Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS WILEL~M B. ROUSE ABSTRACT This paper is concerned with the conceptual design of support systems for humans in advanced manufacturing systems. A design methodology is presented that includes five major steps or phases: characterizing users' tasks, assessing demands of tasks, identifying approaches to support, determining likely obstacles, and anticipating user acceptance problems. This methodology is discussed in the context of its previous applications in aerospace and process control domains and its potential application in manufacturing. INTRODUCTION This paper considers a methodological framework based on an integrated view of human decision making in complex systems and explores its applicability to advanced manufacturing. This framework, which can be of great value for identifying potential problems and likely solutions in the design of advanced manufacturing systems, has been formalized and applied over the past 5 years (Rouse and Rouse, 1983; Rouse et al., 1984; Rouse, 1986~. This framework has been used for almost 20 years in aero- space and in process and power systems, as well as for several applications in public service systems. Many of the problems associated with the human role in complex systems are due to the technology-driven nature of most sys- tem development efforts. The "technology spiral" shown in Figure 1 illustrates the cen- 148 tral phenomena in technology-driven de- velopments (Rouse, 1985~. An understand- ing of these phenomena helps to explain why the human role in complex systems becomes confusing. The use of advanced technology is usu- ally motivated by perceived performance or productivity requirements or the availabil- ity of new technology. Although the infu- sion of technology is the "standard" solution to most problems, the added technology may overwhelm the operators and manag- ers who must work with it and those who must maintain it. This situation can be il- lustrated by the following examples. Pilots of F-15 fighter aircraft and B767 commer- cial aircraft can be overwhelmed by the number of modes of the radar system and the flight management system. In process and power systems, operators can be con- fused by the many failure modes of the au- tomation that is supposed to help them. On
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS 149 Performance/ Productivity Requirements \ Operators Overwhelmed by Alternatives \ Increased Automation \ FIGURE 1 The technology spiral. a high-tech injection molding control sys- tem, the introduction of color graphic dis- plays and keyboards changed a relatively easy job into one that was quite difficult for the operators. Finally, of course, most of us have had frustrating experiences dealing with the multitude of cryptic commands for word processing and spreadsheet pro- grams, with the typical result being that we learn a minimal set of commands and avoid the rest. The plethora of functionality that can overwhelm an operator can also be a prob- lem for those who must maintain the sys- tems. More technology packed in smaller Designers Hi/ Infuse Technology / ~ \ MaintaIners Overwhelmed by Complexity 1 Desl~ners Produce Technology "Fixes" Increased ATE \~ Performance/ Productivity Shortfall Technology " Opportunities" Managers Overwhelmed by Dam \ Increased M15 boxes in tighter spaces presents obvious dif- ficulties in access and installation. Beyond these traditional maintainability problems, however, the complexity of advanced hard- ware and software requires that the trou- bleshooter have sophisticated knowledge and skills. This problem is exacerbated when complex subsystems are combined to produce integrated systems, such as for avi- onics or the multivariable control of large systems. The result is that increased main- tenance hours are required per hour of op- eration, with a corresponding reduction in operational readiness or plant availability. Although managers seldom directly op-
150 orate and maintain complex systems, they are greatly affected by the infusion of tech- nology into those systems. Computer and communications technologies allow collec- tion, compilation, and transmission of enormous amounts of data. Much of this is not filtered and sampled in ways that pro- vide managers with exactly the information they need. It is often unclear what infor- mation is needed to manage a complex sys- tem. As a result, everything imaginable and measurable is compiled, and managers tend to receive large amounts of data but little information. The problems are evident in many do- mains; the issue is not whether they are real, but how to solve them. As shown in Figure 1, the technology-driven approach to system development dictates that these problems be "fixed" with more technology. The tendency is to eliminate operators by automation, robotize maintainers by auto- matic test equipment (ATE), and procedur- alize managers by intelligent management information systems (MIS). As might be expected, however, these types of technological fixes create the same types of problems as the technology infu- sions that produced the need for the fixes. As a result, performance and productivity shortfalls emerge, in part from technologi- cal problems and in part from inflated ex- pectations. Although it might be imagined that this situation would cause the overall approach to be reconsidered, the more common response is to use the shortfall as a basis for reformulating requirements while also searching for new technological oppor- tunities. Of course, this response ensures that the spiral continues! Technology should be viewed as a means rather than an end in itself. From this per- spective, the system development process should be objectives-driven, with technol- ogy considered only after objectives and re- quirements are formulated. Further, the design objectives should be oriented toward providing means to help users achieve the WILLIAM B. ROUSE operational objectives for which they are responsible; i.e., they should be user- oriented. Few people will disagree with this as a philosophical position. Everyone wants his or her system to be ergonomically designed, user-friendly, and so forth. However, many people believe that user-centered, objec- tives-driven design is not practical. To an extent, this point of view is due to a percep- tion that one must first make sure that the technology works. Of course, by the time this assurance is received, a de facto com- mitment has usually been made to the tech- nology of concern. Beyond this preoccupa- tion with technology, there is also the feeling that the design tools and methods do not exist for achieving the well-intended objective of being "user-friendly." This crit- icism is well-founded since, until recently, human factors engineers, ergonomists, and engineering psychologists, among others, have had virtually no methodologies to con- tribute to the conceptual design phases of system development. The framework out- lined in this paper is offered to ameliorate this deficiency. BACKGROUND A review of the current state of the art in advanced manufacturing systems and the salient issues being debated in the manufac- turing community provides a context for some of the earlier statements while also motivating many of the methodological is- sues that are discussed later. State of Me Art Twenty years ago, when an undergrad- uate student was provided the opportunity in a manufacturing course to program a numerically controlled (NC) milling ma- chine, it was viewed as a high-tech adven- ture. Since then, NC evolved into com- puter-controlled NC (CNC), while com- puter-aided design (CAD) was evolving to
