Artificial Intelligence for
Employee Surveys

by
Norman D. Costa, Ph.D., President
eXpert Survey Systems, Inc.
and
William E. Dodd, Ph.D., Vice President
eXpert Survey Systems, Inc.

The next generation of employee opinion surveys will use artificial intelligence to solve morale problems and add value to the management of employee relations. This new generation of survey processors will go beyond displaying percents, norms, and graphs. Using a subset of artificial intelligence systems, known as expert systems, employee surveys will inform management about: what their problems are, where they are, what the probable causes are, recommended actions to take, and what additional information or tutoring might bring an understanding of the problems.

Expert systems, also known as knowledge-based and ruled-based systems, will also speed up the total survey process and dramatically cut expenses. For example, expert systems will relieve supervisory and middle management of the enormous information processing burden involved in analyzing survey results, and will do it quickly and accurately. This enormous economy, in time and money, will be achieved by presenting employee survey results in as final a form as possible. It will obviate the need for time-consuming analysis, and the time it takes to train management to do the job properly.

In another example, expert systems can short-circuit the up-front process of determining the issues and boundaries for survey administration. The labor-intensive (and expensive) process of interviewing and moderating focus groups, coupled with committee and management conferences, can be sharply reduced in time and cost. An expert system can very quickly propose recommendations for the inclusion of survey topics and item wording, sampling versus census polling, what groups to survey, the level of management to receive analytic reports, minimum group size for analysis, and what data to feed back to participants.

What makes an expert system possible and practical for employee opinion survey applications? Two things allow this to be realized: A knowledge base, and advances in desktop computing and software.

KNOWLEDGE-BASED SYSTEMS

A knowledge base is very easy to understand. This is the accumulated experience, wisdom, lore, data, and research results that make up the body of knowledge of any discipline. These data and information reside in several places. They can be in the standard repositories like journals, text books, published articles, and the like.

A knowledge base also resides in the heads of experts. This knowledge base is tapped in several ways. Shadow observers can take notes and record the knowledge that is used by an expert and how it is processed. Research in artificial intelligence for medical diagnosis has used this approach in the past. Another way is to have experts articulate their experiences, organize them in some fashion, and then transcribe them into rules that can be used by computer programs.

Here are several examples of a rule structure which might be a part of an artificially intelligent employee survey system. The first two would apply to typical employee morale-type surveys. The third rule structure is for a survey of company executives prior to the administration and feedback of an employee survey. This is an example of how an expert system might facilitate decision making in the administrative process.

EXAMPLE RULE STRUCTURE I

SURVEY RESULTS: 42% unfavorable responses to, "How would you rate your company as a place to work?"

RULE: This item is an exception if less than 60% favorable or greater than 30% unfavorable.

POSSIBLE CONSEQUENCE: Employees may suffer a loss of esteem and pride in the company.

PROBABLE CAUSE: Recent business performance may have been poor, or lower than in the past.

RECOMMENDED ACTION: Improving your business performance will go a long way to changing employee attitudes. Don't just concentrate on better earnings. Employees will respond positively to new product introductions, territory growth, or new business.

 

EXAMPLE RULE STRUCTURE 2

SURVEY RESULTS: A score of 55 on a scale which measures work place stress.

RULE: This stress scale is an exception if the score is above 45.

POSSIBLE CONSEQUENCE: Employees may complain more of headaches, backaches, digestive upsets, or insomnia.

PROBABLE CAUSE: There may be too many constraints in how employees go about doing their work. They may have little or no say about things that are important to their jobs.

RECOMMENDED ACTION: Give employees more control over how they do their jobs. Allow flexible scheduling of work hours.

 

EXAMPLE RULE STRUCTURE 3

SURVEY RESIJLTS: Executives give a high score to their line supervisors on using employee survey data.

RULE: An acceptable cut-off score is determined for this rating.

POSSIBLE CONSEQUENCE: Employees will react favorably to seeing survey results at the level of their immediate supervisors.

PROBABLE CAUSE: Employees may have had positive experiences with supervisors who shared survey data with them before.

RECOMMENDED ACTION: In a census survey, generate and distribute survey results to the supervisor level.

 

These rule, or knowledge, structures can be accumulated and organized in large data bases and made as general or specific as one likes. They can be tailored to specific companies, or divisions within companies. They can be adapted for various skill groups, product lines, and economic cycles. They can be based on research and empirical evidence, or on the folksy, yet enlightened, anecdotes of experienced practitioners.

ADVANCES IN DESKTOP COMPUTING

The second thing which makes it possible to apply principles of artificial intelligence to employee surveys is progress in desktop computing. On the hardware side, processor speed, system memory, storage media (now measured in multiple gigabytes), and low-cost commodity prices bring the computational burden within the reach of any company.

Recent advances in relational data base management systems round out the software half of the equation. These low-cost desktop software tools make it easier for companies and professional consulting organizations to:

Generate data bases of survey data, item banks, knowledge structures, and professional experiences;
Coordinate data collection and survey administration;
Interrelate all of the relevant data; and
Generate graphical and narrative reports on paper or electronic or magnetic media.
Desktop relational data base management systems do not only handle the numbers and percents of employee surveys, the text of employee comments, and the compilation of item banks and rule structures. They can also handle infinitely large text images like magazine articles and book chapters, graphical images like pictures and charts, and full motion audiovisual images like those on movies and videotape.

A manager can view employee survey results, interactively, from a diskette or CD-ROM provided by the survey processing program. The manager will not only see what the problems are and how to fix them, but could also be offered the option of reading an article or book excerpt tailored to problems manifested in the manager's survey data. The manager could also be given the option of viewing a short film or interactive video tutorial to hone skills for the impending survey feedback meeting. Relational data base management systems can now integrate all these data and image forms.

Expert system applications for employee opinion surveys are not dreams for a distant future in survey technology, awaiting future advances in hardware and software, and eventually lower costs for data processing. The technology and cost reductions have arrived.

The knowledge base is also here, but it awaits our inventiveness in organizing and structuring it. Research and experience have accumulated over the past 60 years covering such concepts as: job satisfaction, the consequences of having too little work to do, why some employees intend to leave their jobs, the consequences of a distrusted management, what happens when you improve hygiene factors. Some of these issues may be addressed with precise results from replicated studies, while others yield to the intuition of skilled practitioners.

Beyond the meaning of results, anyone who has had significant experience with full-scale employee surveys understands that to have a survey energize an organization several things must come together at the same time. Top management must insist on accountability for the results. Middle management must be prepared to carry out a serious auditing role. Supervisors must have effective communications and interpersonal skills. The knowledge bases for how to tune these conditions, although not as widely known, also exist and can be integrated into an expert system for employee surveys.

Finally, the integration of existing technology with existing knowledge bases will benefit consumers of employee opinion surveys. The benefits will be faster service, lower prices, higher quality, and a greater value-add in the management of employee relations and achieving business objectives.