Alla Petrova

Jeff Wartes

Patricia Winters

Proposal

The expert system we want to create will help diagnose one-on-one relationship problems. We will be extracting our knowledge base from a practicing psychologist. Being that this subject is too large in scope, we will narrow down our system to include problems on:

Description

The system will be confined to dealing with one-on-one relationship problems and will be targeted to users over 18 years of age. The system will ask questions that will attempt to identify relationship patterns and what problems they may create. Information gathering questions will be in a menu format. Once the system recognizes one or more patterns it will give feedback to the user. The patterns will be recognized by a numeric score. The score will be derived by taking the user’s answers and weighing them with a predefined value. Our recommendations will not be derived from the user, but will instead be derived by the diagnosed problem. In other words, there will be a few predefined recommendations for any given problem recognized by the system. If the user rejects the given recommendation, a different one will be given.

Design

The main execution will call on a number different modules based on the user’s response to introductory questions. The system will be divided into the following tasks:

Demographic information gathering – Gets personal information about the two people involved in the relationship conflict. For example, name, age, and gender.

Main driver – Determines suspected problem area(s) based on answers to overview questions. Calls appropriate subroutine a particular problem is suspected.

Detailed problem testing – A collection of modules where each are focused on a particular problem. The problem areas are:

Diagnosis phase – Uses the score to determine if a particular problem is significant. If the score is within predefined ranges, the problem will be flagged for resolution.

Recommendation – Provides one or more recommendations for any problem that is flagged in the diagnosis phase. Here the user can reject the recommendation, or accept. If the user rejects, the system will give at least one alternative recommendation.

Problems/Shortcomings

Scope – Since psychology is such a large field with highly complex subjects, we have drastically limited our scope to relationship problems. Even this may be challenging to complete in the allotted amount of time provided.

Domain – There are no clear definitions for a lot of the conflict patterns that we intend to address.

Project division – All modules need to share fact/variable names. Since the project is being split between three people, good communication is essential for successful integration of the modules.

Flow control – The questions must be given extensive thought in order to address all possible problem areas without bias. Once a problem area is suspected, it may be difficult to control the flow within the program using CLIPS.

Subject (psychology) – Using scores for each of the targeted problems requires an accurate scoring method. Answers to psychology questions are not easily quantifiable to a hard numeric score.

Time schedule – If time gets short we may have to skip the recommendation phase.

 

References

ALLA – FILL IN