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.
DISCUSSION: DESIGNING FOR THE FACE OF THE FUTURE: RESEARCH ISSUES IN HWM~NCQMPUTER INTERACTION Judith Reitman Olson The hardest part of generating a research agenda now for issues in human-comput~r interaction for the Space Station is not En finding important issues and unanswered questions that are in need of careful research. It is selecting those research issues and the approach-= to them that will answer the questions we have in the year 2000. In the year 2000, we will have devices that we can only dream of today; the Space Station environment will have a mission, size, and complexity that today we can only begin to sketch out. Our job, therefore, is not to rcoammenl a research program that will answer specific questions that we know will arise in the design of the future Space Station. Rather, it is to prepare for that future with a rich plan that lays the foundation, a sound theoretical base, that will make specific results both easy to predict and simple to confirm empirically. Additionally, the research has to produce a development environment, a flexible hardware platform and programming environment, that allows rapid prototyping for empirical testing and easy final implementation. These bases will serve us well when we have to make specific designs for the year 2000. INTERFA~F~ OF THE SYSTEMS OF THE FUTURE It is important to begin by noting those things that are likely to be different in the Space Station environment than they are in the environments we focus our research on now. The most obvious differences, well discussed by both Poison and Hays=, are that the Space Station environment is weightless (with concomitant difficulties in forceful action and countermotion), perhaps noisy (with difficulties for the implementation of speech recognition and sound productions, and complex (with ~ small number of people doing many, varied tasks with the help of cc mputers, some of which they will be expert in, some of which they will not). In addition, the tasks performed In the Snace Station differ in . . . . . . . other, more run~amenta' ways prom the tasks we use today in our laboratory research on human-com~uter interaction. By far the largest amount of current research focuses on the behavior of people doing Operational basks: wordprocessing, spreadsheet formulation and 201
202 analysis, database search ~ support of constellating a report. Our current ~s~s~h focuses on office Lasts. The Space Station, ~ contrast, is likely to have very little need for Emotional t=.=ks; Seward everyday tasks are more likely to be ac ~ Cliched Or go ~ personnel. Space Station personnel are more likely to be involved in: the monitoring and control of onboard systems te.g., life support, experiment/manufacturing control), the occasional use of planning and decision systems (e.g., expert systems for medical diagnosis or for planning for changes in the missions, and m e nearly constant use of communication systems (i.e., for both mission related information and for personal contact with friends and family), for both synchronous conversation and asynchronous messages. ISSUES There are important research issues that we common among these systems and the operational systems that we focus on today, but there are other, additional issues that are unique, requiring particular emphasis. The common issues, important to all future human-computer interaction, include: 1. How to design a system that is easy to learn and easy to ~ e. One core feature of such a system is "consistency". Poleson's paper mates the case for consistencY--a detailed argument for , ~ e , _ e _ e _ _ _ _ e . ~ One Importance or specl~lcally morel Meg the user Is goals and the methods necessary to accomplish the goals with a particular system. This is a very important research approach that promises to give the right level of answers to questions about consistency that will arise in future designs. 2. A second core feature in making a system easy to learn and use involves a straightforward mapping between the way the user thinks about the task objects and actions and the way the system ~ , ~ For example, the mapping between the objects of wordprocess m g, such as letters, words, and sentences, correspond much more closely to the objects in a visual editor than they do to the strings and line objects of a line editor. Moran (1983) has made a beginning in delineating this type of analysis; more theoretical work and em~rri~a~ e ~ e e e Verl; :lCa lOn IS necessary. rcouircs the user to specify them.
