Human factorsin engineering

Part IIEAdvanced man-machinesystems and conceptsThe advanc ed manl-machinie svsteims-o01 -line mali-conmptlter intei-actionz, mannied space systems, satelliteconununications, advanced displays-nonie of which ex-isted 20 years ago, present challenzging nlew problenmsto hinmani factors in1vestigators

Nilo L indgren Siaff Writer

This second and concluding article scans over someof the new kinds of human factors problems that haveemerged as a result of recent technological advances.The article discusses aspects of these new problemsin the following major areas: man-computer inter-action (psychological problems related to on-lineusage, the need for more "natural" interactionlanguages, graphical communication); man-vehiclecontrol and guidance systems (better models of humancontroller, novel systems); communications (com-puters in telephone information services, satelliterelays); advanced space systems ("learning" and new intellectual adventure-namely, how to come todisplay problems related to incorporation of com- terms with the computer and make it a more naturalputers in prelaunch activities). Last, a recent study ally; in other words, how to bring the functions of manin human information processing is cited that points and computer into a truly symbiotic association. As Dr.up a general problem-the need for rational optimal J. C. R. Licklider recently described that ally, to thosemodels of human functioning. who have been displaced by the computer, he "appears in

the form of a superhumanly fast, accurate, methodical,The folk wisdom, "Give a little boy a hammer and tireless worker. To most of those who have prepared the

suddenly everything needs pounding," has its modern procedures for computers to execute, [he] appears in thecounterpart in "Give a little boy a computer and form of such a worker who is also illiterate, perverselysuddonly everything needs computing." It is the story literal, devoid of initiative, and so high salaried that oneof our age. In fact, many of the advanced man- must strive continually to keep him from running out ofmachine systems with which human factors investigators work."'must now deal involve computers in some form-as For a normal human to get along with such a creature"workers" with whom communication is necessary, as would appear to demand an inhuman patience andadjuncts or components of huge complex systems, as perseverance. But until very recently, computers weretools for the simulation of devices and of aspects of indi- still so exorbitantly costly in relation to human labor, andvidual and social behavior, and so on. computer designers were scrambling so fast to keep up

with technological developments, pressed on by sharplyMan-computer interaction competitive demands, that the inhuman demand had toThere can be little dispute about the statement that be tolerated to a large extent, and was not attacked in any

the computer is the most fascinating machine man now logical, organized fashion. Now, however, the picture hasinteracts with. It is to the generation of this mid-century changed. In general, computers are becoming so economi-what the heavier power machines were to the mid-19th cal, relatively speaking, that ways must be found ofcentury. It has crashed in on the modern consciousness making better use of the people who use the computers.with the force of a silent atomic explosion, liberating In short, the determination of what the human factorsterrific energy and setting many investigators olf on a problems are in man-computer problem solving and

62 IEEE spectrum APlRIL 1966

fast, men go very slow" is at the crux of the speed-costmismatch. To allow a single man to monopolize the timeof a large, expensive computer is simply not economical-this would be like giving over an entire electric powerplant for the use of one individual-and the larger societyis too tyrannical in its economic structure to tolerate sucha situation, even for "a very creative man." And yet verycreative individuals do feel that this is precisely the bestway to interact with computers-to engage in prolongedand intimate "conversation" with them. By contrast, inmost present man-computer exchanges, there exists along interval between experimental steps because of theneed to reduce all communications to typed "microstate-ments," a process that Dr. Ivan Sutherland compares tothe operation of the postal systeM.3 He says, "We havebeen writing letters to rather than conferring with ourcomputers." This correspondence system has involved anarduous translation process between our natural languageand the programming language. Dr. George A. Miller,the very well known authority in what mighit be calledthe field of "psycholinguistics of man-computer inter-

- z m theimplementaction,"'says of the binary language of computers:"It would be difficult to design a language that was moredifficult or unnatural for human beings to learn to use."14This feature, he believes, poses a major obstacle to man-

computer symbiosis. In effect, the language mismatch_4 . .alreadycompounds the speed-cost mismatch problem.One answer to this essentially multivalued problem of

nonsymbiosis, and one that is interlocked with the"step-display-look" approach that is emerging, has beenthe implementation of the concept of time-sharingwhereby many users at different locations are able to tapthe same central computer "power" source. There arealready in existence in the United States a number oflarge, experimental time-sharing systems; probably thebest known is the Project MAC ("machine-aided cogni-tion" and "multiple-access computer") system at theMassachusetts Institute ofTechnology.366

information p-rocessing has been drawn into the center In time-sharing systems, the speed mismatch betweenof the target. Although it is far beyond the scope of this man and computer is turned to advantage. The comlputerslight survey to go deeply into the fascinating psycho- can work on one "customer's" problem for a short timeve,logical or human factors problems inherent in the man- switch to the next user, thien to the next, and so on, andconiputer interaction situation, we shall at least point out come back to work on the first user's problem so quicklysome of the present major directions, that it appears as though the computer has been working

Just six years ago, in the lead article of the very first continuously on his problem alone.'issue of the IRE TRANSACTIONS ON HUMAN FACTORS IN Perhaps it should not be surprising that in such systemsELETRONICS, it was also Dr. Licklider who set forth the "software" and human factors problems of an entirelybroaed problems of "nman-computer symbiosis," a term new sort are encountered and, typically, the discussers ofand a paper that have since been much cited.2 The major the time-sharing systems seemi disposed to prefer theproblems he described then-the "several hurdles that elucidation of the software problems-i.e., thoese problemsstand between the nonsymbiotic present and the antici- are more interesting and crucial in the successful devel-pated symbiotic future"-are still with us, although they opment of these informational "public utilities."do not loom as lairge because in the interim much in- Generally speaking, it is now both necessary and pos-tensive work has gone into whittling them down. sible to determine just what a person does when he inter-

Licklider blocked out the problems into a number of acts with a computer. In former days, Dr. Lickliderareas: the speed-cost mismatchi between mian and points out, fantastically little attention was given tocomputer; the physical interface, that is, the console how programming was done.,, How a programmerwith its displays and controls; the language mismatch worked-his mental processes, his methods for plan-between man and computer. In each of these areas, there ning an attack on a problem-were hidden in obscurity.are profound problems that excite the interest and con- However, in the on-line situation, it is possible to findcern of human factors investigators. Furthermore, the out exactly how he proceeds, step by step. What heproblems are of a radically new order, for they involve does becomes overt. Thus, the habits, procedures, andthe analysis of man's mental processes to a far greatter needs of many on-line users are now the object of inten-degree than has previously been necessary. sive investigation. At M.I.T., for instance, there is in

