| |
1
|
DEFAULT was prepared for use by a legal clinic under the auspices of Project Pericles, a project funded initially by Digital Equipment Corporation and Harvard Law School. John Getsinger, Ellen Hays and the author participated in the develop ment of DEFAULT.
|
| |
2
|
To some extent, this knowledge might be viewed as a set of heuristics for expediting a search within a state space representation of possible actions. S_ee P. Harmon, and D. King, Ex_x_Re~k_S s~s, at 31, (1985). I think here the knowledge represents more than just ability to speed up solution processes. It also reflects the accumulation of experience embodying an evaluation of probabilities of desirable outcomes, and an assessment, often unconscious, of how to adapt the system to make best use of one's own skills and abilities.
|
| |
3
|
A major goal of the project was to make the use of the program educational, allowing the user to test and compare a variety of expert knowledge. More experimentation with allowing such use and game-playing using different expertise is underway in another project.
|
| |
4
|
The program runs on an IBM PC-AT. Our AT had two megabytes of Ram and a twenty megabyte hard disk. Because the program, and the knowledge bases exceeded the two megabyte limitation, it became necessary to load and unload modules during execution, slowing the operation considerably. A variety of means might be applied to speed up the process, such as simply obtaining a faster micro, increasing efficiency of some of the coding, and compiling portions currently interpreted. At least this has demonstrated that using a microcomputer is feasible, especially as micro technology advances.
|
| |
5
|
Buchanan and Shortliffe stress the importance of modularity for production systems, especially for modification of the knowledge base. B. Buchanan and E. Shortliffe, Rule-Based Expert Systems at 39-40 (1984). Modularity of this type also facilitates easy access to different portions of the knowledge.
|
| |
6
|
Compare the activities described by Vrecion in V. Vrecion, "On Possibilities of Exact Analysis and Computing of Decision-Making Processes in Law", in Artificial Intelliqence and Le~l information Systems, 193, at 195 (Ciampi, C. ed. 1982).
|
| |
7
|
For example, the user may currently obtain the suggestions of any one of three experts, and follow that expert's decision process, or view how that expert would rank possible choices at each stage of the decision process.
|
| |
8
|
AdditionB to the system have been outlined, but not yet implemented, which are intended to help the user actually prepare the documents, make the arguments, or otherwise carry out actions identified as desirable to implement selected tactics. See discussion below of "help" predicate.
|
| |
9
|
As noted in the following text however, what we saw the experts do is not precisely what the system ended up doing.
|
| |
10
|
|
| |
11
|
Cf. H. Gardner, The Mind's New Science, at 89 (1985), explaining G.A. Miller, The Maqical Number Seven, Plus or Minus Two: Some limits on Our Capacity for Process~nq Information, 63 Psychological Review 81 (1956). " in C Walter,
|
| |
12
|
Cf. M. Dyer and M. Flowers, "Toward Automating Legal Expertise, ComDutinq Power and Leqal Reasoninq at 57 (1985).
|
| |
13
|
Our analysis of the interviews, however, suggested this approach was a useful first step in other areas the experts worked in as well as the limited domain we were undertaking. Interview protocols suggested, for example, that, for other landlordtenant problems, such factors as the landlord involved, the geographic area of the rental property, or others might suggest a classification of cases.
|
| |
14
|
R. Waldinger, "Achieving Several Goals Simultaneously" in B. Weber and N. Nilsson, Readinqs in Artificial Intel!iqence at 250 (1981).
|
| |
15
|
The experts also seemed to use changing perceptions of the world as they played out two person games in their heads -- predicting acts of other actors, hypothesizing responses and reactions, and so forth. Upon making a prediction about another actor's action, the expert would then analyze the problem again in the light of the altered state suggested by that action. While data structures were designed to allow storage of such predictions and to facilitate game-playing, considerably more work needs to be done to implement such a capability. This aspect of the process is discussed at greater length below.
|
| |
16
|
Another part of the overall project was working with the problem of record keeping more generally, and the intention was eventually to port the information back and forth between these aspects of the overall system. While our portion designed a scheme for storing the information in a portable form, the actual use by the other system has never been tested.
|
| |
17
|
Figure 2 depicts a sample portion of a hypothetical network using representative terms described here.
|
| |
18
|
At the very beginning of the network, obviously there is no previous branch. The part of the program which classifies the case according to selected facts identifies the initial "node" argument from which the operating network will then extend.
|
| |
19
|
Or for this rule to fire -- one could see these terms as individual rules in a modified production system.
|
| |
20
|
If the first condition fails, that is if no one has checked the journal entry, other terms will become viable, including one that leads to a path in which the user is instructed to check the entry, and given guidance as to what to look for, and where to look.
|
| |
21
|
As suggested above, not all of this part of the program has been implemented, part of it is dependent upon interaction with other phases of the project that are more concerned with databases of records and forms. A discussion of part of the process for finding defenses and counterclaims can be found below.
|
| |
22
|
Cf. Vrecion, supra note 5, at 194.
|
| |
23
|
Whether it was from this particular fear, or the natural skepticism of lawyers, lawyers cooperating with us insisted on citations to authority, and explanation of reasons for any advice or decisions. Moreover, they wanted flexibility to make their own decisions. This system allows for such freedom.
|
| |
24
|
". . . {W}orkers in this field speculate that the most 8i~ificant bT-product of expert systems work will be the codification of knowledge." (Emphasis in the original) F. Hayes-Roth, D. Waterman, and D. Lenat, Buildinq Expert Systems, at 27- 28 (1983).
|
REFERENCES
Adobe Acrobat
QuickTime
Windows Media Player
Real Player
Pdf
I.2