Proceedings of the Human Factors and Ergonomics Society Annual Meeting

In this paper, a theoretical framework for interface design for complex systems is proposed. The approach, called Ecological Interface Design (EID), is based on the skills, rules, knowledge framework of levels of cognitive control. The fundamental goal of EID is to develop interfaces that provide the appropriate support for all three levels, but that do not force cognitive control to a higher level than the demands of the task require. The framework, consisting of a set of prescriptive design principles, is discussed, and an example of its application is presented.

A major goal of the DECwindows program is to provide a consistent, state-of-the-art user interface for workstation software.¹ This interface extends across operating systems and many different types of application programs. Within the DECwindows program we have addressed both the technical and organizational aspects of developing consistent user interfaces across applications. Traditional methods for developing user interface consistency, such as the use of an interface style guide and toolkit, were supplemented with more innovative techniques. An exhibition and catalog of DECwindows application designs helped to develop a DECwindows school of interface design. Electronic conferencing software played an important role in facilitating communication among DECwindows contributors throughout the company. Preliminary user interviews suggest that the DECwindows interface style gives a consistent, usable feel to Digital's workstation applications.

Recent research has established the importance of defining usability requirements as part of the total requirements for a system. Instead of deciding in an ad hoc manner whether or not a human-computer interface is usable, measurable usability requirements are established at the outset. It is common to state such requirements in an operational manner: U% of a sample of the intended user population should accomplish T% of the benchmark tasks within M minutes and with no more than E errors. The formal experiments needed to test compliance with the requirements makes this method costly. This paper presents an alternative method of specifying usability requirements currently being developed and tested on a large software project at Virginia Tech.

Briefly, usability requirements are specified by having every member of the software design team and the user interface design team specify the ease of use desired for each proposed functional requirement of the system under development. The individual ratings are then compared in order to arrive at a consensus. It is this consensus that leads to the formal usability requirements which the interface must meet or exceed. As the interface is built, it is rated in the same manner as that used originally to specify the requirements.

This method thus provides a structured means of specifying measurable usability requirements and a means of determining whether or not the interface satisfies those requirements. Several other benefits of this method are presented as well.

Over the past four years, the User Systems Engineering Group and the Industrial Systems Division of Texas Instruments have worked jointly on designing and developing a computer aided software engineering (CASE) tool to enhance the productivity of process and control engineers. This paper discusses the analysis and design procedures employed during this project and how these procedures focused on customer needs. Specifically, the areas addressed are the multidisciplinary design team, the task analysis, minimal-constraint analysis, and prototype design and iteration.

“…providing electronic equipment accessibility for individuals with disabilities who have special needs is an idea whose time has come” (GSA/ED Guidelines, 1987). The objective of the panel is to provide information to the Human Factors community on (1) the new GSA/ED Accessibility Guidelines with which all government procurements initiated after September 30, 1988 shall comply, and (2) selected issues related to accessibility of electronic equipment by those with disabilities.

This paper presents results from two experiments which compared gestural and keyboard interfaces to a spreadsheet program. This is the first quantitative comparison of these two types of interfaces known to the author. The gestural interface employed gestures (hand-drawn marks such as carets or brackets) for commands, and handwriting as input techniques. In one configuration, the input/output hardware consisted of a transparent digitizing tablet mounted on top of an LCD which allowed the user to interact with the program by writing on the tablet with a stylus. The experiments found that participants were faster with the gestural interface. Specifically, subjects performed the operations in about 72% of the time taken with the keyboard. In addition, there was a preference for the gestural interface over the keyboard interface. These findings are explained in terms of the fewer number of movements required to carry out an operation with the gestural interface, the greater ease of remembering gestural commands, and the benefits of performing operations directly on objects of interest.