Time-sharing systems. The fact that "computers go very progress a special Technical Information Project that

LindIgrcn---Hurnan faictors in eiginieering-Il 63

centers around the MAC time-shared computer, and oriented languages to problem-oriented languages morewhich among other objectives is investigating the role akin to natural language and human users. The psycho-of human factors in literature search.9 The MAC facil- logical principles guiding this evolutionary shift areity consists of a central computer with about 100 re- traced out by Dr. George A. Miller in one of his (as ismote consoles that are available to about 400 people usual with him) marvelously lucid papers.4 He explicates(only about 30 can use MAC at the same time). The in- the "psychology" behind the introduction of coding,teraction between user and the system proceeds by means relative addressing, and the hierarchical structuring ofof a language very close to natural English. The user programs, in which the development of subroutines (by"operations are observed and monitored by means of means of which programmers could borrow wholetest procedures and feedback. The information thus ob- chunks of programs from one another) led to the con-tained will guide the future evolution of the system by cept of the compiler. With compilers it has been possiblesuggesting component and procedural innovations."' to construct languages that are much closer to the way aThat on-line users of time-shared computer systems are man thinks (FORTRAN, COBOL, ALGOL, IPL, COMIT, etc.).

entering a "new world" quickly becomes evident from the Essentially, the programmer selects from a library ofamusing stories of what is happening to these users. subroutines the ones he needs, and the compiler translatesAs Clark Weissman points out in a discussion of the it for him into the microsteps of machine language.experiences of the System Development Corporation Thus, the man is able to plan his program in larger(SDC) Time-Sharing System (TSS),10 there can be a hierarchical chunks that are more natural to him, and thegreat deal of helplessness and frustration in on-line pro- machine does the dogwork translation into binary mi-gramming efforts when there are communication failures crosteps more natural to it.owing to the user's ignorance of proper procedures, and The development of compilers has had a revolutionaryhe loses his programs. Some of the stories have positively impact for it brought the interface between machine andChaplinesque overtones: users drop in on each other's man a giant step closer to the man, and in the same stepprograms, get lost, wipe out another user's programs by brought whole new classes of users12 into interaction withaccident, peer over each other's shoulder, and what not. computers-workers in a wide variety of disciplines forIt all sounds fascinating and frustrating-a world in whom the "linguistic" barrier had previously been morewhich one is likely to run into wandering ghosts and or less insurmountable."lost souls." Although much of this evolution from machine-

In one of Weissman9s anecdotes, the computer oper- oriented to problem- and man-oriented languages pro-ators received this cryptic message from a remote user: ceeded empirically, without it occurring to the program-"I'm alone; please decipher the above error message." mers to call in the psychologists and the human factorsAnd, of course, the error message was only on the user's people, the problems they were coping with were es-terminal. At another time, during an important dem- sentially psychological or human factors problems.onstration of one system setup, the demonstration was Nowadays, with the heightened sophistication of man-punctuated by a message announcing the score of the computer interaction, the workers from these disciplinessecond game of the World Series. The system users are are very much in the act. They feel that the trend towardvulnerable to messages from "jokers" or from users who more natural languages of communication betweenare "careless," ignorant, or confused. "When computers man and machine will continue, although they are not sowere 'downstairs' and proper experts were 'around the sure just how this will come about. Miller says, "We arecorner,'" writes Weissman, such failures "might not today at a point where the subtler aspects of humanhave been as disconcerting to the off-line user. But, to language and human psychology are beginning to intrude,the remote time-sharing user, possibly thousands of and we are not clear enough in our understanding ofmiles from the computer or the expert, communication human languages and human language users to bein its broadest sense is of paramount importance."10 confident of what step to take next in reducing the gap"4Of course, the message in such tales for those who look between man and machine. He cites some of the features

on computers with less than unbounded enthusiasm is of natural language that might be introduced into com-to wait until the avant-garde investigators have per- pilers yet to be developed-e.g., recursion and greaterfected these potentially marvelous information assistants use of context-but in this direction, the investigatorsto the point where we can plug into them from a console are brought to the brink of profound problems involvingby our office desks and converse with them in our native them in, among other things, the new theoretical studiestongue. That day, says Licklider, though it may sound in linguistics."Ia bit fantastic now, is coming. Another direction being taken by some investigators isThe language mismatch problem. In his 1960 paper, the study of existing programming languages to dis-

Licklider cited the "basic dissimilarity between human cover what is common in them, with the aim of develop-languages and computer languages" as being probably ing a homogeneous language that will understand all thethe biggest impediment to man-computer symbiosis. languages in the computer libraries, but this work is in aToday, despite the advances, the language mismatch re- primitive state. Dr. Licklider speculates that there "ismains a serious obstacle. The realization of computer not going to be a single language"8 in the man-computersystems that will understand natural speech, which stands interaction process.as a kind of ultimate goal, seems far off." Still another tack is taken by Allen Newell. He stressesThe dissimilarity between man and computer languages the need for investigating the ''planning process"~in

has been the source of new psychological problems, man'l -how man goes about getting his ideas for solvingproblems that were "scarcely dreamed of twenty years problems-so that the computers can be made moreago." What has happened is that there has been a gradual "intelligent" in the interaction process. Newell pointsshift in emphasis from computer-oriented and procedure- out that the problem is one of psychology. "First,"

64 IEEE spectrum APRIL 1966

he says, "we must understand the structure of humanplans: What is it that the human has when he has the'idea of solution'? Only when considerable is knownabout this can we understand what the computer mustdo to take a plan as input and carry it out. We need todevise a language of plans sufficiently similar to the COR'natural language' of human planning so that the man is IN

able to utter his plan to the computer. If this language is ".' - jtoo difficult, then the human will face a translation taskto put his plans into the language. Thus we need to know _ ___in considerable detail what information is available in a .human plan and how these plans are reduced to action."'3As a consequence, he says, "the work on artificial intel- -Aidligence becomes directly related to the work on man-computer communication." And, he says, "if there isone final conclusion to be drawn ... it is the importanceof psychological investigation to the problem of man- _computer communication. The task is not one forcomputer technology alone."'3