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS include computer-aided manufacturing (CAM) and eventually CAD/CAM with an implicit direct link to CNC. Somewhere along the way, robots emerged to help with assembly and material handling, and this, when combined with CNC or CAD/CAM, led to flexible manufacturing systems (FMSs). Finally, combining all of these technologies with networked data bases for engineering, production, purchasing, and so forth, led to computer-integrated manu- facturing (CIM) and the "lights-out factory of the future." With these technological trends and the proliferation of acronyms came promises that manufacturing productivity and com- petitiveness would soar, leading to impres- sive "bottom lines" in the near future. As might be predicted from the earlier discus- sion, technology infusion was viewed as the panacea. Although progress has been impressive, it certainly does not meet many expectations. It appears that CAD is well ensconced, in- tegrated CAD/CAM is still in development, computer-aided engineering (CAE) is still in the laboratory, and CIM faces Great challenges in implementation (Blumenthal and Dray, 1985~. High downtime is a gen- eral problem (Shaiken, 1985), and inherent software incompatibilities across functional areas are impeding CIM (Conaway, 1985~. "Islands of automation" are emerging (Lowndes, 1985), just when they might be expected to disappear, and plans for totally automated factories are running behind schedule (Lowndes, 1986~. Nevertheless, solid technological progress is being made (Lowndes, 1985; Parks, 1987), and the technology spiral appears to be function- ing. Management Issues Aside from the incremental progress in technology, advanced manufacturing tech- nology presents many other issues. One of these is how this technology should be con- 151 sidered within the strategic plans of the company. The normal staff level approach to strategic planning appears to have prob- lems with the technological discontinuity that can result when implementing the new technology. Much more than in the past, technology may have to be acquired from outside sources and adapted to particular applications, while also adopting a much longer term perspective on productivity im- provements and return on investments (Ayres and Miller, 1983~. It also appears that a "champion" is necessary if such rela- tively radical changes are to become part of the strategic plan (Meredith, 1986~. For advanced manufacturing technology to be effective, it may be necessary to re- think organizational structures. Traditional managerial and organizational contexts can impede innovation (Davis, 1986~. Decisions concerning the degree of centralization ver- sus decentralization can also be affected. In an analysis of the applicability of military command and control concepts to FMS or- ganization, it was noted that decentralized structures tend to minimize the information processing requirements for each individual in the structure, while centralized struc- tures can more easily recover from de- graded operations (Armstrong and Mitch- ell, 1986). A particularly interesting trend is the im- pact of advanced manufacturing technol ogy on the role of middle management. Im- plementation of the new technology has tended to broaden the scope of jobs on the shop floor to include planning, diagnosing, operating, and maintenance (National Re- search Council, 1986) or, using somewhat different descriptors, analysis and program- ming (Fraser, 1986~. As a result, operating decisions are being delegated to the work- ing level (Fraser, 1986, National Research Council, 1986; Parks, 1987). With this ad- ditional responsibility and authority, as well as the requisite information, work teams are tending to displace middle management (National Research Council, 1986; Pola-
152 koff, 1987~. It has been suggested that this dramatic reduction in the role of middle management in information compilation and management need not eliminate this level of personnel if their responsibilities can be shifted to emphasize communicating with employees, directing preventive main- tenance, and improving quality control (Polakoff, 1987~. Until recently, applications of computer technology in manufacturing have been fo- cused on reducing direct production costs. As a result, direct labor costs are becoming a decreasing percentage of total costs. The emphasis is therefore now shifting to using automation to cut overhead costs, including materials, which can account for 75 to 90 percent of the total cost (Dornheim, 1986; Lowndes, 1986~. The implication of this trend is that the roles of both white-collar and blue-collar workers are being affected by advanced manufacturing technology. Human Resources Issues From the shop floor to upper manage- ment, humans should be viewed as re- sources within manufacturing sYstems. rather than potential Luddites. It is not simply a matter of being humane and so- cially minded. A lack of consideration of human resources issues can undermine the implementation of advanced manufactur- ing technology (Davis, 1986; Shaiken, 1985~. Further, a recent study found that innovative human resources practices are often associated with what are judged to be successful implementation efforts (National Research Council, 1986~. It has been ar- gued that reluctance to embrace the tech- nology could be assuaged by a general hu- man resources policy that lessened uncertainty about job security by, for ex- ample, guaranteeing retraining (Ayres and Miller, 1983~. The issue of retraining is particularly problematic. There is widespread agree- ment that reeducation and retraining are WILLIAM B. ROUSE essential to any human resources plan (American Management Association, 1986a, 1986b; Ayres and Miller, 1983; Mar- gulies, 1985; Salvendy, 1985~. It is argued that such training efforts are important for shop personnel, middle management, and all other levels including upper manage- ment (American Management Association, 1986a). Although there appears to be a consensus on the need for training, the success of such efforts depends on there being an adequate population from which to select trainees. With the new technology, manufacturing jobs are changing to require more cognitive and reasoning abilities as well as levels of literacy that surpass mere reading and writ- ing (American Management Association, 1986b). It has been estimated that less than 50 percent of current workers have the ap- titudes to be trained for the new jobs (Sal- vendy, 1985~. Considering new entrants into the manufacturing work force, the ma- jority will be immigrants for whom at least the literacy requirements may present problems (American Management Associa- tion, 1986b). Ironically, at the same time that knowledge and skill requirements are outpacing the population of workers, many highly skilled machinists, albeit with differ- ent skills, are being relegated to the status of assembly-line workers for the purpose of monitoring CNC machines, robots, and FMS systems (Shaiken, 1985~. These are important problems that require further ex- amination and study (Committee on Sci- ence, Engineering, and Public Policy, 1987~. A final human resource issue concerns the resistance to change. It has been ob- served that the introduction of CIM systems has encountered stiff resistance (Blumen- thal and Dray, 1985~. Even well-intended, user-centered approaches to design and im- plementation have encountered substantial reluctance (Margulies, 1985~. A useful in- sight is that change is a problem only in that it presents uncertainty (Nadler, 1986~.