The term “direct manipulation” (or DM) often evokes images of interfaces which are intuitive, obvious, and easy to learn. We conducted an experiment to determine whether subjects could learn to use a DM interface without instruction, i.e., whether they could learn the interface syntax on their own merely by inspection and exploration of the interface. The research vehicle was a prototype DM applicationdesigned to allow end users to customize a telecommunications application. Three variations of the interface were created by manipulating elements of the DM syntax, specifically, moded operations and rules about selectingobjects before acting on them. Subjects carried out a set of five tasks in the presence of an experimenter, who was allowed to provide structured help when the subject could not make further progress. Results indicated that the syntax manipulations affected both the number and type of user errors and the amount of help needed to complete the tasks: the use of modes and selection rules significantly interfered with learning, and only four subjects out of thirty were able to perform the complete set of tasks without experimenter assistance. We also found, however, that more than half of the errors made by subjects were not directly related to syntax manipulations. These errors appear to stem more from conceptual problems, i.e., mismatches between the user's developing model of the interface and the model instantiated by the interface designer in the rules of interaction. These conceptual problems were observed across syntax manipulations and represent a significant portion of user's difficulties in learning the interface. Thus, our results shed light on the relationship between interface syntax, learning and usability in the DM paradigm, but they also point out the need for a cognitive account of the processes by which users acquire knowledge of interface characteristics and how that knowledge is related to interface design elements.

Experienced and inexperienced computer users used seven different interaction methods to specify dates of events. Key results were that the three entry methods were faster, more accurate, and preferred over the four selection methods—by both experienced and inexperienced computer users. The rank order of performance with these methods was about the same for both groups. Number of keystrokes required by each method was a good predictor of performance time. For selection tasks, decomposing them into separate fields is advisable.

Five different human-computer interaction techniques were studied to determine the relative advantages of entry-based and selection-based methods. Gould, Boies, Meluson, Rasamny, and Vosburgh (1988), found that entry techniques aided by either automatic or requested string completion, were superior to various selection-based techniques. This study examines unaided as well as aided entry techniques, and compares them to selection-based methods. Variations in spelling difficulty and database size were studied for their effect on user performance and preferences. The main results were that automatic string completion was the fastest method and selection techniques were better than unaided entry techniques, especially for hard-to-spell words. This was particularly true for computer-inexperienced participants. The database size had its main influence on performance with the selection techniques. In the selection and aided-entry methods there was a strong correlation between the observed keystroke times and the minimum number of keystrokes required by a task.

Two experiments were conducted to investigate display variables likely to influence the effectiveness of computer-based procedure displays. In Experiment 1, procedures were presented in three formats, Text, Extended-Text, and Flowchart. Text and Extended-Text are structured prose formats which differ in the spatial density of presentation. The Flowchart format differs from the Text format in both syntax and spatial representation. Subjects were required to use the procedures to diagnose a hypothetical system anomaly. The results indicate that performance was most accurate with the Flowchart format. Although overall completion times did not differ significantly across formats, the Flowchart format required significantly less time for step implementation. In Experiment 2, procedure window size was varied (6-line, 12-line, and 24-line) in addition to procedure format. In the six line window condition, Experiment 2 replicated the findings of Experiment 1. Procedures in the Flowchart format were completed with greater accuracy than procedures in either of the text formats. As predicted, completion times for Flowchart procedures decreased with increasing window size; however, accuracy of performance decreased substantially. Implications for the design of computer-based procedure displays are discussed.

The goal of this research was to evaluate the effects of hypertext as an intra-document text format. Subjects read two articles on a color VDT, one in linear and one in hypertext format. Half of the subjects read general interest articles with instructions for casual reading and the other half read technical articles with instructions to learn the material. Afterward, subjects were given free recall tests, question probe tests, and a preference questionnaire. Results indicated that for both types of article, the linear format resulted in greater memory for the basic ideas contained in the articles. Also, for the general interest articles, subjects reading in linear format assimilated more of the macro-structure than subjects reading in the hypertext format. Consistent with these findings, a majority of the subjects expressed a preference for the linear format and perceived it as requiring less mental effort than reading hypertext. These results suggest that, at least under some circumstances, hypertext can interfere with normal cognitive processes.

This research examined the effects two different cognitive individual differences (perceptual speed and spatial scanning) on information retrieval performance under two matched and two mismatched database format/query conditions. A graphic and a tabular form of an airline database were constructed, along with questions that required users to search through the database to determine the correct response. Two types of questions were designed - graphic and tabular. The data indicate that users are faster when the format of the information in the database matches the type of information needed to answer the question and that cognitive individual differences are differentially predictive of performance in the matched and mismatched conditions. Recommendations for database design are presented.