Rather than having come to the end of a line, then, itseems that those who are in search of a more intimateman-computer symbiosis have broached some "beauti-ful" new problems. In general terms, one can concludethat the next order of business in man-computer inter-action problems is to bring better organization into thelanguages, theories, methods, and devices of program-ming. Significantly, Dr. Ivan Sutherland, now directorfor information processes technology at the AdvancedResearch Projects Agency (and successor to Dr. Lickliderin the post), says that his office is stressing the softwarearea in its many contracts aimed at finding ways in whichmen can make better use of computers.3The physical interface. In discussing the language

mismatch problem, we have failed thus far to point out amost significant advance that may be viewed as a comple- Fig. 1. Computer-controlled displays. Top figure ismentary hardware and language development-namely, multiple exposure. Bottom figure is somewhatman-machine graphical communication devices. About similar to that shown in different projections on

1960, an interest began to build up in developing com- cover.puter display systems whereby man and computer could"converse rapidly through the medium of line drawings."14The first really striking advance in this direction was Fig. 2. The ESL display console, being developed at"Sketchpad" (at the M.I.T. Lincoln Laboratory), the M.I.T., is a specialized computer that automaticallywork of Ivan Sutherland.'4 With Sketchpad, it was pos- converts three-dimensional drawing commands intosible to make two-dimensional sketches with a light pen arbitrary two-dimensional projections as in Fig. 1.

The three-dimensional rate-control joy stick (crystaldirectly on a computer CRT display, to modify and move ball, near CRT) allows the user to rotate (in real time)parts of the drawing as one wished, while preserving the the figure appearing on the scope face in almost anytopology of the drawing, and more important, to carry fashion. Light-pen tracking is fully automatic.out computations on the figures so drawn. These ca-pabilities showed that such graphical language devices '~could serve as tools of extraordinary power in design andother applications. For instance, Sketchpad could beused to draw basic bridge shapes, which were thensubjected to various loading conditions, so that thestresses and strains in the constituent members could becomputed automatically. The Sketchpad system was

subsequently extended by T. E. Johnson to permit theuser to work with three-dimensional figures. Thesegraphical communication systems have clearly just openeda Pandora's box, and now many organizations are work-ing on such systems. Another very well known one, forinstance, is the RAND tablet."- Dr. Licklider reportsthat IBM is now in the process of developing five differentadvanced display systems,8 although details about themare still in the realm of trade secrets.A most extraordinary and sophisticated system is the

Lindgrcn-Human fnctors in enginecring-I1 65

M.I.T. Electronic Systems Laboratory display console, cal networks.'8"19 In this program, being carried forwardwhich since January 1964 has been in operation in the by Prof. M. L. Dertouzos and his colleagues at M.I.T.,Project MAC Time-Sharing System.'6 With this spe- an electronic circuit designer interacts directly with thecialized computer console, it is possible to make directly computer through a typewriter and CRT graphicalthree-dimensional drawings and convert them into input-output equipment. The designer communicatesarbitrary two-dimensional projections, to rotate them in his circuit configuration an element at a time to thereal time, to translate them, to change their scale, and so computer by placing the light pen on the CRT, as shownon (see Fig. 1). Figure 2 shows the console. The three- in Fig. 3, and by pressing the appropriate elementaxis globe in front of the scope is used to rotate and button. Through this process, he can compose on thetwist the drawing about as desired, so that one can get screen any circuit-including resistors, inductors, ca-a feeling of the figure from all sides (the SPECTRUM cover pacitors, diodes, and tunnel diodes-without knowledgeshows different projections of a complex form). John of computer programming; he can then ask the computerWard, assistant director of ESL, points out that this to analyze the circuit by a typed command. Prof. Der-console, which was designed with special attention to the touzos points out that a crucial point in this approach isneeds of computer-aided design under the restriction of a that it involves simulation rather than analysis. The circuittime-sharing system, is so flexible that it can be used for simulation is achieved by representing each electricalmany applications. For instance, it is being used in the element by an appropriate block of storage registers anddesign and evaluation of ship hull forms.'7 In this setup, by embedding the laws of circuit theory within the com-three-dimensional hull surfaces displayed on the CRT puter. Then, for any given excitation, the computercan be altered in a few seconds by typed-in changes in "walks the circuit through" its behavior much as a realparameters, and can be rotated to any desired viewing circuit operates. A most significant consequence of thisangle for closer study. man-machine interaction process is the short turnaround

Computer-aided circuit design. Another sophisticated time, of the order of seconds, between a modification ofgraphical-type research program that uses the Project the circuit by the designer and the computer response,MAC computer involves the on-line simulation of electri- whereas in the past this time was of the order of days. 20

The moral of this advance should wallop design engineerson the head-i.e., the 99.99 percent of engineers who aredesigning circuits without on-line graphical-languagefacilities are, in one sense, already "living in the past."As Prof. Dertouzos notes, "this reduction of delay

Fig. 3. Prof. M. L. Dertouzos and C. W. Therrien in the design loop has obvious economical and technicalshow how with a light pen and their new computer repercussions."20 He is careful to point out, however,program tied in to Project MAC, they are able to in- that breadboarding and testing will not be eliminatedvent, analyze, and display immediately electrical net- . .works. Computer-aided circuit design should have by this design process, but all the same he is temptedprofound repercussions in the electronics art. into speculating on the future of man-machine inter-

action along these lines: "It is not difficult to imagine anot so distant time when a designer seated before aconsole attains, after some man--machine dialog, acircuit design he considers satisfactory. He then asksthe computer to search component-manufacturers'tapes for components that will meet circuit and designerrequirements. Having done so, he then commandsnumerically controlled tools to fabricate a prototypeof the circuit, while all related financial transactions are

completed automatically."20And the exciting part about Dertouzos' projection is

that it represents but one meaning of that Anglo-Saxon-Latin-Greek mouthful, "man-computer symbiosis."For he goes on: "What is harder to imagine and per-

1 haps more challenging to speculate about is the contri-bution of the man-machine dialog to the creative orconceptual stages of circuit development. Certain mech-anisms of creative design are observable and to someextent comprehensible. For example, a designer con-fronted with the design of a circuit to meet certain speci-fications resorts to his own memory for 'past experience'in designs of 'this sort.' He may then postulate a first-order circuit configuration that should accomplish thedesired task. In evaluating his 'experience' or ideas sug-gested to him by the problem, the designer often usessome measure of goodness to reject or accept the pro-posed configuration. In decomposing a large system hemay start from the output specifications and advancetoward the input, or vice versa. We see that from aqualitative point of view, the notion of 'experience' is