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS Human resource policies such as those dis- cussed earlier may help to decrease uncer- tainty (Ayres and Miller, 1983~. However, at the level of detail necessary for system design, much more concrete measures and methods are needed. An approach to antic- ipating and dealing with user acceptance problems is outlined in a later section of this paper. System Design Issues As noted in the introduction to this pa- per, there have been many proponents of user-centered system design in the manu- facturing domain, but little methodology has been developed for pursuing this philos- ophy (Blumenthal and Dray, 1985; Brod- ner, 1985; Margulies, 1985~. Nevertheless, various research efforts have provided in- sights and potential elements for more com- prehensive methodologies. Several related sets of guidelines have been developed that provide a taxonomy of the abilities and limitations of humans and computers, with particular emphasis on ro- botics and FMS applications (Hwang et al., 1984; Kamali et al., 1982, Nof et al., 1980~. The primary use of these guidelines is in allocating functions between humans and computers. These types of guidelines were popular in aerospace applications in the 1950s and 1960s and a bit later in process control. It was found, however, that such taxonomies provide, at best, only a first step in resolving potentially difficult function al- location trade-offs. More recently, fairly so- phisticated methods have been developed for function allocation (Rouse and Cody, 1986~. Design procedures are needed that pro- vide a nominal sequence, but not a "cook- book," of issues to be resolved and decisions to be made that go beyond simple guide- lines. Representative of a step in this direc- tion in manufacturing is a series of model- based analyses of FMS scheduling. These efforts have included an analysis of the types 153 of decision making in FMS scheduling (Fraser, 1986), the development of a hier- archical approach to integrating human ca- cabilities into the decision-making process (Ammons, 1985), and an analytical decom- position of FMS cell schedule management and inventory management into six super- visory control subfunctions (Mitchell et al., 1986~. Methods and models should, to the ex- tent possible, be evaluated relative to em- pirical data. Such data, unfortunately, are scarce for advanced manufacturing sys- tems. Two studies of the use of computer- generated displays for FMS control con- cluded that the effects of the displays were much more subtle than anticipated (Mitch- ell and Miller, 1983; Sharit, 1984, 1985~. Computer-generated information displays do not necessarily improve the performance of the operators and can degrade perfor- mance, unless additional assistance is pro- vided for using the information in this form. Similar results have been found for proce- dure displays in flight management (Rouse and Rouse, 1980; Rouse et al., 1982) and searching aids in data-base retrieval (More- head and Rouse, 1983~. The allocation of responsibilities for machines between hu- man and computer in FMS control has also been studied (Hwang, 1984~. Not only are modest amounts of data available, but they are also often plagued by considerable variability. Further, the trends in the data often defy intuition. It appears that this is a result of the nature of the problems being studied, as opposed to the skills of the investigators. It is simply not possible to study a system as compli- cated as FMS control by merely choosing naive experimental subjects and observing how they respond in a low-fidelity simula- tion. To be realistic and provide meaning- ful results, much more job training and aid- ing should be provided. In fact, the design and evaluation of such training and aiding are much more interesting and important issues than the effects of display formats,
154 graphics, color coding, etc. (Govindaraj and Mitchell, 1985; Hwang, 1984; Mitchell and Miller, 1983~. Summary of Issues The "factory of the future" is economi- cally appealing although perhaps socially unsettling. The technological challenges ri- val, and may exceed, those in aerospace and process systems. Because of the com- plexity and costs involved, humans will still have many, if not more, roles in manufac- turing systems. Fewer people may be doing more, which would seem to portend in- creased productivity. However, the level of understanding of the role of the human in manufacturing, as well as the tools and methods for supporting those roles, is se- verely limited. The remainder of this paper outlines an approach for helping to over- come this deficiency. A FRAMEWORK FOR USER-CENTERED DESIGN The important motivations for the user- centered design point of view fit in two classes: (1) human abilities and (2) human inclinations. Human manipulative skills continue to surpass those of automation, particularly in adaptability and flexibility, as contrasted with precision and consis- tency. Considering the ways in which ad- vanced manufacturing technologies are af- fecting jobs, the perceptual, judgmental, and creative abilities of humans are likely to be more important than manipulative skills as reasons for humans to retain central roles in manufacturing. Various theorists and practitioners in ar- tificial intelligence continue to assert that computers will eventually supplant almost all human manipulative, perceptual, judg- mental, and creative abilities. Even if this is true, which I strongly doubt, the incli- nations of humans are such that they will continue to have important roles in com- WILLIAM B. ROUSE plex systems. In particular, the inclination of humans to make commitments and ac- cept responsibility for their actions, as well as the actions and well-being of others, make them unique relative to imaginable hardware and software alternatives. Thus, I believe that user-centered design must not be considered a "holding action" until total automation is possible. There are four key conceptual elements of the framework for user-centered design. The initial element relates to understanding user-system tasks and characterizing these tasks by means of a constrained set of ter- minology. This terminology provides direct links to the next two conceptual elements, which relate to identifying means for en- hancing human abilities and overcoming human limitations for the tasks of interest. The fourth and final element relates to fos- tering user acceptance of the means identi- fied for enhancing abilities and overcoming limitations. These four conceptual elements are the basis for the design methodology summarized in Figure 2 and described in the remainder of this section. Characterizing Users' Tasks Traditional approaches to describing users' tasks include direct observations and interviews for tasks where job incumbents are available and analytical decomposition of task scenarios for those that are new and not directly analogous to similar existing tasks. The value of these methods is limited during the early stages of design when a system concept is not yet available. There is a need for general procedures by which designers can choose a subset of tasks that appear to be the most important for the application domain of concern. By pro- ceeding in this way, the roles for the hu- mans become the first issue studied rather than the last. With a goal of developing a general set of tasks, 120 publications that reported on decision-making and decision-support sys-
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS Characterize users' tasks L l , Assess relative demands of tasks Identify potential approaches to support l 1 Determine likely applications obstacles _ _ r l l Anticipate user acceptance problems FUGUE 2 User-centered design methodology. tems, primarily in the aerospace industry, were reviewed (Rouse and Rouse, 1983~. All of these studies were concerned with one or more of the general tasks shown in Figure 3. These can be summarized as fol- lows: Situation assessment is concerned with the formulation of the problem or deciding what is happening; planning and commit- ment are concerned with devising a solu- tion to the problem or deciding what to do about it; and execution and monitoring in- volve implementing the solution or plan and determining whether the consequences of 155 the solution will be acceptably close to the desired results. As one might expect, Figure 3 is much too simple, particularly in relation to the interactions among tasks. Figure 4 is a re- finement and expansion of the simpler rep- resentation. First, it reflects the fact that the a priori situation and human stereo- typical plans and expectations govern much of behavior. Most situations and subsequent behaviors are fairly routine and, fortu- nately, considering the effort involved, sit- uation assessment as well as planning and commitment need not be invoked. Occa- sionally, the consequences of actions devi- ate significantly from expectations, and routine behaviors are insufficient. Humans then must be concerned with explaining the observed deviations and choosing among alternative courses of action, whereupon the situation may, or may not, revert to routine. Another important distinction shown in Figure 4 is the differentiation of task behav- iors from the user-computer interface. User- centered design is much more than the analysis and synthesis associated with the Sltuatlon assessment Planning and commitment Execution and monitoring FIGURE 3 Three general user-system tasks.