Although hypertext and hypermedia have been discussed for a number of years, it is only recently that such systems have become generally available to interface designers. A user base of thousands exists at the present time, the most widely used system being HyperCard, from Apple. Hypermedia systems differ from traditional software systems in the complexity of its domains of access, its reliance on visual presentation, and in the use of sound. These differences are beginning to revise previously held ideas and guidelines about the processes of designing and evaluating software interfaces. This session presents some of the approaches to coming to grips with this problem… a problem that will continue to grow as technological advances continue in the development of readily available hypermedia software on inexpensive hardware platforms.

Now that large stores of information in a variety of media can truly be at our fingertips, how do we know what we have and how to get at it? Hypermedia databases present new challenges to interface designers whose goals are to make such databases truly accessible and easy to use. In this paper, using HyperCard (1987) as an example, I discuss some of the user interface issues that arose as we observed children and adults interacting with computerized hypermedia environments. And I describe some design principles and examples that have emerged from our work in the Human Interface Group at Apple Computer, Inc.

User-system interface (USI) guidelines are emerging as a tool for user interface design. The roles of levels of abstraction (concrete and abstract) and of guideline presentation medium (hard copy or on-line) on detection of USI guidelines violations in user-system interface evaluation were investigated. Overall, less than 50% of the guideline violations were detected by software engineers. Abstract guideline violations were more difficult to detect than concrete guideline violations, and on-line presentation resulted in relatively higher guideline usage during evaluation than hard copy presentation. It was concluded that improvements in USI guideline form and media are required to make guidelines a more useful USI design tool.

DoD-HDBK-761 “Human Engineering Guidelines for Management Information Systems” was extensively updated and revised to reflect 1) significant changes in computing technology, including user interface techniques and display technology, and 2) recent user computer interface (UCI) design literature. The document was updated based on literature reviews, mail-out surveys of UCI designers and users, and interactive computer interviews (using question and answer dialog). The updated document contains two main sections. The first presents a general process for conducting UCI analysis, design, development, and evaluation activities. These activities include; conduct of system functions analysis, function allocation, user task analysis, user surveys, use of UCI prototypes, and product testing. The second section contains over 1000 human engineering guideline statements which can be applied to the detailed design of UCIs. The guideline section addresses dialog design, computer control, data entry and display, job performance aiding, expert systems interface design, and data communication and protection. This paper discusses the content and organization of the document.

Following a brief review of how the human factors community achieved its present focus and circumstance, the discussion deals with what future preoccupations and opportunities may face human factors practitioners, such as computer software and documentation evaluations, the development of standards and guidelines, the avoidance of exposure to product liability cases, and the presentation of legislative testimony, among other concerns and challenges. Also examined are the roles that human factors funding and organizational structures may play in facilitating or inhibiting the effectiveness of human factors research and acceptance of the results obtained.

The purpose of this study was to evaluate the effects of spacing and case (use of upper- or lowercase letters) on search time for menus presented on a contemporary high resolution PC display. Twenty Bellcore staff members, all experienced computer users, were presented two sets of four menus each (representing the combinations of case and spacing). Search time was recorded and participants were asked for their preferences among the four menu styles. Analysis of variance results indicated that double-spacing yielded significantly shorter search times in both menu sets. Although case was significant only for the first menu set, a significant case by spacing interaction for the second set indicated that double-spacing had more effect on search time for uppercase options than for lowercase options. Preference data also indicated that 85% of the participants preferred the double-spaced menus.

