr. Ashley and his colleagues
and how it can be applied to let sortition juries or the whole population
replace the Supreme Court.
At the end of a Thoughtful Thursday on Judiciaries and Judges of February
25th, I closed
with problems in Constitutional
interpretation by high courts in Illinois, Italy and Indonesia.
A Case decided
for the plaintiffs
F21. A second case has
F1 and third
case could identify
F15 but did not
F21. A lawyer could distinguish
the second case by the presence by the presence of
And an arguer with the third case for a defendant would point out the lack
As Drs. Aleven and Ashley pointed out, boiling down the judges opinion
in a case to a set of yes/no values on factors does not capture all the
subtlety of legal reasoning or the theories involved.
(Although irrelevant to a participatory democracy, it is fair to point out than in a controlled test of argument skills and querying a legal database, the students who had the instruction from CATO did better than the students who had a conventional law lecture.)
Drs. Aleven and Ashley used trade scecret law in their studies and the codes corresponded to:
whether a product was unique
[F15], whether the plaintiff
disclosed the trade secret in the course of a business negotiation
and whether the defendant knew the information was
The first case might have these three factors (
The concept of factors was also illustrated by Fourth Amendment law in Rissland's wonderful review article on case-based reasoning, analogy and similarity in the law. She pointed out resoning in dealing with fourth amendment cases, those that determine whether the police can do a search without a warrant. The Supreme Court had already decided that the police can search a car when they make a stop. On the other hand, a warrant is required to search someone's home. But what about a mobile home? What about a mobile home being driven down the highway? What about a mobile home moored semipermanently in the mobile home park? In the words of Rissland, what about "a vehicle..hooked up to water and electricity but with its wheels still on," a "camper's tent with the person's things" and "a van with a bed and upholstered chairs."
One can develop hierarchies to help organize factors, e. g. combining
several factors into the uberfactor of "home" Then "home" is another
factor, along with "emergency" "permission" and who did the search.
(A person not affiliated with the police might find evidence of a crime
and alert the police officers--this could be a person with permission to
be in the dwelling, a roommate, a family member, or a landlord coming in to
do a repair. It could also be a person who is not authorized, a vigilante
or a common burglar.)
Then these factors might be the ones combined to determine whether evidence
is supressed, the outcome of the case.
Figure Four of the conference
paper >shows the factor hierarchy used in CATO, a tutoring system
to help law students argue cases appropriately from precedent.
F15 (Unique product) positively influences (supports) the abstract factors of
F104 that information is valuable and negatively influences
that "information is known."
These in turn positively impact the determination
that there was a trade secret.
F21, that the party knew "the information
was confidential" influences
F115 that there was a "notice
of confidentiality" which in turn positively influences
F114 that there was a "confidential relationship."
F101, leaves of the directed
graph, could be considered final legal factors that
determine the case.
Ashley and Aleven identified eight sargument strategies with cases for lawyers
and law students to use. But what are
the strategies for judges or sortition jurors to use. I will
use the term "case at bar" for the new case for which the jury is voting
on a decision and legal argument.
And any of the prior cases is a "precedent case." (generically
P 2, etc.)
The jury could simply say, choose
or vote that the "case at bar" matches one of the "precedent cases"
and match up the factors. That is
F3. The Case at Bar also has factors
P1was decided for the plaintiff, without loss of generality(wlg). Therefore case at bar would also be decided for the plaintiff. Presumably, this is what will happen in many cases.
The jury could distinguish.
That is the precedent
F3. They would say the case at bar has
F4They would in doing this by labling
P1as having factors
F4. (by "not"
F4, they might mean that the case has not demonstrated
F4, not necessarily that the lawyers in that case demonstrated conclusively or presented sufficient that
F4did not occur.)
The jury could change the reasons for a precedential case.
Assume the precedent P1 was decided for the plaintiff factors
F3. However, from the writing in the opinion, they also find that the case had factor
P1was decided for the plaintiff.
But the case at bar has factors
F3as well and the jury would like to vote it for the defendant. They don't want to overrule it. They could decide that the precedential case was decided on the basis of
F4. The jury then decides the case at bar on the basis of factors
F3for the defendant.
- And, lastly, the jury could simply overrule the precedential case.
These strategies are not as arcane as they might appear. Parents do them all the time, at least in homes with siblings. Bob, Age 18, is allowed to say out past midnight. His sister, Janet, Age 16, is not allowed to do the same thing. Their parent distinguishes. Bob is age eighteen, you are sixteen. Of course when Janet becomes eighteen and wants to say out late, she is going to cite that precedent. The parent really doesn't want to let her out and then says, Oh, you are a girl, and ... Or perhaps, she might want to add a new factor, "Just last month, I heard that there is someone preying on. Back then, our suburb was peaceful." Or she could overrule, "I was wrong then, I shouldn't have let him out-- we were lucky he came back safely."