156 Routing and \ monitoring Updating ~ ... expectations (, outing Planning] and commitment \ chosen /~ F]GURE 4 Relationships among user-system tasks. hardware and software for displays and controls. Although these aspects of design are important during the later, more de- tailed stages of design, they will not be ad- dressed in this paper, which is primarily concerned with the support of a structured approach to problem formulation and re- quirements analysis. The process shown in Figure 4 can be further decomposed into the set of 13 tasks in Figure 5. This set of tasks was sufficient to classify and describe all decision-making and decision-support efforts reviewed (Rouse and Rouse, 1983~. It is reasonable to ask, however, whether this set of tasks is unique to the aerospace applications from which it emerged. WILLIAM B. ROUSE A prlorl slSuatlon, plan, and expectations Execution | ___ \` Interface ,/ / Devlatlon \ acceptable / \ Situation | ~ assessment ~ ( computer ) I ~ Interface / 1 ,~- ___ - \ computer ~ In an effort to test the applicability of this task taxonomy to other problems, it was used to describe a set of decision-support systems in the process control domain (Rouse et al., 1984~. Two independent an- alysts reviewed the documentation for sev- eral existing support systems and classified the functionality of these systems using Fig- ure 5. The results of these two independent analyses for the process control example were virtually identical. Although the ap- plicability of this task taxonomy to both aerospace systems and process control does not ensure its applicability to discrete parts manufacturing, my perception is that this taxonomy is applicable to advanced manu- facturing systems. The accuracy of this per-
THE HUMAN ROLE IN ADVANCED MANUFAC TURING SYS TEMS ception can be finally demonstrated only through repeated testing of this approach in a variety of manufacturing circum- stances. Two characteristics of Figure 5 are of particular importance. First, most of the tasks involve generation, evaluation, and selection among alternatives. The use of this standard terminology will be shown to be helpful for identifying approaches to en- hancing abilities and overcoming limita- tions in these tasks. The second noteworthy characteristic of Figure 5 is the emphasis on alternative interpretations of deviations, in- formation sources, explanations, and courses of action. Thus, the structure of Figure 5 is Execution and Monitorlng 1 1. Implementation of plan 2. Observation of consequences 3. Evaluation of devlaUons from expectations 4. Selection between acceptance and relectlon Sltuatlon A~ment: Inform. Icon Seeking 5. GeneratlorUldonUticatlon of altemaUve Inforrnatlon sources 6. EvaluaUon of altemaUve Infonnatlon sources 7. Selection among' altemaUve Information sources Sltuatlon A_ment: Explanation 8. Generation of altemaUve explanations 9. EvaluaUon of altemaUve explanations 10. Selection among altemaUve explanations Planning and Commitment 11. Generation of altemaUve courses of action 12. EvaluaUon of altemaUve courses of action 13. Selection among altemaUve courses of action FUGUE 5 Subtly of genera mar-system toy. 157 based on both the process depicted in Fig- ure 4 and a three-by-four array of action words and objects of action. These comple- mentary methods of organization bring an important degree of structure to the process of characterizing the tasks of the users. Assessing Relative Demands of Tasks It could easily be argued that every task involves all of the elements shown in Figure 5. Although this may be true, such a con- clusion does not help in identifying the "bottlenecks" in user-system performance. Table 1 was prepared as an aid for deter- mining which tasks are likely to present problems. This table is meant to prompt the thinking of the analyst rather than re- place it. Context-specific knowledge should, of course, preempt any of the assessments in this chart. To illustrate the use of Table 1, it will be useful to consider the emerging roles of the FMS operator. In his or her original job as a machinist, execution and monitoring were dominant, with occasional elements of the other tasks noted as "moderate" in the op- erations column of Table 1. More recently, CNC, CAD/CAM, and FMS have caused the machinist's job to include more of a maintenance role, involving aligning parts, clearing debris, and dealing with snags in the flow of parts and materials (Shaiken, 1985~. From Table 1, one can see that this lessens the relative demands for some of the more cognitive tasks, thereby increasing the requirements for manipulative and percep- tual skills as contrasted with judgmental and creative abilities. It would appear, however, that the even- tual role of the FMS operator will increase the demands on judgmental and creative abilities, in particular by shifting the em- phasis to elements of planning and commit- ment (Ammons, 1985; Fraser, 1986; Mitch- ell et al., 1986~. Thus, the emerging roles of the FMS operator would appear to in- clude some aspects of the roles of manage-
158 TABLE 1 Relative Demands of User-System Tasks WILLIAM B. ROUSE Type of User User-System Tasks Operations Maintenance Management Design Execution and monitoring Implementation High High Low Low Observation High Moderate Low Low Evaluation Moderate Moderate Moderate Moderate Selection Moderate High High Moderate Situation assessment: Information seeking Generation Low Low High Moderate Evaluation Moderate Low High High Selection Moderate Moderate Moderate High Situation assessment: Explanation Generation Moderate Moderate High Low Evaluation Low Moderate High Low Selection Low Moderate Moderate Low Planning and commitment Generation Low Low High High Evaluation Moderate Moderate High High Selection Moderate Moderate Moderate Moderate ment and design. My perception is, how- ever, that the projected demands for situation assessment indicated for the man- agement and design roles shown in Table 1 will be "moderate" rather than "high" for the EMS operator. This illustrates the ear- lier point that context-specific knowledge should preempt the entries in Table 1. This brief analysis suggests that technol- ogy will initially move the work content for the machinist toward that of lower-level positions, comparable to that of assembly workers (Shaiken, 1985), but that these po- sitions will eventually evolve to a higher level, such as FMS cell supervisors (Am- mons, 1985; Fraser, 1986; Mitchell et al., 1986~. This trend will obviously require great flexibility from the workers. More central to the theme of this paper, however, is the conclusion that the changing roles of these workers dictate changing approaches in assisting them to achieve their opera- tional objectives. Identifying Approaches to Support At this point in an analysis, Figure 5 and Table 1 have been used in conjunction with domain-specific knowledge to identify one or more tasks that appear most in need of support. The design team could now sit around a table and brainstorm to produce support concepts. However, such an ap- proach would ignore the thousands of pre- vious efforts to develop support systems. What is needed is an easy method of access- ing these previous efforts. Figures 6 and 7 were synthesized from ongoing reviews of hundreds of support sys- tem development and evaluation projects as well as many years of experience in support systems R&D. Although these tabulations are useful for prompting ideas, they are also used to access a card file and, subsequently, a small library of documents on support systems. This method of identifying and re- trieving information could easily be com-
