Computers may carry out various forms of information integration, for example, by configuring separate channels of information graphically in the configural or object display. Computers can draw inferences from raw data – using relatively lower levels of intelligence in interpolating or predicting data, or using more sophisticated algorithms and heuristics in the process of diagnosis.
Across all three functions – attention, integration, and inference -human performance data support the conclusion that higher levels of automation, supporting higher levels of situation awareness are of near universal benefit, as long as such automation is reliable. Careful task analysis can insure that highlighting and attentional cueing will guide users to the most relevant information, while low-lighting or backgrounding of less relevant information, will still make that information accessible should it be unexpectedly needed.
Both integration and inference can substantially offload human working memory demands. In operations management automation systems offer positive examples of proposed functions that will integrate information for controllers, and draw inferences about future events (prediction) that can improve decision making, while reducing workload. The scale of automation level of decision choice, which ranges from full authority of humans to consider and select all authorities, to mid-levels in which computers can recommend options, to still higher levels at which computers fully decide.
As with the scale of information integration automation, a key feature with this scale is the reduction of operator workload that is realized at higher levels, along with the possibility that decisions might be made better by computers. Yet research that is both relatively old, in the domain of autopilots and manual control (Madu 1993), as well as that of more recent vintage (MacCarthy 2001), has addressed the concerns that result when such automation at its higher levels fails.
If the decisions made are those with little risk (e. g., the decision to implement one display viewpoint over another), then high levels of automation may suffer only minor consequences even if there are failures. However, if decisions have high risks associated with them, then three sources of human performance deficiencies at high automation levels mitigate against the use of those higher levels: (1) humans become less likely to detect failures in the automation itself, or in the processes controlled by the automation, if they were only passively observing the automation, (2) humans lose some awareness (Neely 2002).
Natural language and expert systems both have significant potential applications for operations management in this decade. The hope for both kinds of systems is that, by allowing people to give commands and communicate with computers in everyday language, widespread use of the computer will be substantially easier. Fewer people would need to learn specialized computer languages, and fewer computer experts would be needed as intermediaries between computers and those who wish to use the computer as a tool.
For example, without a natural language front end, a plant manager who sought the answer to the question, “Which products in the 2000 series were sold in volumes of more than 1,000 last year? ” would probably refer the question to a programmer, who would write a short program in a computer language to process the request. With a natural language feature added to the DBMS, that plant manager could type his request, more or less exactly as he would say it, into a computer terminal, and the requested information would appear on the screen (Nersesian 2002).
Finally, expert systems can allow the use of computers in situations normally thought to be so complex that they require human judgment or “intuition. ” AI researchers have found that in a narrowly defined problem area, it is sometimes possible to simulate much of the judgment of human experts by breaking down that judgment into hundreds of rules for the information to look for under different circumstances, and how to weigh that information.
Though the commercialization of expert systems is only beginning, many industrialists have high hopes for their use in operations management. A myriad of applications have been proposed, including systems which could mimic the performance of a human machinist; systems for advising designers and preventing design errors; systems which would act as a linkage between manufacturing and design data bases; and even systems for overall factory control.
With current high levels of interest in expert systems, and evolving tools and techniques to streamline their development, it seems likely that these tools will be used in several areas of operations management. However, it is unlikely that expert systems will in the near future meet the many expectations which their recent successes have generated. It is easy both to underestimate the development effort and skills needed to construct such tools, as well as to imagine new applications in areas which are too broad or ill-defined for current technology to handle.