66 IEEE spectrum APRIL 1966

mechanizable by an organized memory. Evaluations of functions to be assigned to the man and to the machine.conceived configurations or 'good moves' are mech- In the past, however, this goal has been easier to postulateanizable as heuristic weightings in the decision process; than to realize. Much less has been known about thewhile the conception of heretofore unknown configura- engineering characteristics of human operators thantions may be accomplished by a somewhat constrained about the engineering characteristics of automaticrandom search."20 guidance and control systems. As Prof. T. B. SheridanThe possibilities and experiments with on-line facilities, of M.I.T. has so picturesquely put it: "It must be ap-

with such a degree of man-machine symbiosis, have preciated that in dealing with the human operator wealready become so numerous that no adequate account are dealing with the 'blackest' of 'black boxes' and thecan be given. A glimpse of the breathtaking spectrum of most complicated of physical systems."25 Because "mansuch studies, embracing wide classes of users in both the is a nonlinear, discrete, adaptive, time-varying, statisticalphysical and social sciences, can be obtained from a controller with a 'mind of his own,' "26 it has been difficultProject MAC Progress Report.2' To scan but a few of to develop adequate models for human operators, sothe others: at Bolt Beranek and Newman, Dr. Jerry that one of the major avenues to man-vehicle designsElkind is heading up work on teaching through man- has been the simulation of vehicle control for varyingcomputer interaction22 and is carrying out on-line control degrees of human operator participation. There canof human factors experiments;22 at the MITRE Cor- hardly be an engineer alive who hasn't seen at least oneporation, Dr. Edward Bennett and his colleagues have form of such simulators. Thus, in the evolution of man-been developing an experimental on-line information ually controlled vehicles, designers were very much de-control system called AESOP, which serves as a prototype pendent on subjective opinion ratings expressed by theof a class of management or command information human operators as to the vehicle's handling qualities.systems capable of giving the members of the using The aura of the test pilot did not grow throughorganization (all the way up to the highest executives) public relations gimmickry-an engineer could not knowimmense on-line control over system performance;93 the what the man-vehicle situation would be really likequestion of decision making in man-machine systems,24 until the pilot took off, flew the vehicle, and came backtaken up in a current special issue of the IEEE TRANS- to tell him. Thus, there have been assembled extensiveACTIONS ON HUMAN FACTORS IN ELECTRONICS, is re- catalogs of vehicle dynamic parameters given as func-viewed in this month's "Scanning the issues," page 120. tions of pilot ratings,27 which no doubt will continue

In discussing this exploding field, released by the genie to be used. However, as Duane McRuer and Dunstanof man-computer interaction, one must bring oneself to Graham (both of Systems Technology, Inc.), and Ezraan artificial stop. So many concepts and possibilities are Krendel and William Reisener, Jr. (of the Franklinnow emerging that one must ask: "Just how far can this Institute) point out in a recent report, "In a fundamentalthing go?" Dr. Sutherland gives us one view. He has sense, these catalogs are only reports of specific testspeculated that the "ultimate displays" may allow a results-they fail to explain adequately the mutualuser to draw abstract systems that operate under any interactions between the pilot and the vehicle, and theylaws he wishes to impose, and then, through special are difficult, if not impossible, to extrapolate to newtelevision headsets, he might "feel" his way into these situations and novel vehicle characteristics."27 Accord-abstract environments. Such fantastic systems might ingly, of late there have been increasingly extensivebring about invaluable discoveries by investigators who efforts to generate better theoretical models of humanfail now to understand certain physical phenomena operator dynamic characteristics, especially in view ofsimply because they cannot "see" the underlying dy- the increasing space work in which man-vehicle designnamics of the phenomena. Although such ultimate sys- failures would be more expensive and devastating thantems (one can hardly describe them as mere man- anything that has gone before.machine "interfaces") are still far in the future, creative The general situation is summed up by Prof. L. R.people are now thinking about them seriously for the Young of M.I.T. as follows. He points out that psycho-simple reason that, owing to the developments of the logical testing has been very fruitful in that it has led topast few years, they are no longer impossible. a certain level in devising models for the human oper-

ator. Now, however, "for some of the adaptive andMan-vehicle control and guidance systems optimization control problems, which the human operatorUnlike the field of man-computer relationships, the involves, we have little appropriate mathematics available

study of the human factors involved in the manual con- to treat them. I think the time has come for the humantrol, guidance, and stabilization of vehicles has been factors people to look more intensively into mathematicalgoing on for a long time, and for all kinds of vehicles- possibilities. We must build up some theoretical backlogmotorbikes, automobiles, aircraft, and now spacecraft. strictly for the manual control problems rather than con-Although formal study of the "controlling function" tinue to live off what has been ready-made for us inof the human operator began about 1900, well before automatic control."2'the development of formal theories of servomechanisms, 25 Most analytical studies of human operators haveand although every aerospace organization (as well as viewed the man as a simple controller in a simple closed-iust about everybody else) can boast whole bibliographies loop operation-the man gets information on the vehicleof manual control studies, we can include here only a motion through external sensors, such as gyroscopes,few points about recent investigations in this rich and radar, etc., and through his observation of the error be-"venerable" domain of human factors. tween his actual and reference orientation he stabilizes

Traditionally, for each vehicle design, the designer his craft by means of a control stick or whatever. Forhas had to decide, on as rational a basis as possible, this type of closed-loop system, McRtuer and Krendelthe degree of the guidance, control, and stabilization developed in the late fifties a set of quasilinear transfer

