Expert Systems With Applications, Vol. 10, No. 1, pp. 55-62, 1996 Copyright 1996 Elsevier Science Ltd
Printed in Great Britain. All rights reserved 0957-4174/96 $15.00 + 0.00
A Knowledge-Based Support System for Psychological Analysis
G. DARYL NORD* AND JERETTA HORN NORD
College of Business Administration, Oklahoma State University, Stillwater, OK 74078
Abstract--An expert system was designed and developed to assist a cfinician in the administration and preliminary analysis of the Minnesota Multiphasic Personality Inventory (MMPI) test. The inventory is a standardized test developed in the late 1930s designed to assess attitudes, character, behavior, and in general personality traits of individuals. Today the test is widely used not only for clinical purposes but also for employment screening and research. The expert system encompasses a computer program that: (1) administers the personality inventory via computer, (2) scores the test results based on standard norms, and (3) passes the scores to a rule-based expert system that renders a preliminary psychological profile of the test taker to the clinician. The system can be exploited for its educational value as well as provide valuable assistance to the clinician.
EXPERT SYSTEMS designed to mimic and replace human experts have attracted extensive attention in the past several decades. Their development advanced from efforts of artificial intelligence (AI) researchers to achieve the goal of building computer programs that could "think", that is, solve problems in a way con- sidered by human beings to be intelligent (Holsapple & Whinston, 1987; Pigford & Baur, 1990).
As early as the 1950s, scientists began writing symbolic, domain specific computer programs. Then in 1956, Newell, Shaw and Simon developed an AI program named Logic Theorist that used reverse reason- ing to solve problems in propositional calculus (Olson & Courmey Jr, 1992; Quinlan, 1987). Dendral, a system that identifies molecular structures of materials using mass spectrometer data was later developed in the mid 1960s. AI researchers continued to initiate numerous expert system applications in the 1970s. Mycin, a rule- based system consisting of 400 heuristic rules that diagnoses blood diseases, is one of the more notable projects (Newell, Shaw & Simon, 1963).
Beginning in the 1980s and continuing into this decade, many researchers decided that AI had matured to the point that it belonged in business, government, and related applications. This decision has accelerated the
* Requests for reprints should be sent to: G. Daryl Nord, College of Business Administration, Oklahoma State University, Stillwater, OK 74078-0555. USA.
research and development of AI and specifically expert system technologies.
This paper discusses the development of an expert system designed to test and analyze psychological traits of individuals. First, the domain in the form of the psychological test will be examined; then the expert system development software that was utilized will be discussed; and, finally, the detailed architecture of the expert system will be presented.
2. THE PSYCHOLOGICAL TEST DOMAIN
In the late 1930s, Starke R. Hathaway, a medical psychologist, and John C. McKinley, a neuropsychiatrist, developed a personality study called the Minnesota Multiphasic Personality Inventory (MMPI). The psycho- logical inventory consists of a test of over 500 questions designed to focus and identify specific personality traits of individuals. The test assesses character, attitudes, and behavior by the patterns of responses to true, false, or cannot say questions.
Originally designed to help identify those with psychological problems, today the test is widely used in employment screening, clinical practice, and research. Prospective law enforcement officers, nurses, nuclear power plant personnel, and even the Russian cosmonauts are required to take the MMPI. The test is used in 46 countries around the world and has been translated into 124 languages. MMPI scores have been assembled on nearly every kind of medical and psychiatric population as well as on the young, the old, prisoners, religious
56 G. D. Nord and J. H. Nord
groups, and ethnic subgroups. Additionally, over 5000 published research studies have been conducted on the MMPI (Holden, 1986; Parker, Hanson & Hunsley, 1988).
Items that compose the primary clinical scales of the MMPI were selected essentially by empirical means. Response patterns obtained from criterion groups with known psychological and behavioral characteristics were compared with contrast groups of normal people, medical patients, and persons with various psychiatric diagnoses. Thus, the various scales of the MMPI contain diverse item content; in addition, some test items are included on more than one scale because of the empirical selection procedures that were implemented. The origi- nal test included 13 clinical scale domains briefly described below (Colligan and Offord, 1988):
1. Social Maladjustment. The item content of this scale approximates the concept of introversion-extraversion. High scores tend to indicate that the person is socially bashful, shy, embarrassed, self-conscious, and extremely resereved; and low scores indicate that the person is probably gregarious, confident, assertive, and fun-loving and relates quickly and easily to others.