203 3. How to make decisions about what modes of 1nput/output (and their combinations) are appropriate for a given environment and task. Hayes' paper discusses a number of considerations that must be taken into account when deciding among speech/visual/keyboard input and output modalities, as well as the use of appropriate combinations of these modalities. 4. What characteristic= of the human information processor are primary determinants of the range of accept~hie interface designs. One way of evaluating a design of an interface is to analyze it on the basis of the major processing that a user engages An On order to understand the output and generate the next input. For example, we can analyze an interface for its perceptual clarity (e.g. adherence to Gestalt pr mciples of grouping for mea m ng), its load on working memory (e.g., how many sub-goals or variables must be retained for short periods in order for users to accomplish their goals), and its requirements for recall from long-term memory (e.g., how many specific rules must be Spurned and how similar they are to each other). This approach, the cognitive science of human-computer interaction, by its generality across all application interfaces, promises to provide a theoretical thread through a number of e~piri~1 investigations. With a body of emp Id tests of its predictions, this approach can both provide a robust base for future design situations and grow in sophistication and precision as a base for understanding complex cognition, even outside the domain of human-computer interaction. Progress on these topics will make substantial contributions to our understanding of how to design human-computer ~nterfa ~ for the Space Station in the year 2000, just as they will for those interfaces in offices and on the factory floor. As discussed above, however, the systems on the Space Station are less likely to include operational systems, like those used in research on the above "common" topics, and more likely to include planning and decision, monitoring and control, and communication systems. Additional, important research issues arise in considering these latter three types of systems: 1. What characteristics of an interface app~vpriat-1y alert users to abnormal situations in systems that must be monitored. What advice, information, or immediate training can be given users of a monitoring system that will guide them to behave in a creative but appropriate manner. 2. How are voice, video, keyboard, pointing devil==, etc. to be used singly and in combination in each of these three types of systems? Certainly voice and video have begun to be explored in synchronous communication systems (e.g., picturephone and siow-scan video teleconferencing). How can these modalities be
204 used to best advantage to support the need for Jong-term contact with friends and family when individuals are separated for a long -~me? How are privacy issues accommodated in such systems, both for personal communication and operational ccmmunication? 3. If users have to consult an expert system or if some intel~iger~ce is incorporated into a system, how is information conveyed to the user about whether the system is to be believed? Since current intelligent systems are "fragile," that is, easily put In situations for which their advice is not appropriate, we need to convey to the ~ er information about the system's boundaries of capabilities. Or, better yet, we need to build intelligent facilities that allow the user to query or access the stored knowledge in ways that can make the advice fit new situations more flexibly. 4. Since the systems that Space Station users must d=~1 with will be varied and the users will have varying expertise On either the task at hand or the particular system to be used, it is important to have the system provide requisite context or train m g. Traim ng need not be a formal docile that one accesses explicitly, as software training mcUNIes are designed today. m e systems could be initially designed to be transparent (i.e. with cbiects and actions that fit the way the ~ , _ user thinks about the task), not requiring training. Or, they could be built to include a "do what I mean" facility or embedded "help" or "training" facilities, accessible either when the user requests it or when the system detects that the user is confused or doing things inefficiently. Most of the current theoretical bases for the design of human-computer interfaces consider tasks that are well-known to the user. The Gods analysis of Card et al. (1983), for example, is for skilled cognition. Kieras and Polson's (1985) production system formalism similarly considers only skilled performance of cognitive banks. However, in the Space Station environment, users will be doing few routine tacks. They will be doing tanks that involve navel situations, situations that invoke creative problem solving, not routine cognitive skill. Space Station personnel, for example, may try to alter a system that their monitor has shown is malfunctioning; they may use the advice of a medical expert system to attend to a colleague who has an undiagnosed iciness; they may ~ e communication channels to a ~ e additional expertise form the ground crews to solve onbcard problems or plan new missions. In order to understand how these interfaces should be designed, more emphasis should be made in research in the area of human problem solving. The focus should be, for example, on how to build systems that, minimally, do not interfere with the information the person needs to keep track of during complex problem solving. Ideally,