Lindgren-Human factors in engineering-II 67

functions that quite accurately predicted the control The bicycle project is but one of a broader program ofcharacteristics of the human operator over a wide range studies of the adaptive functions of man in vehicle con-of inputs and vehicle control dynamics.2't This pilot dy- trol systems. In many respects, adaptive control systemsnamics model, in fact, found considerable application, attempt to imitate the adaptability of humans. (M.I.T.'sand just recently these same investigators completed a Prof. T. B. Sheridan is credited with the first study of thecomprehensive five-year programn of developing better adaptive characteristics of the human controller-inpilot models for use in handling qualities and manual 1960.) But, Professor Li and his colleagues point up thecontrol system analyses.2- These new models, which are irony that it was not the versatility of human adaptivemuch more precise than the circa 1960 model, are ex- control but rather its limitations that led to the greatpected to have far-reaching applications in the future. surge in the development of automatic adaptive controlThey embody, among other things, new knowledge that systems in the last few years. In general, the "human out-"should have significant impact both on the content and shines the automatic system with his huge capacity ofnature of the information displayed to the pilot and on open loop or programmed control; but he lacks thethe design of the manipulative devices with which the capacity and speed for making on-line computation whichpilot imparts his desires to the vehicle." ' is needed in the operation of active continuous adaptiveOther very extensive investigations of control models systems.""2 Thus, "the objective of adaptive control

of the human operator have been going forward at the systems is to reduce the range of adaptation required ofM.I.T. Man-Vehicle Control Laboratory, which is part the human operator."32of the Center for Space Research. They have been work- For space workers only. One of the most novel human-ing on, among other matters, the extension of models of operator-vehicle systems now under investigation is thathumans in orientation control tasks to include the de- of a completely foot-controlled platform that free-floatingscription of the role of the nonvisual motion sensors space workers might eventually use.33 The study into this(particularly the tactile and vestibular sensors) .2 3o use of the human balancing reflex for stabilization andFigure 4, which represents a general block diagram of the control of vehicles in one-g and sub-g environments isman -vehicle control problem, shows the major areas going on in the Life Sciences Section of the Grummannow under research at M.I.T. It should be remarked in Aircraft Research Department under the sponsorshippassing that a very good, readable, general discussion of of NASA. The research embraces triple aims: the use ofthe many questions involved in the assignment of roles the foot-balancing concept in one-g vehicles; its applica-for men and machines in space systems appears in a recent tion to zero-g conditions; and basic studies of the humanpaper by C. S. Draper, H. P. Whitaker, and L. R. Young, balancing mechanism. The appeal of a jet-propelledall of the M.I.T. Department of Aeronautics and Astro- foot-controlled space scooter, says Thomas Keller,nautics. 31 head of the Life Sciences Section, is that it makes use of aAmong the more earthbound, but sophisticated, neuromuscular response that has already become a re-

studies in progress at M.I.T. is that of a stabilized motor- flex. "We feel," he says, "that the foot-control vehiclebike, which is being used as a test vehicle for dynamic would be far superior to the Buck Rogers type space-orientation. In these studies, which have been going on pack."34 Theoretically, the big advantage is that the spaceunder Prof. Yao Tzu Li's direction for the past few years, worker would require no special control training, andthere are two aims: one is to find out more about how ahuman rides a bicycle (that is, to identify the rider'scontrol function); and the second is to build an autopilot,an adaptive automatic stabilization device that will du-plicaite the human rider's performance. 21 Fig. 4. New investigations going on at M.l.T.'s Man-

Vehicle Control Laboratory aimed at extending con-trol models of the human operator in dynamicspace orientation are indicated here by circled items.

aQ~~~~~~~~~~~~~~~Vehicle osition and orientan a v:naIeu system

68 IrEE spectruLIm AP'RII 1966

his control reflex would be better under conditions of ening" term. Any modification to a closed-loop man-stress. In addition, such a scooter would leave the space machine control system that reduces the need for theworker's hands free. Figure 5 shows a zero-g simulator human operator to perform analog differentiations isthat is being used in the research project. defined as quickening. The concept has been applied inNew aircraft landing aid. Just to relieve any impression many human factors designs.)

that the reader may be getting that all modern human fac-tors problems are concerned with complex systems, we Human factors in communication systemsshall cite an at least deceptively simple development that We have seen something of the human factors andhas been under way at the U.S. Naval Research Labora- psychological problems involved in communicationstory.35 It is an optical-geometrical system, called the between man and machine. The communications betweenRainbow Optical Landing Aid, which its designers (H. P. man and man through machines as intermediaries bringsBirmingham, A. W. Baldwin, and Miss Barbour Lee different types of human factors into the foreground ofPerry of NRL) hope will increase the safety of landing attention. In this respect, the problems ofvoice communi-high-performance aircraft aboard aircraft carriers in cation through telephone systems are especially interest-nighttime operations, a hazardous feat at best. ing.The new landing aid is designed to provide a pilot with Some of the investigations carried out at Bell Labora-

information about the appropriateness of his descent tories, for instance, where a Human Factors Research De-during his landing approach, information which present partment was established in 1948, include: the measure-landing aids do not adequately provide (i.e., accident ment of human channel transmission characteristics;rates have been very high). Altitude-rate or sink-rate error studies of telephone users' habits and desires; prob-information is transmitted to the incoming pilot by a lems of echoes and time delay on telephone conversa-dynamic tricolor light beam (see Fig. 6), which by its tions; users' responses to push-button telephone keysets;color sequence tells him not only his angular error from verbal behavior in retrieving telephone information;glide path, but which is also coded to tell him contin- how customers use number services; and many others.uously whether or not his angular rate of descent is too More recently, with the experimental study of trans-small, too large, or just right. This latter quality of the mission through satellites, and with the prospect of com-landing-control system-the inclusion of "lead" or error- puter services in the telephone system,36 new human fac-rate information-is important, for it reduces the diffi- tors problems have come to the fore.culty of the control task required of the pilot. (In NRLterminology, this represents the inclusion of one "quick-

Fig. 5. (Right) Zero-g simulator at Grumman Air-craft Engineering Corporation's Research Depart-ment. The man is turned on his side so as to elimi-nate the gravity vector. This research is aimed at thedevelopment of a small space platform that a "spaceworker" mightcontrol by means of his foot-balancingreflex, thus obviating special control training.

Fig.6. (Below) RainbowOpticalLandingAid,beingdeveloped at U.S. Naval Research Laboratory,simplifies pilot's tasks in nighttime carrier landings.The dynamic array of color-coded light beamsguides the pilot from any path of approach into thecorrect steady-green glide path. The Rainbow unit(U) would be space stabilized to compensate forship pitch, roll, and heave movements.