2. Depression. High scorers have classic depression with feelings of guilt, regret, worry, unhappiness, and loss of concentration and motivation; reduced self- esteem, anxiety, and apprehension are present; these persons are oversensitive and likely to feel misunder- stood, unworthy, and deserving of punishment.
3. Feminine Interest. High scorers describe liking femi- nine games, hobbies, and vocations and deny liking masculine games, hobbies and vocations. High scores also may result from expressing a preference for a broad range of interests and those viewed as socially desir- able.
4. Poor Morale. High scores indicate a lack of self- confidence, feelings of failure and despair, and tendencies to give up hope. Persons with high scores are likely to be oversensitive to the feelings and reactions of others, to feel misunderstood and useless, and to be socially suggestible.
Although there is overlap among Poor Morale, Social Maladjustment, and Depression, Poor Morale empha- sizes a lack of self-confidence and hypersensitivity to the opinions of others, whereas feelings of guilt and apprehension are stronger on Depression and a lack of social poise is prominent on Social Maladjustment.
5. Religious Fundamentalism. A high score indicates that the person is religious and churchgoing and has many fundamentalist beliefs and a view of his or her faith as the true one.
6. Authority Conflict. High scorers tend to view life as a jungle and to view other persons as unscrupulous, dishonest, hypocritical, and motivated only by personal profit. They are likely to be competitive, believe that rules are for others, have little respect for experts, and be generally mistrustful.
7. Psychoticism. Persons with high scores admit to hallucinations, strange experiences, loss of control, classic paranoid delusions of grandeur and persecution, feelings of unreality or daydreaming, and the sense that things are wrong and to feeling misunderstood by others when these views are expressed.
8. Organic Symptoms. High scorers describe symptoms of headache, nausea, dizziness, loss of mobility and coordination, poor concentration and memory, difficulty in speaking and reading, unusual sensations of hearing, smell, and touch, and loss of consciousness, which may be indicative of organic involvement.
9. Family Problems. High scorers believe that their home life is unpleasant and characterized by a lack of love. Their parents are viewed as unnecessarily critical, nervous, quarrelsome, and quick tempered.
10. Manifest Hostility. High scorers admit to sadistic impulses and tendencies to be cross, grouchy, com- petitive, argumentative, uncooperative, socially aggressive, and retaliatory in interpersonal relationships.
11. Phobias. High scorers admit to many fears of the classic phobic variety (e.g. heights, the dark, and close spaces).
12. Hypomania. High scorers are characterized by feel- ings of excitement, well-being, restlessness, tension, enthusiasm, broad interests, seeking change, cheerful- ness, great energy, and difficulty with temper control.
13. Poor Health. High scorers express concern about their health and admit to various gastrointestinal symp- toms, primarily related to an upset stomach and difficulty in elimination.
Over the 57 years of the test's history, dozens of specialized subscales have been added, including sub- scales that evaluate such items as basic trust, violence, dominance, worry and obsession, to name a few. Additionally, there are scales that measure validity factors on the answers to the main items. These validity scales take into account fabrication of test answers, incompetence, and defensiveness that may have an effect on the final test results. Each scale is scored separately but is meaningful only in the framework of the entire profile.
Psychological Analysis Support System 57
FIGURE 1. Model of the PICTURES Expert Assistance System.
3. EXPERT SYSTEM DEVELOPMENT SOFTWARE
In the early development of computer programs for expert systems, it became evident that many of the programming tasks were duplicated in every program, while the coding of the problem-specific knowledge base changed from one program to another (Zahedi, 1993). Over the past 25 years, these tools have evolved from low-level system development shells to high-level knowledge engineering facilitators.