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS GENERATION OF ALTERNATIVES o For a Gwen situation, a support system might retrieve previously relevant and useful alternatwes. O For a given set of attributes, a support system might retrieve candidate alternatwes with these attributes. 0 Given feedback with regard to suggested alternatives, a support system might adapt Rs search strategy and/or tactics. EVALUATION OF ALTERNATIVES 159 o For a Gwen alternative, a support system might assess the alternatwe's a priori characteristics such as relevance, intorrnation content, and resource requirements. 0 For a grven situation and alternative, a support system might assess the degree of correspondence between situation and alternatwe. o For a Gwen alternatwe, a support system might assess the likely future consequences such as expected impact and resource requirements. o For given multiple alternatwes, a support system might assess the relatwe merits of each alternatwe. o Given feedback of appropriate variables, a support system might adapt its evaluations in terms of time horizon, accuracy, etc. SELECTION AMONG ALTERNATIVES 0 0 FIGURE 6 Alternative approaches to supporting generation, evalua- tion, and selection of user systems puterized, but the investment in such an effort cannot be justified until there are more users of this information. Many of the entries in Figures 6 and 7 are self-explanatory, but a few clarifica- tions are needed. Although the entries con- cerned with the generation of alternatives appear straightforward, the suggested sup- port is difficult to provide (Madni et al., 1985) in that there are few previous efforts to draw upon. This difficulty appears to be due, for the most part, to the difficulty in specifying the attributes that are desired. Some progress has been made in using For given criteria and set of evaluated alternatives, a support system might suggest the selection that yields the ~best. allocation of human and system resources. For Gwen individual differences and time-var~tions of crRena, preferences, and evaluations, a support system might adapt Rs suggestions to reflect these variations. pattern-recognition methods to infer attri- butes of desired alternatives from a set of examples (Freedy et al., 1985; Morehead and Rouse, 1985~. It is fairly straightfor- ward to retrieve the examples if users can define them appropriately e.g., byrequest- ing information on all flight directors for jet fighter aircraft or all of the types of robot manipulators that are currently available. The evaluation of alternatives is easy to understand in that this type of activity is common in engineering analysis. The feasi- bility of supporting evaluation depends on the availability of appropriate models, cal-
160 WILLIAM B. ROUSE INPUTS TO THE USER o For Gwen information, a support system might transform, format, and code the information to enhance human abilities and overcome human limitations. 0 0 For a given set of evaluated information, a support system might filter and/or highlight the information to emphasize the most salient aspects of the information. For a grven sample of information, a support system might fit models to the information in order to integrate and interpolate within the sample. For given constraints and indw~ual differences, a support system might adapt transformations, models, etc. OUTPUTS FROM THE USER o o For a Gwen plan and information regarding the user's actions, a support system might monitor implementation for inconsistencies and errors of omission and commission. For a grven plan and information regarding the user's actions and intentions, a support system might perform some or all of the implementation to compensate for the user's inconsistencies, errors, or lack of resources. o Gwen information on intentions, resources available, priorities, etc., a support system might adapt its monitoring and/or implementation. FIGURE 7 Alternative approach to supporting the input to users and the output from the users of systems. culation techniques for the alternatives, and measures of interest. Finite difference meth- ods and geometric modeling techniques in CAD/CAM/CAE represent evaluation sup- ports for designers. Spreadsheet models pro- vide similar support for managers. The majority of previous efforts to de- velop support systems have focused on the selection among alternatives, in part be- cause this type of support is most tractable. If all of the alternatives have been specified and the probability distributions associated with the consequences of choosing each al- ternative are known and the decision mak- ers' criteria can be assessed, it is usually easy to determine the best or optimal alterna- tive. For alternatives involving multiple stages, locations, or the like, this optimiza- tion problem is less straightforward but can, nevertheless, be treated using standard con- trol theory and operations research tech- niques. Although the techniques used to support selection among alternatives are important, experience suggests that identi- fication of feasible alternatives and their likely consequences is often sufficient for decision makers to choose immediately with- out resorting to optimization. Thus, despite the great attention that has been given to selection among alternatives, this task is usually not the most difficult task faced by humans. Good support for generation and evaluation is typically more important but unfortunately is seldom available. Although the support of generation, evaluation, and selection is central to this user-centered design framework, these types of support are not sufficient for a compre- hensive approach to user-centered design. Figure 7 provides guidance in choosing ap- proaches for supporting inputs to the user and outputs from the user. With regard to inputs, display design has long been the stock in trade of human factors engineers.