Red-blue-whitesequence

Red-white-blue

B > Steady red Red-white-blue

W = WhiteR ~~~~~~~~~~~~~~~~~~~~~~~~~~R = Red

W ~~~~~~~~~~~~~~~~~~~~~~~~~~~B = BlueG = Green= Flashing red

G'= Flashing green

Horizontal

Lindgren-Human factors in engineering-11 69

One of the problems involves the use of computers to realization of a human factors dream.store telephone numbers. In the proposed system, The other big human factors study at Bell involves thewhen an information operator receives a customer re- Early Bird satellite, and looks into the question of howquest, she keys the request into a console by pushing but- well people can talk over a satellite connection. Over atons, and a set of numbers appears on a display before period of some months now, normal, commercial tele-her from which she selects the number fitting the request. phone traffic has been relayed over Early Bird as wellHowever, since computers are so expensive, one prob- as over cable. After each call, a subscriber is queriedlem being investigated by the Bell human factors group about his experience with the connection (he does notis whether or not this method is enough faster to com- know which system handled his call). Results of thesepete with telephone books on cost. The design of the experiments will be announced within the next two toconsole input and output-the various keys and display three months.37characteristics-constitutes one area of the problem, butit is, according to John Karlin, head of the Bell group, Advanced space systemsthe least important. The greater share of work is that More than any other man-machine system, the ad-of looking into the best search strategies for the computer. vanced manned spacecraft missions due to come offA major problem arises because the computer cannot before the end of this decade represent the most crucialthink as a human operator can. If given information is problems yet encountered. They have conscripted humanincomplete or inaccurate in any way, the computer will factors efforts across the entire spectrum-man-vehiclebelieve what the operator tells it. Accordingly, the really control dynamics, advanced displays, computer backup,tough problem is to devise an appropriate search strategy sophisticated real-time mission simulations, communica-and to determine how complete and accurate the informa- tions, life support systems, atypical environmental effects,tion the operator gives the computer must be. The big allocation of tasks to several crew members, ground sup-cost savings would come from determining how much port equipment. There is almost no way of gettingof a customer's request the operator should key into the a human measure of the effort involved. They say 30 bil-computer. Karlin reports that if all goes well in the study, lion dollars worth, but what does that mean in humanthey hope to have the final results this summer.37 terms? Grumman Aircraft, which a few years ago hadThe Bell project also presents an interesting sidelight. no human factors people around to speak of, now has

Human factors engineers always stress the importance a permanent staff of 50 people, 30 of them working onof being in on design studies as early as possible so that problems associated with LEM (Lunar Excursion Mod-their investigations can properly influence the final sys- ule) alone. And LEM itself, as big and fraught with prob-tem design. In the Bell case, the engineering department is lems as it is, sits like a pea atop the colossal Saturn ve-not even going to begin to implement the computer hicle that is to start it on its mission. But size alone doeslookup system until they have seen the complete results not convey very much meaning. It merely symbolizesof the human factors study.... which amounts to the the level of human busyness. Saturn and its associated

problems are to our day and our culture what, perhaps,the Great Pyramid was to its day and the people of theFourth Dynasty. Yet, such analogies don't provide muchground to stand on either. Finally, in discussing engi-

Fig. 7. On this kind of display surface (developed neering efforts of such magnitude, what one must do is se-by means of computer programs), errors or abnor- lect (somewhat arbitrarily) a single aspect and let itmalities show up instantly as "bulges." Informa- stand for the rest.tion displayed here is equivalent to that contained Human factors in prelaunch activities. Now that the dayin 42 000 eight-digit numbers. If the same informa- of the "really big" space programs has arrived, many oftion were in "page" format, it would take a man the adhoc space prelams has develofhours of analysis to discover the errors. the ad hoc procedures for prelaunch checkout developed

during the earlier "smaller" programs, such as Mercuryand Gemini, must now become much more formalized,while the highly automated ballistic missile checkout pro-

\.0000 -

cedures (e.g., Minuteman) are inappropriate to man-5.0000- rated vehicles. In the earlier manned space programs, the

use of ground computers was practically nil, but for theApollo program, for instance, in which the launchingalone will involve something like 10 000 men, and forwhich thousands of data points (upwards of 6000) will

3.0000- / -be monitored at central stations, the assistance of com-lineputers ha, become mandatorY; and the problems of on-

i E EED D D ~~~~~~~~~~~~~Althoughcomputers are being incorporated into missileEWBDEDGDm/XM7 ~~~~~~andspace programs, questions arise as to what level of°°°° 1~~~~~~~~~~~~~~automation will be most effective in prelaunch checkout.

From a human factors point of view, this presents a newkind of problem for which there exist no adequate guide-lines for the appropriate allocation of tasks between the

I ~~~~~~~~~~~~~~~menand the computers, or for the selection of effective-1.000 - control and display techniques. In addition, the launch

-1.000 1.0000 3 0000 5.0000 7.0000 checkout crew finds itself faced with less "direct contact"

70 IEEE spcCtrumn APRIL 1966

with the space vehicle, due to the interposition of check- Sciences Division of the Illinois Institute of Technologyout computers used primarily for sequencing and Research Institute. This plot displays information con-go/no-go determinations. In the face of an obvious need tained in approximately 42 000 eight-digit numbers.for, and recently demonstrated success with, automated Through use of a computer program, a display surface isspace vehicle testing, checkout personnel nevertheless formed that instantly reveals errors or abnormalities bymust retain ultimate responsibility for launch decisions. bulges on that surface.This responsibility, which weighs most heavily in themanned space program, can be carried out most effec- Visual sampling behavior-a paradigmtively when the vagaries and special characteristics of In these two articles, we have referred fairly often toindividual vehicles are fully understood. Thus, for the stress that human factors investigators now place onexample, the period between delivery of a space vehicle to the need for greater scientific generality in their modelsCape Kennedy and the launch is a period not only of of aspects of human behavior. With better theoreticalintensive activity but of continual learning. In a recent models, they say, it should become possible ultimately tointerview, Joseph G. Wohl, of Dunlap and Associates, solve analytically many human factors design problemsInc., a firm that has made many studies of prelaunch that have had to be treated empirically through largelyman-machine problems, stated the problem in a form ad hoc methods. They want, in other words, to make avery close to that of a human factors principle: "The full science out of this business.significance of missile and space vehicle preparation and Examples of this trend have been cited under the dis-prelaunch checkout is understood only when one realizes cussion of man-computer24 and man-vehicle systems.that these learning activities are indeed the primary means An example of the weighty power of newer theoreticalby which the residual unknowns, having escaped the methods in modeling human pilots has also been cited infilters of engineering analysis and factory testing, may be a recent SPECTRUM article.6 Still another recent investiga-discovered, understood, and brought under control."38 tion, which can have an important bearing on the analysisCheckout computers and display systems can aid this of manned systems, is that undertaken by John W.learning process. Thus, man-computer interaction in Senders of Bolt Beranek and Newman, Inc.prelaunch activities represents in fact a general problem One of the questions with which Senders has beenconfronting all future advanced missile and space systems. concerned is how people absorb information; i.e., human