205 we want to be able to build systems that augment a person's abilities to explore and evaluate new actions in novel situations. 6. Furthermore, as Hayes' paper pa Mets cut, most of our current research on human-computer interaction focuses on the use of a system ~~ a ['tool" not ~s an Agent Our understanding of cooperative human behavior is woefully thin. Theoretical bases near to be established so that we can build systems that cc operate well with the human problem solver, so that systems can augment the intelligent human to produce an even greater level of understanding and action. APPROACHES FOR THE UNDEFINED FUTURE 1 As stated at the beginning of this discussion, the most difficult aspect of the task of listing research issues that the Space Station of your 2000 will benefit from concerns predicting the Space Station environment and the technology that will be available at the time. We just don't know what the alternative design elements will look like. The best we can do at this time, therefore, is to recommend a research agenda whose results promise to be useful no matter what the environment and technology will be. At the cone of these reoo=mendations is research that ~ nters on the capabilities of the human information processor, both as an individual and ~ a cooperative environment. m e human will not have changed substantially by the year 2000. Consequently, cur understanding of human-computer interaction will benefit from research that accumulate= results frog a common th~nretira~ core that: 1. del~n-=tes in detail the functioning of the human information processor, with particular emphasis on the interaction among cognitive resources and those resources involved in attention (for monitoring systems), problem solving (for expert systems and decision support systems), and communication, 2. within the domain of expert systems, explores the information a user Negro and determines how it should be presented so that the user can assess the believability of the advice given, and deters mes ways to help casual users of a variety of systems to use them without a great d-~1 of "start up" effort, either' through transparent design; effective, easy training; or embedded intelligent aids. A -Client aspect of this type of research is that it IS bared on cognitive models, not on design pr mciples. Cognitive models allow the examination of the interaction of features of the task or interface, which principles cannot do. mese cognitive models characterize
206 details of what the task rehires and details of the human information processor. By running these models, the designer or researcher can de~nine in detail areas of difficulty In the interaction berg., Ante the working Tory is overloaded with subgoals ark pare ~ t ~ s to be retained). Certain changes to the interface design could be tested by running these models without having to invest An the expense of a full-fledged usability study. The number of researchers approaching issues in human-computer interaction with cognitive models is currently very small; their numbers should be encouraged to grow. Furthermore, research should have as one of its goals the transfer of the knowledge developed in the laboratory to the design and development process. This calls for development of: 1. analytic tools for assessing consistency in a particular design. analytic tools for assessing the amount of effort required in mapping the users' natural way of thinking about the task (i.e., an object/action language) into that required by the system, and 3. guidelines that will assist the designer in decisions about which modality or combinations of modalities are appropriate for a particular ta.=k and situation. And, if systems are to be built for an evolving future, they must be built with scars and hooks, as Hayes notes. Software should be designed so that it he places that will allow easy growth in capabilities or input/output devices. Furthermore, research is needed to develop: 1. a method and language that allows the system designer to incorporate good human factors into the target system (e.g., toolkit with components that have been designed with consideration for research on their human factors), and a method that allows system developers to rapidly implement trial 1nterfa~=, so that they can ~ bested with real end-users, and then turned quickly into production code. It is clear frump the papers in this session that funds devoted only to simple emp Apical studies of users' behavior with new, Increasingly complex technology will not be sufficient for answering the questions of the year 2000 and beyond. In contrast, research that focuses on: 1. the abilities of the human information processor with concommitant widespread, specific, robust cognitive modeling, and
207 2. additions to the developrner~t life Cycle to make Jche production of good software rapid can produce research that can make the h~nan~nputPr interfaces on the Space Station of the highest possible quality for their time. REM- Card, S. K., Moran, T. P., ar~ Newell, A. 1983 The Psychology of Han - Carnputer Interaction, Hillsdale, N. J.: Er~amn. Kieras, D. E., arm Polson, P. G. 1985 An approach to the formal analysis of user complexity. International Journal of An - Sac hire Studies 22: 365-394 . ~ran, T. 1983 Getting into a system: external-~nternal task crappie analysis. Pp. 45-49 in Exceedings of the 1983 CHI Conference on Man Factors In Cc~outincr. New York: ACM.