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS Fairly recently, methods based on expert systems technology have been developed for on-line, intelligent information manage- ment. These methods have the potential to filter, transform, and format information automatically. This capability will, for ex- ample, make it feasible for operators of complex systems to cope with the enormous amounts of data available through com- puter and communications technologies (Rouse et al., 1986~. On the output side, it is feasible to have a computer monitor ac- tion sequences for reasonableness and con- sistency. An example of this support con- cept is an onboard information system for aircraft that was developed and evaluated a few years ago (Rouse et al., 1982~. More recently, this concept has been generalized to a comprehensive architecture for error- tolerant interfaces (Rouse and Morris, 1987~. To summarize this discussion of identify- ing approaches to supporting users, all of the vast support system literature describes theories, design concepts, and evaluative results that relate to one or more of the capabilities summarized in Figures 6 and 7. Support concepts for selection and input are the most common; concepts for evaluation and output are not uncommon; and con- cepts for generation are fairly rare. TABLE 2 Obstacles to Application of Support Concepts ]6] Determining Likely Obstacles Although there is a wealth of support concepts, there are several types of obsta- cles that can limit feasibility of applying these ideas. Table 2 provides some guidance for identifying likely obstacles. Potential problems with the nature of the knowledge base relate to the extent to which it is likely to be difficult to understand and perhaps model or capture knowledge and skills for each type of user. Obstacles associated with the nature of the interaction between the user and the system are related to the diffi- culties of developing flexible and fast intel- ligent interfaces. There are also risks asso- ciated with inadequate resolution of the knowledge base and interaction problems. The consequences are usually immediate and can be traced for operations and main- tenance. For management and design they are typically delayed and are not traceable. Although Table 2 is not comprehensive and does not provide guidance for over- coming major obstacles, this compilation of experiences is useful for determining the scope required for a support system devel- opment effort. In other words, Table 2 can be used as a guide for budgeting resources in anticipation of likely obstacles. As noted before, this type of chart should be used to Type of User Potential Problems Operations Maintenance Management Design Nature of knowledge base Lack of structure Low Low High Moderate Less than comprehensive Low Moderate High High Inaccessible Moderate Moderate High Moderate Nature of interaction Heterogeneity of users Low Moderate High Moderate Unacceptability of prescriptions Moderate Low High High Real-time requirements High Moderate Low Low Nature of risks Potential immediate misfortune High Moderate Low Low Potential long-term misfortune Low Moderate High High Lack of traceability Low Low High Moderate
162 prompt questions rather than as a replace- ment for thinking about the specifics of the application of interest. Anticipating User Acceptance Problems The review of R&D efforts in support systems suggested that few support system concepts are ever used operationally. Stud- ies are performed, reports are written, and the effort is frequently terminated. In at- tempting to understand this situation fur- ther, it was found that many of the support systems that are fielded encounter indiffer- ence or opposition from the personnel for whom the support is intended. This finding led us to a detailed examination of user ac- ceptance (Rouse and Morris, 1986~. There appear to be four determinants of user acceptance. Most obvious is the users' perceptions of the impact of the support system on the quality of their job per- formance. In other words, does the system work as advertised and, if so, does this functionality help the individual user? Another fairly obvious dimension is the per- ceived ease of use. Specifically, is the per- ceived effort required to learn and use the support system outweighed by the per- ceived benefits? The remaining dimensions are more sub- tle. One of these is the perceived impact on desired levels of discretion. In particular, is the support system such that users retain the desired opportunities to exercise their manipulative, perceptual, judgmental, and creative skills? The final dimension is per- ceived peer group and organizational atti- tudes toward the support system. Users are more likely to accept a support system if their colleagues and their supervisors extol its features and benefits. The development of this four-dimen- sional view of user acceptance led to the development of a structured approach for anticipating user acceptance problems and attempting to resolve them in parallel with the overall design process. This structured WILLIAM B. ROUSE approach is shown in Figure 8. The process starts with a set of candidate functions in which automation or other technology in- fusions appear feasible and warranted. The process then proceeds through the struc- tured set of considerations shown in Figure 9 to prune and modify the candidate func- tions, while also preparing and involving eventual users in planning and implement- ing the changes. The basis for the entries in Figure 9 is discussed in detail elsewhere (Rouse and Morris, 1986~. Within the scope of this paper, it is sufficient to indicate the availability of a method for dealing with the important problem of user acceptance. Summary of Framework The user-centered design framework pre- sented in this section is basically a struc- tured approach for characterizing users' tasks, assessing relative demands, identify- ing approaches to support, determining Candidate functions 1 Front-end analysis Automation decisions ~ 1 Implementing change FIGURE 8 Fostering user acceptance.