Displays. One of the important concerns of prelaunch information processing in relatively complex tasks.activities is the detection of nonnormal (but within tol- Senders says that he "spent a number of years dealing witherance) conditions, a process that is very much dependent more or less classical kinds of experiments in which youon the manner in which the test information is displayed. hold everything constant and vary one thing at a time andSays Wohl, "It is clear that detection of cyclic distur- so on. These days people have discovered that you can bebances and second-order trends in sampled data can be led into traps when you try to extrapolate the results ofenhanced by some display techniques and suppressed by single-valued functions to multi-valued functions, forothers."38 In their studies of existing display capabilities, the interaction terms are very great indeed."39 Thus,Wohl and his colleagues point out potential inadequacies he has been led in his more recent efforts into a study ofand suggest improvements based on human pattern- the nature of the effects of complication, and he has beenrecognition capabilities. Present and planned test moni- trying to "generate mathematical models, which aretoring facilities differ from the Gemini and Mercury rational models as opposed to purely empirical ones, in anfacilities (which used indicator lights, meters, and strip- effort to characterize and predict what will happen whenchart recorders) in that they include CRT displays packed [human operators and monitors] are confronted withwith information in alphanumeric form. Such so-called very complicated information flows, complicated dis-page- or list-formatted CRT's, although they allow the plays and complicated controls."39'40presentation of a great amount of information, submerge These studies lead to a rather comprehensive theorythe test engineer's capabilities of observing data trends, about human visual sampling behavior in particular andoscillations, and other abnormal indications. In such about the nature of attention in general. The extendeddisplays, a value that goes out of limits is indicated only theory incorporates ideas about what Senders callsby a periodic interruption of the appropriate line of in- "iconditional sampling behavior, in which an observerformation. Nor does such a display give the human inter-sample interval is a function of the value of themonitor information about instantaneous rate of change signal read on the previous sample." What this means isor leave behind a time history. Even at the relatively slow that an observer breaks up the environment in terms ofpace of manual checkout, this display makes it difficult to attentional demand. In effect, he says, "This demandsobserve trends in test data. For fully automated tests run more attention than that." For instance, if he is observingat "system speed," these displays are nearly useless in this dials and displays, he is most interested and attentiverespect, and would certainly not enhance the test engi- when their values indicate that a variable is approaching aneer's understanding of what he was observing. Wohl boundary condition. The closer a dial needle, say, getsstates: "To help test engineers detect anomalies while to a boundary, the m,ore often the observer should look atattempting to monitor a large mass ofCRT data at system it. This behavior is what Senders calls conditional sam-speeds, we recommend the development of a real-time pling, and its mathematical formulation might be identi-pattern recognition format for CRT's in which test point fled in textbooks one day as "Senders Sampling," a namereadings are 'mapped' onto a two-dimensional display in that has more than mere alliteration to recommend it.a coding scheme designed to enhance detection of im- Senders advances the idea "that as a consequence of thepulses or other fleeting noncyclic disturbances."38 One single-channelness of attention, queueing theory pro-example of the type of thing Wohl is talking about is vides a general method of analysis of the switching ofshown in F;ig. 7, a display developed at the Computer attention, of the attentional demand of a stimulus source,

Lindgren-Human factors in engineering-II 71

of the probability of simultaneous demand from two or 8. Licklider, J. C. R., Private communication.more sources of stimuli, and of the notion of overload. 9. Kessler, M. M., "The MIT technical information project,"The conditional sampling models provide the probability Phys. Today, Mar. 1965.

10. Weissman, C., "Communication techniques for an interna-distributions which enter into the queueing model."40 tional time-sharing users group," presented at IEEE 6th AnnualThrough the mathematical formulation of sampling Symp. on Human Factors in Electronics, May 6-8,1965.

behavior, Senders points out that it should be possible to 11. Lindgren, N., "Machine recognition of human language-predict how often an observer will need to look at the Part I," IEEE Spectrum, vol. 2, pp. 114-136, Mar. 1965; Part II,predlc how ften n obsrver lll ned tolook t the pp. 44-59, Apr. 1965.various control instruments and information displays in a 12. Rowe, W. D., "Why systems science and cybernetics ?," IEEEparticular man-machine situation. Thus, in a complex Trans. on Systems Science and Cybernetics, vol. SSC-1, Nov.display situation, it would be possible to calculate the 1965.

probability thatoninstrumenthasbeenleft"scream 13. Newell, A., "The possibility of planning languages in man-probability that one instrument has been left "screaming computer communication," in Communication Processes, F. A.for attention" while another is being attended to. Fur- Geldard, ed. New York: Pergamon, 1965.thermore, it may be possible to get an analytical measure 14. Sutherland, I. E., "Sketchpad: A man-machine graphical

of he elibilt o th mn i a ompex ystm,to "see communication system," Tech. Rept. No. 296, M.I.T. Lincolnof the reliability of the man in a complex system, Lab., reissued May 19, 1965.what the probability is that he will miss an event of in- 15. Davis, M. R., and Ellis, T. O., "The RAND tablet: A man-terest, that he will clobber into something,"39 a result machine graphical communication device," Memo. RM-4122-that, as Senders points out, "strips the problem out of ARPA, The RAND Corp., Aug. 1964.psychology."39 Thus, one can measure an operator's 16. Stotz, R. H., and Ward. J. E., "Operating manual for the ESLdisplay console," M.I.T. ESL Internal Memorandum 9442-M-129,actual performance, calculate the theoretical optimum, Mar. 9,1965.and then characterize the man's performance in terms of a 17. Hamilton, M. L., and Weiss, A. D., "An approach to com-percentage of the theoretical optimum. Through similar puter-aided preliminary ship design," M.I.T. ESL-TM-228, Jan.

