Course 5895: Artificial Intelligence and Legal Reasoning Seminar
Syllabus: January 1, 2017
Time and Place: Spring Semester, 2017: Wednesdays, 9:00 AM - 11:20 PM, Law Bldg G18
Professor: Kevin D. Ashley, Professor of Law and Intelligent Systems
3900 Forbes Ave., Room 525, (412) 648-1495
Learning Research and Development Center, 3939 O'Hara St., Room 519, (412) 624-7496
Seminar description and rationale:
How will the technology of legal practice develop in the near future, and what challenges and opportunities will these changes present for young attorneys and technologists? Machine learning applications and start-ups are revolutionizing e-discovery in litigation. IBM Corp. may soon apply its “Jeopardy!” game-winning Watson technology to question-answering and argument-making in law. These and other developments in natural language processing, argument mining, and automated information extraction offer tools for designing new process models for delivering legal services, promising greater efficiency and public accessibility. Legal reasoning, however, does not change as quickly. How will the new tools accommodate, employ and adapt the structures of legal knowledge such as arguments from statutes, regulations, cases, and policies? How, for example, will Watson reason with these legal sources?
Researchers in Artificial Intelligence (AI) and law are addressing questions like these as they design computer programs that can perform legal reasoning or that can solve, or assist attorneys in solving, legal problems. The researchers address how to represent what a legal rule means so that a computer program can decide whether it applies to a situation, how to separate “hard” from “easy” legal issues, and the roles that cases and values play in interpreting legal rules. Their answers, however, are not philosophical but scientific; they design computer programs that model the tasks and conduct experiments to evaluate how well the programs perform. Increasingly, they integrate their models with the underlying legal texts for purposes of improving legal information retrieval, prediction and decision-making.
In this seminar, law students and graduate students from other disciplines will consider these questions after reading excerpts from comprehensible research papers. The seminar will introduce the fundamentals of AI to law students and of legal reasoning to graduate students. Law students will more fully appreciate the techniques and assumptions employed by attorneys, and by AI and law researchers, to deal with the uncertainties inherent in legal reasoning. No familiarity with computer programming is required. Students will be asked to prepare short summaries (1-2 pages) of selected readings and a seminar paper. In lieu of a seminar paper, students with approval of the instructor may design and build an AI-related program for a legal application using new tools that do not require programming experience. (Such a program will not satisfy the legal writing requirement.) Graduate students are invited to propose relevant seminar paper topics close to their research interests.
Seminar topics overview:
This seminar provides an overview of work in AI and Law, and then focuses on very recent developments in information extraction from text for purposes of computerized question-answering and generation of arguments. These developments promise to break the knowledge acquisition bottleneck that has long hampered progress in fielding practical applications in AI and Law. They raise an interesting question, however. Now that we can soon extract information automatically from legal texts, what can programs do with it? And, exactly what kind of information should be extracted? Certainly, the extracted information will be used to improve legal information retrieval, helping to point legal professionals more quickly to relevant information, but can more be done? Can computers reason with the legal information extracted from texts? That has been the goal of much of the work in AI & Law: developing computational models of reasoning with legal information extracted from legal texts but extracted primarily manually, at least up to now.
Thus, in the first weeks of the seminar, we will focus in some detail on the nature of legal decision-making and some rule-based or case-based approaches to modeling it computationally. We briefly study an AI model of the kind of issue spotting law students perform in taking exams and then focus on some formal logical models of statutory reasoning and various lessons learned from these efforts. The seminar then takes a more detailed look at reasoning with legal cases or precedents, an important phenomenon in common law jurisdictions. We will compare and contrast a number of computational models of case-based legal reasoning in terms of how the legal information is represented and the extent to which they take into account underlying legal values. In particular we will address the following questions:
Š What is the process of reasoning with precedents in common law legal practice?
Š What aspects of this process do the computational models of case-based legal reasoning address? What is missing?
More recent work on computational models of argumentation unifies reasoning logically with legal rules and analogically with legal precedents. The models generate legal arguments represented diagrammatically for purposes of teaching or public discussion of legal issues. Finally, we turn to some practical applications of AI-supported tools for legal information retrieval from text, e-discovery, corporate compliance, network analysis of legal institutions, and helping judges and legal professionals with document drafting, sentencing, and other aspects of legal decision-making.
TWEN Course Website:
There is a TWEN website for this course, Artificial Intelligence and Legal Reasoning Seminar, at lawschool.westlaw.com. In order to access the site, you will need to use your Westlaw password and a course password to be distributed via email. Beside the syllabus, the website will contain materials for distribution and a discussion board.
The readings are primarily from Artificial Intelligence and Legal Analytics: New Tools for Law Practice in the Digital Age, a new book by Kevin D. Ashley, to be published in 2017 by Cambridge University Press. The readings will be distributed electronically via the TWEN course website. This material is an uncorrected final draft of the book. It is for personal use only. Seminar participants should not copy or distribute it.
Evaluation and Requirements:
Attendance will be taken on a regular basis throughout the semester. The main requirements are (1) a seminar paper or project and short presentation, (2) weekly short write-ups on selected readings, (3) classroom participation, and (4) participation in the collaborative design problem described below. Seminar grades will be based on a student’s seminar paper final draft or seminar project, oral presentation, classroom participation, one-page reading write-ups, and good faith participation in the collaborative design problem.