Personal-computer applications-software often requires people to navigate and select options using their keyboard's function keys where context-dependent meanings for these keys are assigned by guides or menus labeled on the screen. The physical layout of function keys on standard PC-compatible keyboards differs from the most common layouts of screen labels. This study examined user performance consequences of this simple, spatial, inconsistency. In a simulated order entry task, 36 participants each completed 240 trials, 40 with each of six different combinations of two keyboards and three screen guides with different spatial arrangements of function keys and screen labeling. One keyboard used the standard 5×2 function key pad and one used a single horizontal row of function keys; the screen guides were either a horizontal row, a vertical list, or a grid consistent with the standard key pad. We collected measures of response time, errors, and user preferences. Analysis of errors showed no reliable results. Analysis of response times showed several significant effects. Responses were faster with the two combinations of key pad and screen-guide layouts that were spatially consistent than with the four inconsistent layouts. Response times were also faster with the keyboard with horizontal function keys than with the standard layout, and slower with the vertical screen guide than with either of the other two guides. Over 80% of the participants thought the task was easiest when the screen guide matched the function key layout.

Modal block clustering (MBC) is proposed as an approach more suited to the derivation of menu structures than hierarchical clustering techniques. Problems with the application of hierarchical techniques and pairwise similarity ratings (PWSR) from which the clusters are derived are discussed. MBC defines clusters based on the pattern of common command attributes and provides an objective way to determine the composition and number of menu panels to include in a menu structure. The method also objectively defines command redundancy for the menu panels. The method of MBC was applied to the 97 commands that comprise the CMS operating system resulting in 17 menu categories. The menu categories were used to design a help menu system. The MBC procedure provides a viable methodology for complex systems, such as CMS, which derive increased functionality from numerous command options. System designers can fruitfully and efficiently apply this methodology both to current systems and to proposed systems for which there are no expert users.

A study was conducted to determine how well the display format effects described by Tullis (1983, 1984) and the resulting prediction equations could be generalized to other display situations. Task complexity and visual monitoring load were identified as task variables which could potentially moderate the format effects and, thus, were varied factorially. The current study also sought to extend Tullis's findings to tasks where the use of several pieces of information from predictable display locations is required. In general, the data indicate the need to study Tullis's format dimensions more fully before using his regression equations to evaluate display designs for use outside the task situation in which the equations were developed. Also, subjects were unable to evaluate their performance accurately under alternative display designs. Their evaluations seemed to be determined mostly by the perceived ease with which information was extracted from the display. This outcome should serve as a warning to system designers. That is, empirical human performance research should be conducted when performance is the paramount design criterion and a validated prediction system, such as the one developed by Tullis for search, is not available.

New PC based design and prototyping tools are making it easier for user interface designers to rapidly implement prototypes of new user interfaces for complex systems. We used such a tool to design the user interface for an on-line data base that was being developed to replace an existing batch system. The tool not only made the design of the user interface easier, but it also allowed us to present a prototype of the system to end users and to conduct design reviews in an interactive fashion. This allowed us to actively involve users in the design of the user interface and resulted in a better design that met both the users requirements and our own internal design goals.

Rapid interface prototyping (RIP) involves the simulation of potential user-interface designs for demonstrating and evaluating design concepts and iteratively modifying the interface designs without the burden of labor-intensive code generation and modification. In the past few years, many interface designers have had to use expensive equipment to help them perform RIP. Well, thanks to Bill Atkinson and the folks at Apple Computer Inc., you can purchase HyperCard® ¹ software for your Macintosh® ² for $49 and use it right out of the box for many of your RIP projects, without need of additional hardware (just add a cup of creativity). A special feature of HyperCard® allows the designer to create software-controlled buttons which, along with the graphics capabilities and the layering properties, permit the dynamic simulation of virtually any control panel that can be operated by discrete user inputs. Changes to the prototype can be made very rapidly (almost instantly) by an experienced HyperCard® user, making it possible to modify an existing prototype “on-line” in a design meeting. This paper will describe this serendipitous application of HyperCard®, its potential as an equipment-interface design tool, and describe how it was used to simulate the user interface for a weapons field tester, designed at Sandia.

For many years the User Systems Engineering Group at Texas Instruments has been using software user-interface prototypes as a primary tool for the design and development of user interfaces for complex systems. Our model of the prototyping life cycle is introduced and its benefits are discussed. Current issues in prototyping often center around the introduction of new prototyping tools such as HyperCard™ or Prototyper™. Our discussion focuses on how these tools fit within the prototyping life cycle.