means,it als becoms posible t evalute thehuman1965.means, it also becomes possible to evaluate the human 18. Dertouzos, M. L., and Therrien, C. W., "CIRCAL: On-line

engineering design of an instrument or a pilot's cockpit. analysis of electronic networks," M.I.T. Electronic Systems Lab.The significance of these results should not be under- Rept. ESL-R-248, Oct. 1965.estimated. What they mean is that it is now possible to 19. Reintjes, J. F., and Dertouzos, M. L., "Computer-aided designof electronic circuits," presented at WINCON Conf., Los Angeles,quantify things that have always been somewhat fuzzy Calif., Feb. 2-5,1966.in the past. As Senders says, "This is why human en- 20. Dertouzos, M. L., Private communication.gineering and human factors, and so on, have always 21. Project MAC Progress to July 1964. M.I.T., 171 pp. (Availablebeen rather fuzzy, too, because we have had no optimum, from Defense Documentation Center, Document Service Center,

no '0 percentno limitingconditionagainstwhichto Cameron Station, Alexandria, Va.)no 100 percent, no limiting condition against which to 22. Elkind, J. I. Private communication.scale the behavior or to scale the quality of the design."39 23. Bennett, E., et al., "AESOP: A prototype for on-line user con-At the present time, Senders and his colleagues are trol of organizational data storage, retrieval and processing,"

mechanizing the mathematical structure of this kind of MTP-23, The MITRE Corp., Nov. 1965: also Proc. Fall Joint Com-samplng beavio (tha is,simulting he smplin and

puter Conference, 1965.sampling behavior (that is, simulating the sampling and 24. Edwards, W., "Introduction: Revision of opinions by men andscanning processes) and testing it against what people man-machine systems," IEEE Trans. on Human Factors in Elec-actually do. tronics, vol. HFE-7, p. 1, Mar. 1966.

Investigations like this illustrate the general direc- 255t Sheridan, T. B., "The human operator in control instrumenta-tion," in Progress in Control Enginzeering-I. London: Heywoodtion of the human factors field-it is leaving behind the & Co., Ltd., 1962, pp. 143-187.descriptions of ad hoc experiments that clutter the 26. Young, L. R., and Li, Y. T., "Studies of human dynamic spaceliterature, and moving toward the unifying solidity of orientation using techniques of control theory," Man-Vehicle Con-science. trol Lab. Rept., Dec. 1964.

27. McRuer, D., et al., "Human pilot dynamics in compensatoryThe writer expresses his appreciation to the following people who systems," Tech. Rept. AFFDL-TR-65-16, AF Flight Dynamics

generously gave assistance and information during the preparation Laboratory, Wright-Patterson AFB, Ohio, July 1965.of this survey: B. H. Manheimer, consultant; H. P. Birmingham, 28. Young, L. R., M.I.T., Private communication.R. Chernikoff, J. H. Hill, B. L. Perry, NRL; R. L. Deininger, 29 McRuer, D. T., and Krendel, E. S., "The human operator as aJ. E. Karlin, Bell Labs.; J. C. R. Licklider, IBM; I. Sutherland, R. 29. system D.eT, J. Frankl, E. volh 267app.rat49 a aTaylr, APA;. L.Derouzo, W.R. Frrel, T B. herian; servo system element," J. Franklin Inlst. vol. 267, pp. 1-49, May,Taylor, ARPA; M. L. Dertouzos, W. R. Ferrell, T. B. Sheridan, June 1959.J. E. Ward, L. R. Young, M.I.T.; E. Krendel, The Franklin Insti-tute; J. I. Elkind, J. W. Senders, Bolt Beranek and Newman, Inc.; 30: Meiry, J. L., "The vestibular system and human dynamic spaceR. W. Pew, Univ. of Mich.; E. M. Bennett, The MITRE Corpora- orientation," Sc.D. dissertation, T-65-1, Man-Vehicle Controltion; J. Fox, T. Keller, Grumman Aircraft Engineering Corpora- Lab., M.I.T., June 1965.tion; M. A. Tolcott, J. G. Wohl, Dunlap and Associates, Inc.; 31. Draper, C. S., et al., "The roles of men and instruments in con-B. Shackel, EMI Electronics Ltd.; K. S. Teel, Autonetics; and R. 0. trol and guidance systems for spacecraft," presented at the XVthBesco, Hughes Aircraft Company. Internat'l Astronautical Cong., Warsaw, Poland, Sept. 7-12, 1964.

REFERENCES 32. Li, Y. T., et al., "Adaptive functions of man in vehicle controlsystems," International Federation of Automatic Control (Ted-

1. Licklider, J. C. R., "Man-computer communication-introduc- dington) Symp., Sept. 14-17, 1965.tion," in Communication Processes, F. A. Geldard, ed. New York: 33. Keller, T., et al., "A study of the mechanics ofhuman balancingPergamon, 1965. for potential application to the control of vehicles," Grumman2. Licklider, J. C. R., "Man-computer symbiosis," IRE Trans. on Research Dept. Memo. RM-299, Oct. 1965.Humanl Factors in Electronics, vol. HFE-1, Mar. 1960. 34. Keller, T., Private communication.3. Sutherland, I., Private communication. 35. Perry, B. L., "The Rainbow Optical Landing Aid," NRL Rept.4. Miller, G. A., "Man-computer interaction," in Communication 6184, Febr. 1, 1965.Processes, F. A. Geldard, ed., New York: Pergamonl, 1965. 36. Romnes, H. 1., "Satisfaction and success in the highly organ-5. Fano, R. M., "The MAC system: the computer utility ap- ized society," IEEE Spectrum, vol. 2, p. 67, Dec. 1965.proach," IEEE Spectrum, vol. 2, pp. 56-64, Jan. 1965. 37. Karlin, J. E., Private communication.6. Senders, J. W., et al., "Men and computers-Sixth annual sym- 38Wh,J .Piaecmuctonposium on human factors in electronics," IEEE Spectrum, vol. 2, 38WolJ.G,Pvaecmuitonpp. 104-108, Aug. 1965. 39. Senders, J. W., Private communication.7. Licklider, J. C. R., "Problems in man-computer communica- 40. Senders, J. W., et al., "An investigation of the visual samplingtions," in Communication Processes, F. A. Geldard, ed. New York: behavior of human observers," Rept. No. 1246, Bolt Beranek andPergamon, 1965. Newman, Inc., May 10, 1965.

72) Lindgren-Human factors in engineering-II