|
 ABSTRACT
The author describes five separate projects he has undertaken in the intersection of computer science and Canadian income tax law. They are:
- A computer-assisted instruction (CAI) course for teaching income tax, programmed using conventional CAI techniques;
- A “document modeling” computer program for generating the documentation for a tax-based transaction and advising the lawyer-user as to what decisions should be made and what the tax effects will be, programmed in a conventional language;
- A prototype expert system for determining the income tax effects of transactions and tax-defined relationships, based on a PROLOG representation of the rules of the Income Tax Act;
- An intelligent CAI (ICAI) system for generating infinite numbers of randomized quiz questions for students, computing the answers, and matching wrong answers to particular student errors, based on a PROLOG representation of the rules of the Income Tax Act; and
- A Hypercard stack for providing information about income tax, enabling both education and practical research to follow the user's needs path.
The author shows that non-AI approaches are a way to produce packages quickly and efficiently. Their primary disadvantage is the massive rewriting required when the tax law changes. AI approaches based on PROLOG, on the other hand, are harder to develop to a practical level but will be easier to audit and maintain. The relationship between expert systems and CAI is discussed.
REFERENCES
Note: OCR errors may be found in this Reference List extracted from the full text article. ACM has opted to expose the complete List rather than only correct and linked references.
 |
1
|
|
| |
2
|
For a detailed description of the course and its underlying CAI principles, see David M. Sherman, "Computer-Assisted Instruction and Examination", Proceedings of the 8th National Educational Computing Conference, ICCE, Philadelphia, June 1987.
|
| |
3
|
"Upper Canada" was the name before 1854 of the British colony which eventually became the province of Ontario upon confederation in 1867. The Law Society of Upper Canada, having been founded in 1797, declined to change its name. It remains the governing body of the legal profession in Ontario and is run by the profession, entirely independent of government.
|
| |
4
|
The original inspiration for Document Modeler was Jim Sprowl's ABF Processor. See James A. Sprowl, "Automating the Legal Reasoning Process: A Computer That Uses Regulations and Statutes to Draft Legal Documents", {1979} American Bar Foundation Research Journal 1.
|
| |
5
|
Sherman, op. cit. (ICAIL-87); and L. Theme McCarty, "Intelligent Legal Information Systems: Problems and Prospects'', 9 Rutgers Computer & Technology Law Journal 265-294 (1983).
|
| |
6
|
Sherman, op. cit. (ICAIL-87).
|
| |
7
|
Income Tax Act, section 256.
|
| |
8
|
For general texts on ICAI, see Greg P. Kearsley, Artificial Intelligence and Instruction (Addison-Wesley; 1987), and Etienne Wenger, Artificial Intelligence and Tutoring Systems (Morgan Kaufman, 1987). A large number of papers on recent work appear in Proceedings of ITS-88, Intelligent Tutoring Systems (Montreal, June 1988), Universit6 de Montr6al and ACM (SIGART, SIGCUE).
|
| |
9
|
Income Tax Act, paragraphs lll(8)(a) and Co).
|
| |
10
|
Retract and assert are the Prolog names for the predicates which remove a rule from the database (equivalent to legislative repeal) and add a rule to the database (equivalent to enactment) respectively.
|
| |
11
|
The reason this solution is non-trivial is that it may be difficult to guard against an infinite loop, if the rules involved are numerous enough. Prolog could end up rejecting one alternative after another, never finding a set of completely unambiguous facts.
|
| |
12
|
As a trivial example, suppose the fact situation generates a dollar figure for employment income and a dollar figure for investment income, and the problem is "calculate total income". If the figures are the same, say $300, then an answer of 300 represents a clash--either the employment income or the investment income may have been forgotten. If the fact constraints require the employment income to be between $10,000 and $50,000, and the investment income between $500 and $9,500, the problem disappears. Whether this kind of approach can be extended to large-scale problems with dozens of possible individual facts is doubtful. Possibly it can be done by using numbers which are multiples of prime numbers and restricting the maximums of such numbers so that we do not end up with a product of two such primes. But we also have to stick to numbers that can easily be calculated with, since we are dealing with law students who are not interested in manipulating "messy" numbers.
|
| |
13
|
Income Tax Act, paragraph 46(1)(b). The proceeds are deemed to be a minimum of $1,000.
|
| |
14
|
John A. Self, "Bypassing the Intractable Problem of Student Modelling", Proceedings of ITS-88, Intelligent Tutoring Systems (Montreal, June 1988), Universit6 de Montr6al and ACM (SIGART, SIGCUE), 18-24.
|
| |
15
|
|
| |
16
|
For an example of a tool designed to do this for any expert system written in M.1, see William J. Claneey and Kurt Joerger, "A Practical Authoring Shell for Apprenticeship Learning", Proceedings of ITS-88, Intelligent Tutoring Systems (Montreal, June 1988), Universit6 de Montr6al and ACM (SIGART, SIGCUE), 67-74. I understand from Bill Clancey that the program described therein, TRAINING EXPRESS, will not be marketed commercially by Teknowledge.
|
| |
17
|
M. J. Sergot , F. Sadri , R. A. Kowalski , F. Kriwaczek , P. Hammond , H. T. Cory, The British Nationality Act as a logic program, Communications of the ACM, v.29 n.5, p.370-386, May 1986
[doi> 10.1145/5689.5920]
|
| |
18
|
For example, Professor Pierre Mackay at the Universit6 de Qu6bec fi Montr6al has developed a stack (in French) for teaching law students about the Canadian unemployment insurance system. That stack can be used by lawyers for reference purposes as well.
|
|
Adobe Acrobat
QuickTime
Windows Media Player
Real Player
Pdf
H.2