Seminar papers (or projects) and short presentations:
Students will be asked to write a fifteen-page (minimum) paper on a topic of their choice approved by the instructor. A non-exhaustive listing of possible paper topics may be found at sampletopics2017.htm. Students should contact the instructor early in the term to discuss appropriate paper topics. This is especially true for those who intend the paper to satisfy their law school writing requirement. Whatever paper topic a student chooses, a student should plan to develop a thesis and include an extended specific example or examples to illustrate his/her points. Graduate students are invited to propose paper topics connecting the seminar material to their own research interests in AI or computer science.
Legal application in lieu of seminar paper: In lieu of a seminar paper, students with approval of the instructor may design and build an AI-related program or module for a legal application. For example, students may use publicly available machine learning programs or resources from the IBM Watson Developer Cloud. (Such a program will not satisfy the legal writing requirement.)
Students will be asked to make an oral presentation of their paper or project at the end of the semester.
In order to stimulate classroom discussion and foster understanding of the readings, for each session, each student should prepare a one-page critique of each session’s readings. Generally, each session’s readings comprise one chapter of the Artificial Intelligence and Legal Analytics book. These one-page critiques should be submitted electronically to the instructor by midnight before each seminar session. Each student should select some approach described in the day’s readings that he/she find interesting and write:
(a) A very brief description of the approach, and short descriptions of:
(b) The strengths of the approach,
(c) The weaknesses of the approach, and
(d) The reasons why the student finds the approach of interest. Students should try to relate it to Cognitive Computing (to be defined in the first class) and some specific legal project, paper, or educational experience in which the student has been involved.
The instructor will assign particular students responsibility for leading discussions of particular readings at the next class.
Collaborative Design Problem:
New text analytic techniques promise to revolutionize legal information retrieval, but the machine learning on which they depend requires sets of training instances comprising manually annotated legal cases and statutes. “Annotating” means marking-up the texts of case decisions or statutes to identify instances of semantic types of information that are both important for conceptual legal information retrieval and pedagogically relevant. In order to support this need for manual annotation of legal texts, it is possible that law students could perform these annotations as part of their studies. As a by-product, they would produce the semantically annotated legal texts with which machine learning programs could be taught to automatically perform similar annotation of new texts.
As an extended collaborative design problem, the seminar will devote time each week to brainstorming and designing a pedagogical intervention involving annotating cases that could be applied in a legal course. This will involve demonstrations of programs and approaches, including:
- An interactive tool to select texts relevant for statutory or trade secret case analysis.
- The WebAnno annotation environment.
- The Value Judgment-based Argumentative Prediction (VJAP) Model, possibly analyzing student-annotated trade secret cases.
- Automated summarization of trade secret cases.
Seminar students will use the WebAnno tool to annotate trade secret cases in terms of sentence roles in legal argument and legal factors (to be defined).
A more detailed description of the collaborative design problem will be found here.
Schedule of Reading
I. Introducing AI & Law and its Role in Future Legal Practice
Date: January 11
Readings: Chapter 1 of Artificial Intelligence and Legal Analytics (available on TWEN site) and Collaborative design problem description
Š Discussion of possible paper topics.
Š Discussion of collaborative design problem re pedagogical uses of case annotation
II. Modeling Statutory Reasoning
Date: January 18
Readings: Chapter 2, Artificial Intelligence and Legal Analytics
III. Modeling Case-Based Legal Reasoning
Date: January 25
Readings: Chapter 3, Artificial Intelligence and Legal Analytics
IV. Models for Predicting Legal Outcomes
Date: February 01
Readings: Chapter 4, Artificial Intelligence and Legal Analytics
V. Computational Models of Legal Argument
Date: February 08
Readings: Chapter 5, Artificial Intelligence and Legal Analytics
VI. Representing Legal Concepts in Ontologies and Type Systems
Date: February 15
Readings: Chapter 6, Artificial Intelligence and Legal Analytics
VII. Making Legal Information Retrieval Smarter
Date: February 22
Readings: Chapter 7, Artificial Intelligence and Legal Analytics
VIII. Machine Learning with Legal Texts
Date: February 29
Readings: Chapter 8, Artificial Intelligence and Legal Analytics
IX. Extracting Information from Statutory and Regulatory Texts
Date: March 1
Readings: Chapter 9, Artificial Intelligence and Legal Analytics
[Note: No class on March 08, Spring Recess]
X. Extracting Argument-Related Information from Legal Case Texts
Date: March 15
Readings: Chapter 10, Artificial Intelligence and Legal Analytics
XI. Conceptual Legal Information Retrieval for Cognitive Computing
Date: March 22
Readings: Chapter 11, Artificial Intelligence and Legal Analytics
XII. Cognitive Computing Legal Apps
Date: March 29
Readings: Chapter 12, Artificial Intelligence and Legal Analytics
XIII. Next Steps toward Realizing Cognitive Computing Legal Apps
Date: April 5
XIV. Collaborative Design Problem Wrap-up
Date: April 12
XV. Student Short Presentations of Seminar Papers
Date: April 19
Š Students’ give short presentations of their papers or projects.