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS 163 FROr4T-END ANALYSIS 1. 2. 4. Characterize the functions of interest in terms of whether or not these functions currently require humans to exercise significant levels of skill, judgment, and/or creativity. Determine the extent to which the humans involved with these functions value these opportunities to exercise skill, judgment, and/or creativity. Determine If these desires are due to needs to feel in control, achieve self-satisfaction in task performance, or perceptions of potential inadequacies of automation techn~c~av in terms of nil~l~v of performance and/or ease of use. , ~ .,., A. If need to be in control or self-satisfaction are not the central concerns, determine ~ the perceived inadequacies of the automation technology are well founded; H so, eliminate the functions in question from the candidate set -- if not, provide demonstrations or other information to familiarize personnel with the actual capabilities of the automation technology. AUTOMATION DECISIONS 5. 6. To the extent possible, only automate the system functions that personnel in the system feel should be computerized or computer aided (i.e., those for which they are willing to lose discretion). To the extent necessary, particularly H number 5 cannot be followed, consider increasing the level and number of functions for which personnel are responsible so that they will be willing to delegate the functions of concern (i.e., expand the scope of their discretion). 7. Assure that the level and number of functions allocated to each person or type of personnel form a coherent set of responsibilities, with an overall level of discretion consistent with the abilities and inclinations of the personnel. 8. Avoid automating functions when the anticipated level of performance is likely to result in regular intervention on the part of the personnel involved (i.e., assure that discretion once delegated need not be reassumed). IMPLEMENTING CHANGE 9. 10. Assure that all personnel involved are aware of the automation effort and what their roles will be after the change. Provide training that assists personnel in gaining any newly required abilities to exercise skill, judgment, and/or create ty and helps them to internalize the personal value of having these abilities. 11. Involve personnel in planning and implementing the changes from both a system-wide and individual perspective, with particular emphasis on making the implementation process minimally disruptive. 2. Assure that personnel understand both the abilities and limitations of the increased automation, know how to monitor and intervene appropriately, and retain clear feelings of still being responsible for system operations. FIGURE 9 An approach for anticipating and avoiding user acceptance problems. likely obstacles, and anticipating user ac- ceptance problems. Various subsets of these five components of this methodology have been applied to several applications in the aerospace and process control domains. They are now being used for several new applications in these domains. The framework is basically a user-
164 centered approach to problem formulation and requirements analysis. Once the pro- cess outlined in this paper is completed, there is still a tremendous amount of design work left to be done. This work includes development of the functional architectures of the support system concepts chosen, in- tegration of these architectures into the overall manufacturing system architecture, and completion of the detailed design. Methodological support for the more de- tailed aspects of the user-system design has been developed and is compatible with the methods presented here. The outputs of the process described in this paper serve as in- puts to more detailed design procedures (Rouse et al., 1984; Rouse, 1986~. CONCLUSIONS Manufacturing is undergoing a meta- morphosis, more slowly than anticipated, but nevertheless it is progressing. Associated with this substantial change are three im- portant trends. First, information technol- ogy is replacing physical technology as the central concern. Second, as a result of the first trend, software is replacing hardware as the key to productivity. Third, cognition and reasoning abilities are replacing senso- rimotor skills as the raison d'etre for the human role in manufacturing systems. Sim- ilar trends are evident and more mature in the aerospace domain and, to a lesser ex- tent, the process and power industries. Thus, the management, human resources, and design issues that these trends pose for the manufacturing industry are neither novel nor unique. The industry, therefore, can benefit from some well-reasonecl tech- nology transfer. This technology should not be limited to hardware and software. There is a strong need for methodology in general, and user- centered design methodology in particular. Clearly, traditional methods of ergonomics and safety engineering are no longer suffi- cient, and often not even appropriate, for WILLIAM B. ROUSE understanding and supporting the human role in advanced manufacturing systems. This paper has suggested a new way of looking at user-system problems, as well as a structured approach for resolving these problems. Although tailoring may be nec- essary to fit the manufacturing context, it appears that transferring this technology to manufacturing would be a good first step. REFERENCES American Management Association. 1986a. Report of the manufacturing council. AMA Council Reports (Winter) :5-6. American Management Association. 1986b. Report of the manufacturing council. AMA Council Reports (Summer) :9. Ammons, J. C. 1985. Scheduling models for aiding real time EMS control. Pp. 185-189 in Proceedings of the 1985 IEEE International Conference on Sys- tems, Man, and Cybernetics. Armstrong, J. E., and C. M. Mitchell. 1986. Organi- zational performance in supervisory control of flex- ible manufacturing systems. Pp. 1437-1442 in Pro- ceedings of the 1986 IEEE International Conference on Systems, Man, and Cybernetics. Ayres, R. U., and S. M. Miller, eds. 1983. Robotics: Applications and Social Implications. Cambridge, Mass.: Ballinger. Blumenthal, M., and I. Dray. 1985. The automated factory: Vision and reality. Technology Review Qanuary) :28-37. Brodner, P. 1985. Qualification based production: The superior choice to the "unmanned factory." Pp. 18- 22 in Proceedings of the 1985 IFAC Conference on Analysis, Design, and Evaluation of Man-Machine Systems. Committee on Science, Engineering, and Public Pol- icy. 1987. Technology and Employment: Innova- tion and Growth in the U.S. Economy, R. M. Cyert and D. C. Mowery, eds. Washington, D.C.: Na- tional Academy Press. Conaway, J. 1985. Integrated data flow in CIM sys- tems. IE News on Computer and Information Sys- tems 20:1-4. Davis, D. D., ed. 1986. Managing Technological In- novation. San Francisco: Jossey-Bass. Dornheim, M. A. 1986. Airframe makers expect com- puter techniques to cut overhead costs. Aviation Week & Space Technology (December 22):56-59. Fraser, J. M. 1986. Effects of flexible, computerized manufacturing systems on decision making. Pp. 1303-1306 in Proceedings of the 1986 IEEE Inter-