The SystemIn the real world, the lawyers would present their arguments and the record would be available. The justices would assign factors to the case and write up an opinion. They would also say whether the opinion supports for plaintiff or defendent. Note that not all justices might render an opinion if other justices prepared one with which they agreed.
The system would ensure that any outcome was consistent with with prior precedents. Justices could apply the strategies above, such as adding factors to prior opinions to distinguish them or perhaps overruling them.
Assume that there was more than one opinion. Then the voting would begin. In the real world, there could be a different set of voters than justices. I proposed that the judges would write opinions and if there was a disagreement or dissent, then the American people would vote which opinion would rule.
(This raises the analogy of concurring opinions. Let
us consider an example with three opinions
and four factors.
A third of the voters might vote
F2 -> plaintiff.
Another third might vote that factors
F4 ->plaintiff. And the remaining third
Obviously the plaintiff will win the case at bar.
In Marks versus United States, as cited by Wikipedia, the narrowest opinion would
rule. IN the above case, there would be no holding but
what if the layout was:
A third of the voters
Another third might vote that factors
->plaintiff. And the remaining third
See the Wikipedia article for some more work and I will hold
Marks versus United States, 430 U. S. 188 (1977) for a future
In experiments, we would probably use squibs or short synopsis of the cases. We might help the authors along by requiring them to use a set of factors that we predefined with the help of legal experts or might allow them to add factors themselves.
Drs. Ashely and McLarenalso did an experiment with factors and a case body of ethics decisions that showed the power of precedents in information retrieval. And Ashley acknowledged that there is more to legal opinions than just a few yes/no values. But I propose an interesting experiment where groups would try and set up opinions and judge cases in various topic areas such as the fourth amendment or trade secret law.
Some groups will just write their opinion in prose. Other groups will just get to assign factors. And the third groupw ill do both. We can measure the predictability by giving people half the opinions and ask them to predict how they would decide the remaining cases, and then see how they do. We will also try splitting the decisions after half the cases were decided. Assume that a specific area has twenty questions to decide. Each group would get the first ten. Then, they would be given the next ten cases to decide. They will be asked to be true both to the precedents but also to apply their own opinion. I would assume there are some sophisticated statistics that could see whether each group is influenced in decisions eleven to twenty by the precedents that they are to follow-or they simply find a way to write or assign the precedents so as to write a justification for what they believe.
When I was at the 2001 International Conference on Artifical Intelligence and Law, Dr. Schauer lectured us in the Old Court House in St. Louis of Dred Scott fame. Aand he said that most of the time an appelate court judge could write the opinion and justify the opinion to meet his personal desire of how the case should be decided, but every so often, they say "it just won't write." And I recall a Clarence Darrow biography where he said that a lawyer must make the judge want to rule in their favor from an emotional sense more than provide the precedents and legal reasoning. To what extent would a system based on factors guide the participants, especially a jury or a whole democracy, to maintain stare decisis?
For Future Thoughtful ThursdayMarks Vs. United States
- Aleven, V., Ashley K., "Evaluating a Learning Environment for Case-Based Argumentation Skills" In the Sixth International Conference on Artificial Intelligence and Law: proceedings of the Conference University of Melbourne Law School, Melbourne, Victoria, Australia June 30 to July 3 1997, ACM Press, New York, NY Pages 170 to 179.
- Vincent Aleven and Kevin D. Ashley, "Doing Things with Factors" The Fifth International Conference on Artificial Intelligence and Law, May 21 to 24 1995, Page 31 to 40.
- Vincent Aleven and Kevin D. Ashley, "What Law Students Need to Know to Win" in the Fourth International Conference and Law, June 15 to 18, 1993 in The Netherlands, Pages 152 to 161.
- Bruninghaus, S. and Ashley, K. Improving the Representation of Legal Case Texts with Information Extraction Methods" In The Eigth International Conference on Artificial Intelligence and Law Proceedings of the Conference (Washington University, St. Louis, Missouri, May 21 to May 25, 2001). ACM Press, New York, NY, 2001, 42-51.
- Ashley, Kevin D. and Bruce M. McLaren, "An AI Investigation of Citation's Epistemological Role" In The Eigth International Conference on Artificial Intelligence and Law Proceedings of the Conference (Washington University, St. Louis, Missouri, May 21 to May 25, 2001). ACM Press, New York, NY, 2001, 32-41
- Rissland, E. L. AI and Similarity. IEEE Intelligent Systems 21, 3, 2006, pages 39 to 49.