An Expert System Statistical Tutor


On June 28th 1995, a presentation titled 'An Expert System Statistical Tutor for (Psychology) Students' was delivered by Dr Lisa Wise, Lecturer in the Department of Psychology at the Clayton campus. She was introduced by Mr Jack Chorowicz, Associate Director of the Computer Centre, who, in his introduction, spoke of the widespread acceptance of PCs in the workplace, at home and in schools, and the fact that this will increase people's expectations. He made reference to the news that Optus Vision is planning to run Ethernet connections into homes, commencing in 1996. He expressed the view that 'students will increasingly come to Monash with well developed skills in the use of computer technology, and will have very high expectations of its use here at the university'. He summarised some of the steps taken by the Computer Centre to prepare for the expected changes in the patterns of learning that will occur in the near future.

Jack described Lisa Wise as a person with a keen interest in the use of computers in education, and one who has devoted considerable effort to improving her knowledge and skills in order to utilise computers in her teaching programs (for example, see the PSY2022 Home Page). 'She has contributed in a very positive manner to the ongoing development of the staff and the student environments of Monash. She is an active member of the TWP (Technical Working Party) and regularly represents the TWP at the Computer Centre's Executive Team Meetings.'


Background

Lisa began by noting that the co-workers on her CAUT-funded project were Dr Brett Degoldi (who has since moved to Edith Cowan University) and Mr Chris Hughes and Ms Renee Gedge (who has subsequently withdrawn from the project because of other work commitments). She then outlined the intended content of her presentation:

Funding

The initial proposal was driven by the pressure created by an increasing number of students, whilst at the same time, decreasing resources. There are fewer teaching staff and higher expectations of what students should be taught along the way. It was considered there was a need to find ways to continue teaching a lab program without compromising the content, whilst at the same time dealing with students who were less well prepared than they used to be. It was felt there was scope for a computer aided tool which could answer routine sorts of questions, thus freeing up staff to help students with conceptual issues. Lisa and the team put together a low-budget proposal, and although this probably improved the chances of being funded, it meant that the project was really under-funded throughout. This compromises quality and / or requires much "unfunded" additional work to produce a quality product. Lisa wished to emphasise the idea expressed by Branko Cesnik (Director of Medical Informatics and a previous HEPCIT presenter) that computer-based learning packages should not be developed without adequate resources, including labour.

Why Statistics?

Turning to the choice of subject, statistics, Lisa pointed out that at the undergraduate level the Dept of Psychology has a need to acquaint its students with statistics. This includes 'descriptive statistics' for describing large data sets, 'inferential statistics' used as a decision making tool when deciding between competing hypotheses, and 'multi-variant statistics' for doing exploratory analysis of huge amounts of data. She pointed out that to become a true statistician requires study towards an advanced mathematics degree. It is not expected that when exposed to one lecture of statistics per week, psychology students will come out at the end with a comprehensive knowledge of mathematical statistics. Therefore, the aim of this project was to attempt to teach the use of statistics as tool, via a computer based package..

There are 'statistics packages' available which, according to Lisa, provide access to huge computational power for people who may have little real knowledge of statistics. So there is the need for students to be able to select the right tool for the job at hand, so that appropriate analyses can be performed, with the acceptance that the average student may not have a full understanding of what the tool is actually doing as it crunches numbers. This leads to the question of defining expertise, and it has to be recognised that in teaching statistics within psychology, the emphasis is not on turning out expert statisticians.

Why an Expert System?

The next question outlined above - why use an expert system? - was addressed by reference to a diagram.

On a Cartesian 'xy' frame, Lisa used the 'x' axis to represent 'level of understanding of a problem', ranging from 'limited' to 'deep'.On the 'y' axis, we can represent the 'scope of the problem', ranging from 'specific' to 'broad'. Lisa asserted that expert systems find application in the region of the x,y space adjacent to the origin, 'where we have very specific problems, and there is a limited knowledge base required for those problems'. But if , with reference to this diagram, you asked 'Where would you find "expertise"?', most lay people would point to the top-right portion of the x,y space, where there is deep understanding of the broader aspects of the problem(s) at hand. But in the way that problem at hand has been described, there is a need for only limited understanding of what the statistical tests involve and therefore the content area is very limited. With these limitations placed on the scope of the problem, the situation is made amenable to use of an expert system approach.

What are the advantages of using this approach?

The expert system will answer questions for as long as students are willing to sit in front of a terminal, and they can even take the system home with them. In addition, there is a permanent record of this expertise, so if a staff member moves on, for example, the expertise remains. There is obviously a fast response at a time required by the student.

There are also indirect benefits. In developing and refining the system, teaching staff need to sit down and really think deeply about the teaching process and what needs to be taught. There is also the fact that through this system, the information delivered to students is consistent across all classes, and it is available for self or public scrutiny.

Educational Impact

At this stage, Dr Wise moved on to discuss the educational goals of the project. An advantage of the expert system over any other is that it allows the student to ask the program about its reasoning processes. The system is a rule-based one, using if-then-else style rules.

Lisa went on to explain that in RULEBOOK, a Windows based package, tree structured knowledge is stored. To illustrate this point, she brought up a display from her computer, and pointed out the 'if' clauses, the choice of variables, and so on, and also the 'then' clauses which may be associated with each test. At any particular 'node', you can look at the 'if' levels negotiated to reach that point, and you can build in as part of the 'then' clause a box directing the student to a further reference for more information about that particular outcome. At this stage, Lisa went into more detail about the tests, and discussed some of the more advanced aspects of the package, including some which have not yet been implemented in the project. (For more information about these details, it is suggested the reader contact Lisa direct - see below.) A criticism Lisa levelled at the current software is that whilst the outcomes (the final choices amongst which the package can choose) will have attached confidence levels, there is no way that uncertainty levels can be attached to the "qualifiers", the individual pieces of information forming the inference chain.

The concepts of 'backward' and 'forward' chaining were considered at this point, and Lisa illustrated backward chaining by an example from the computer. Too much detail was included to be reported here.

Possibly of more general interest, Lisa summarised aspects of the expert system which aids student learning. This includes

It is now thought that the most useful parts of the system, from a student's point of view, are - the hypertext explanations of words and concepts, explanations of why the system is asking particular questions, and for more advanced students, its role as an aid in selecting specific tests, and reminding students of the input requirements of each test.

Other Applications

Lisa finished up by looking at other areas in which this kind of tool would find useful application. Such areas as helpdesks, where straightforward problem solving is required, faculty handbooks, where decision making is based on a set of regulations (e.g. prerequisites for entry to subjects), are ones amenable to the expert system approach. The session concluded with Lisa fielding questions. Whilst attendance at this presentation was fairly low, it became obvious from the questions asked that Dr Wise had communicated her message well to those who were in the process of setting up or contemplating the use of an expert system to meet their particular needs.


The HEPCIT Steering Committee thanks Dr Lisa Wise for her fine presentation, which, whilst dogged by some technical difficulties, communicated a fairly specialised message to the small but dedicated audience which attended.


For further information, please contact Dr Lisa Wise direct.
Phone (03) 9905 3978; E-mail Lisa.Wise@sci.monash.edu.au


Summaries of past presentations

HEPCIT Home Page

Monash University Home Page

Last updated: 6 September 1995
Enquiries: Deidre.Schutz@lib.monash.edu.au

Copyright © Monash University 1995