The EXSYS Professional system development soft- ware was used to develop our expert assistance system. EXSYS Professional is a generalized expert system designed for any problem that involves a selection from among a definable group of choices where the decision is based on logical rules. The rules can comprise relative probabilities of a choice being correct. Furthermore, the knowledge management facilities of EXSYS Profes- sional supports an excellent user interface, development tool function, and external program/data interface capa- bility. The external program/data interface component allows the system to communicate and retrieve data from external sources, which was a major consideration in the development of our expert assistance system.
4. ARCHITECTURE OF PICTURES
PICTURES is an acronym for P__ersonality _Inventory Calculated Through the U__se of a Rule-based E_xpert System. Figure 1, exhibits a model of the logical design of the PICTURES expert assistance system. The system (1) administers the psychological test, (2) scores and computes the scores for each of the 41 personality scales included, and (3) analyzes the final test results.
PICTURES includes approx. 500 test questions administered and scored exclusively by means of the computer. A Pascal program was designed, structured, and written that displays each test question on the computer monitor, scores the test, and passes the computed statistical results to the expert system compo- nent of PICTURES. The test taker answers the questions as they appear on the screen, and the individual's responses are scored objectively in accordance with set norms and standards. The raw scores from the standard validity and clinical scales are transformed to linear T- scores (mean = 50; SD = 10) using data provided in the MMPI Manual (Dahlstrom, Welsh & Dahlstrom, 1972; Duckworth, 1980; Graham, 1987; Hathaway & Heehl, 1972; McKinley & Hathaway, 1943). The responses of
the Minnesota normal group provide the basis for the T- score conversions. The normal population, on which the test is based, was represented by a group of about 1400 men and women randomly drawn from Minnesota, Iowa, and Wisconsin in the early 1940s (Colligan & Offord, 1988).
Scoring of the inventory is objective. Standard scoring procedures generate a test profile composed of four validity scales and 37 clinical or personality scales. PICTURES includes the 13 original scales and an additional 24 of the more interesting scales that have been added over the years. The inventory's purpose is to elicit a wide range of self-descriptions from each test subject and to provide, in quantitative form, a set of evaluations of his/her personality status and emotional adjustment (Colligan & Of-ford, 1988; Newell, Shaw & Simon, 1963). A final score is computed for each scale based on the test subject's responses on each inventory item of the MMPI. The formula used to obtain the final score is:
((((subjects score-mean) * 10)/sd) + 50),
where the subject's score is the score acquired on that specific scale, mean is the average score for that scale based on the normal group, and sd is the standard deviation for that scale (Dahlstrom, Welsh & Dahlstrom, 1972; Duckworth, 1980; Graham, 1987).
Table 1 is a complete list of the 41 scales included in the PICTURES knowledge-base assistance system. The scales range from emotional to physical problems, but predominantly evaluate one's personal responses to oneself and to others. These scales measure both positive and negative traits. Each scale is a quantifiable measure of one trait. For example, the positive trait for depression is cheerfulness, while the negative is called severe depression. As can readily be seen from this illustration, each scale is actually two traits. To classify the subject for a scale, a score will be generated from the test that will show whether the subject is ranked low or average or high, as compared to the norm response for that scale.
Figure 2 shows a representative example of an attribute chart generated for each of the scales. The attribute chart illustrates that abnormally low scores and abnormally high scores are usually considered malad- justed, while mid-range scores are considered within the normal range.
58 G. D. Nord and J. H. Nord
TABLE 1 Profile Scales Within PICTURES
Lie Scale Frequency Scale Correction Scale
? Scale Attitude Toward Self
Ambition Basic Trust Depression
Orderliness Originality Paranoia Phobias Prejudice
Psychasthenia Psychopathic Deviate
Quantitative-Linguistic Balance Dominance
Emotional Immaturity Family Problems
Female Masochism Hypochondriasis
Impulsivity Intellectual and Cultural
Manifest Anxiety Masculinity-Femininity
Schizophrenia Self Criticism
Sensitivity Sexual Deviation Social Adjustment Social Introversion
Suicide Key Underachiever-Overachiever
Violence Worry and Obsession
Figure 3 depicts the specific characteristics of one of the 41 scales, the depression scale. The illustrative count- range preceding each choice shows examples of final test scores that will re...