THE HUMAN ROLE IN ADVANCED MANUFACTURING SYSTEMS national Conference on Systems, Man, and Cyber- netics. Freedy, A., A. Madni, and M. Samet. 1985. Adaptive user models: Methodology and application in man- computer control. Pp. 249-293 in Advances in Man- Machine Systems Research: 2, W. B. Rouse, ed. Greenwich, Conn.: JAI Press. Govindaraj, T., and C. M. Mitchell. 1985. Decision support systems for real time control of flexible manufacturing systems. Pp. 56-58 in Proceedings of the 1986 IEEE International Conference on Sys- tems, Man, and Cybernetics. Hwang, S. 1984. Human supervisory performance in flexible manufacturing systems. Doctoral disserta- tion, Purdue University. Hwang, S., W. Barfield, T. Chang, and G. Salvendy. 1984. Integration of humans and computers in the operations and control of flexible manufacturing systems. International Journal of Production Re- search 22:841-856. Kamali, J., C. L. Moodie, and G. Salvendy. 1982. A framework for integrated assembly: Humans, au- tomation, and robots. International Journal of Pro- duction Research 20:431-448. Lowndes, J. C. 1985. Lockheed installs advanced fa- cilities as part of factory modernization. Aviation Week & Space Technology (May 27~:113-118. Lowndes, J. C. 1986. Management by computer promises unprecedented productivity gains. Avia- tion Week & Space Technology (December 22~: 50-55. Madni, A., M. Brenner, I. Costea, D. MacGregor, and F. Meshkinpour. 1985. Option generation: Problems, principles, and computer-based aiding. Pp. 757-760 in Proceedings of the 1985 IEEE Inter- national Conference on Systems, Man, and Cyber- netics. Margulies, F. 1985. Flexible automation: New options for men, economy, and society. Pp. 14-17 in Pro- ceedings of the 1985 IFAC Conference on Analysis, Design, and Evaluation of Man-Machine Systems. Meredith, J. R. 1986. Strategic planning for factory automation by the championing process. IEEE ences. Transactions on Engineering Management EM- 33:229-232. Mitchell, C. M., and R. A. Miller. 1983. Design strat- egies for computer-based information displays in real-time control systems. Human Factors 25:353- 369. Mitchell, C. M., T. Govindaraj, O. Dunkler, S. P. Krosner, and J. C. Ammons. 1986. Real time scheduling in EMS: A supervisory control model of cell operator function. Pp. 1443-1448 in Proceed- ings of the 1986 IEEE International Conference on Systems, Man, and Cybernetics. Morehead, D. R., and W. B. Rouse. 1983. Human- computer interaction in information seeking tasks. I65 Information Processing and Management 19:243- 253. Morehead, D. R., and W. B. Rouse. 1985. Computer- aided searching of bibliographic data bases: Online estimation of the value of information. Information Processing and Management 21:387-399. Nadler, G. 1986. Breakthrough thinking for the inte- gration engineer. Industrial Engineering (Decem- ber:~22-25. National Research Council (NRC). 1986. Human Re- source Practices for Implementing Advanced Man- ufacturing Technology. Washington, D.C.: Na- tional Academy Press. Nof, S. Y., J. L. Knight, and G. Salvendy. 1980. Effective utilization of industrial robots: A job and skills analysis approach. AIIE Transactions 12:216- 225. Parks, M. W. 1987. Expert systems: Filling the miss- ing link in paperless aircraft assembly. Industrial Engineering January):37-45. Polakoff, J. C. 1987. Will middle managers work in the "factory of the future?" Management Review (January) :50-51. Rouse, S. H., and W. B. Rouse. 1980. Computer- based manuals for procedural information. IEEE Transactions on Systems, Man, and Cybernetics. SMC-10:506-510. Rouse, S. H., W. B. Rouse, and J. M. Hammer. 1982. Design and evaluation of an onboard computer- based information system for aircraft. IEEE Trans- actions Systems, Man, and Cybernetics. SMC- 12:451-463. Rouse, W. B. 1985. The role of human factors in military R&D. Pp. 167-178 in Using Psychological Science: Making the Public Case, F. Farley, and C. H. Null, eds. Washington, D.C.: Federation of Behavioral, Psychological, and Cognitive Sciences. Rouse, W. B. 1986. Design and evaluation of com- puter-based decision support systems. Pp. 259-284 in Microcomputer Decision Support Systems: De- sign, Implementation, and Evaluation, S. J. An- driole, ed. Wellesley, Mass.: QED Information Sci- Rouse, W. B., and W. J. Cody. 1986. Function allo- cation in manned systems. Pp. 1600-1606 in Pro- ceedings of the 1986 IEEE International Confer- ence on Systems, Man, and Cybernetics. Rouse, W. B., and N. M. Morris. 1986. Understand- ing and enhancing user acceptance of computer technology. IEEE Transactions on Systems, Man, and Cybernetics SMC-16:965-973. Rouse, W. B., and N. M. Morris. 1987. Conceptual design of a human error tolerant interface for com- plex engineering systems. Automatica 23:231-235. Rouse, W. B., and S. H. Rouse. 1983. A framework for research on adaptive decision aids. Technical Report AFAMRL-TR-83-082. Wright-Patterson Air
166 Force Base, Ohio: Air Force Aerospace Medical Re- search Laboratory. Rouse, W. B., R. A. Kisner, P. R. Frey, and S. H. Rouse. 1984. A method for analytical evaluation of computer decision aids. Technical Report NUREG/ CR-3655; ORNL/TM-9068. Oak Ridge, Tenn.: Oak Ridge National Laboratory. Rouse, W. B., N. D. Geddes, and R. E. Curry. 1986. An architecture for intelligent interfaces: Outline of an approach to supporting operators of complex systems. Pp. 914-920 in Proceedings of the 1986 National Aerospace Electronics Conference. WILLIAM B. ROUSE Salvendy, G. 1985. Human factors in planning ro- botic systems. Chapter 32 in Handbook of Indus- trial Robotics, S. Y. Nof, ed. New York: Wiley. Shaiken, H. 1985. The automated factory: The view from the shop floor. Technology Review (Janu- ary): 17-24. Sharit, J. 1984. Supervisory control of a flexible man- ufacturing system: An exploratory investigation. Ph.D. dissertation, Purdue University. Sharit, J. 1985. Supervisory control of a flexible man- ufacturing system. Human Factors 27:47-59.