Although the use of a prototype of the user interface of a software application early in the development cycle is a valuable tool in the design of a usable user interface, prototyping can be difficult to introduce into the development process. Furthermore, designers without experience in using a prototype can run into problems which counteract the value of this user interface design methodology.

Designers with experience in a substantial prototyping effort have formulated guidelines for the use of prototyping which can help to minimize these problems.

Three training methods were compared in a computer text editing situation. Training varied on the degree of constraint imposed on the behavior of learners. For complex methods, completely constrained and completely unconstrained training situations led to worse test perfomance than an unconstrianed training condition in which subjects were informed about their compliance with a method and allowed to reset the system at will. The results argue against a strict “training wheels” approach to learning environments but support the general notion of “guided exploration,” especially for complex methods.

Twelve subjects who are not touch typists, but have a median of ten years experience using computer keyboards performed two writing tasks: writing a short memorized passage and copying a four paragraph passage. Subjects performed each task once using a pen and paper and once using a display oriented text-editor. Typing speed was over five words per minute (wpm) faster than handwriting for both memorized and copied passages. Typing and writing were each about ten words per minute faster from memory than from copy. The number of errors was greater when typing from copy than in any other condition. These results suggest that for experienced two-finger typists, typing from a display-oriented document processor can be faster than handwriting.

The present research attempts to identify factors which account for cognitive skill acquisition at the higher end of the cognitive performance curve and understand differences between expert and super-expert skill levels on a cognitive oriented task of computer program modification. Subjects completed two computer program modifications and a battery of five cognitive ability tests. Results indicate that super-experts used a global search strategy to obtain abstract information while experts utilized a minimal “directed” search to obtain task specific information. Results are discussed in relation to knowledge representation and current taxonomies of skill acquisition.

Most software engineers have weak backgrounds in areas where human factors engineers are strong: task analysis, applied psychology, and empirical evaluation. Software engineers can become better builders of user interfaces if they are instructed in these techniques, and that is the main goal of the Software Engineering Institute's graduate curriculum module on user interface development. Parts of the module provide instruction about when, where, and how consultants such as human factors engineers can contribute to the design and evaluation of user interfaces. It is critical that human factors engineers understand how they can contribute during development, so that they can have the greatest positive impact.

The study focuses on the identification of the underlying representational properties of human problem solving and their application to expert systems. In this study the interaction between problem representation (procedural, conceptual, unstructured) and problem type (transformation, arrangement, inducing structure) was observed. The results of this study indicate partly that quantitative and qualitative differences in problem solving performance can be attributed to the form of knowledge representation employed by the problem solver. It is suggested that expert systems could be implemented with different shells or structures according to problem characteristics.

This paper presents the perspective of a user system knowledge continuum which recasts traditional user system components (user interface, context-sensitive help, completion aids, manuals, training) as interrelated knowledge components tasked with appropriately distributing required knowledge between user and system. It suggests that maximizing user system effectiveness is best viewed as optimization of a set of such knowledge components. The paper relies upon a case study showing this perspective at work in the development of APT - Applications Productivity Tool™, an integrated software environment for industrial automation applications.

Through the use of AI technology, the potential exists to make revolutionary changes in military command and control. Knowledge based systems can enable a commander and staff to accommodate large amounts of information efficiently and effectively, to generate operations plans rapidly, and thereby allow exploitation of early maneuver opportunities. This paper addresses the issues of application domain and human factors design practice as they relate to the development of effective user system interfaces for expert systems.

In the supervisory control of a complex, dynamic system, one potential form of aiding for the human operator is a computer-based operator's associate. The design philosophy of the operator's associate is that of “amplifying” rather than automating human skills. In particular, the associate possesses understanding and control properties. Understanding allows it to infer operator intentions and thus form the basis for context-dependent advice and reminders; control properties allow the human operator to dynamically delegate individual tasks or subfunctions to the associate. This paper focuses on the design, implementation, and validation of the intent inferencing function. Two validation studies are described which empirically demonstrate the viability of the proposed approach to intent inferencing.

This panel discusses the human interface considerations associated with hypertext. Panelists will present empirical, observational, or anecdotal data based upon their experiences with hypertext systems and users.