1 March 1962

 

AUGMENTED HUMAN INTELLECT PROGRAM

 

by

 

Douglas C. Engelbart

 

1

I RESEARCH PLAN

 

A. Specific Aims*

The long-term goal of the proposed project is to increase significantly the intellectual effectiveness of human beings. The specific aims of the work proposed are to establish an experimental research facility, to develop preliminary associated methodology, and to pursue first phases of a systematic plan for developing integrated means for the human to take advantage of the best aids that our technology can provide.

 

B. Method of Procedure

Our plan is basically to provide human subjects with the best technological aids possible (which, in the initial conception is represented by a work station having good cathode-ray-tube displays, keysets, light-gun, and controls that are tied directly to a large, general-purpose computer), and to re-design the subjects' way of attacking intellectual problems so as to take advantage of the capabilities provided in these aids. It is to be an empirical approach, guided by an extensive conceptual model that has evolved from more than two years of full-time thought and study, and representing a basic and systematic attack from a carefully chosen initial position.

For this application, the stereotype of the computer as only a mathematical instrument is too limiting--essentially, a computer can manipulate any symbol in any describable way. It is not just mathematical or other formal methods that are being considered. Our aim is to give help in manipulating any of the concepts that the individual usefully symbolizes in his work, of which mathematical concepts constitute only a limited portion in most real-life instances.

Two main phases of research are evident. The first phase, which will probably take about a year, would be essentially the development of equipment, facilities, research methodology, computer programming, etc., to be applied in the second phase. The second phase, of which the second year of work may be considered as representative, involves a progression through research stages according to a systematic plan designed to provide early and continuing practical results while directly, methodically, and continually pursuing the long-term goal. We shall first describe the goals, conceptual models, and methods associated with the second phase (as exemplified by plans for the second year), to provide a background for better understanding the succeeding description of the first-phase plans.

 

* The nature of these aims is explained in "Program on Human Effectiveness," by D. C. Engelbart, a copy of which is attached to this proposal as supporting data.

 

 

2

1. Direct Research

a. Goals

One goal is to develop the lower stages of the process hierarchy of what will be our first augmented man; the other is to advance the research methodology itself. What we mean by "advancing research methodology" is explained in Part 2-C, which discusses methodology and programming in preparation for direct research. The lower stages to be developed are expected to consist of the following:

Information Structuring--Composing and modifying arrays of symbols and lines to derive conceptual portrayals (whether straight text, diagrams, list hierarchies, or whatever) are processes basic to most of the sophisticated intellectual activity in which we anticipate that a computer could be of most help.

First we shall examine the basic methods by which man and computer can cooperate to compose and modify symbol and line drawing displays (limited to frame-size displays) to portray quickly and easily the concepts that the human has in his mind.

Next, we would develop the subject's proficiency at building and modifying portrayals that are larger than can be pictured at one time on the surface of the display. This will involve building a concept-presenting and concept-relating structure within the computer memory that is viewed through a moveable window (the display surface).

The human will be able to inspect the structure through the moveable window, or to "reach in" and add to or modify the structure. An example of such a structure is a multi-page draft of a writeup, including outlines, auxiliary notes, and tentative changes or additions. The auxiliary notes could be inserted or removed freely at any point--and when the user chooses to view his clear text these notes can be omitted from the display and the text squeezed together to close the gaps (if the user chooses special markers can indicate where such notes would come). Trial rearrangements, insertions, or deletions, can be so recorded that any combination of trial draft may be viewed in uncluttered form.

Future research will go on from this to develop ever more sophisticated ways to operate on such portrayals automatically. However, we cannot predict the needs of the human for specific basic processes until we gain experience later in testing them within more sophisticated processes. We will thus stay with this task-proficiency area only long enough to develop a satisfactory research methodology, and to develop a good first approximation of an optimum system of computer processes and human methods and skills for information structuring.

 

 

3

Process Structuring--The next probable stage of research would be development of efficient means for constructing specific functional types of structures--computer programs if you will--that represent computer processes the human wishes to design to obtain certain services from the computer. These computer processes are the human's tools, and he must have a certain proficiency in designing or modifying his tools. He already gets help from the computer in composing his lists of formal statements that make up a program, but the computer can also be helpful in determining the results of each new statement' in checking for logical inconsistencies, in comparing the function designed with the function specified, and so on. We have a methodical progression of such task areas through which our research should progress.

b. Methods

First we shall describe the model from which our approach is constructed, and then we will discuss the actual procedures involved in our research.

Our Model--Consider a given task area in which our research is trying to develop superior performance. In the proposed model, the performance of a given task is considered as the execution of a process. In a very fundamental sense, the term process involves more than just human task performance' it can also refer to the performance of a piece of equipment or a tool (what we shall term generally artifacts). The execution of any process is accomplished by the execution of some set of subprocesses. For instance, the process of writing a letter of the alphabet on a piece of paper can be broken down into the different muscular processes associated with moving a pencil' as well as the process attributed to the artifact (pencil) of making a mark when moved across the paper in a certain fashion. Here the marking process of the pencil would be termed an explicit artifact process, the muscular process that moves the pencil would be considered an explicit human process, and the whole process of writing down the letter would be considered a composite process, since it is composed of both human and artifact processes.

Figure 1 shows schematically the general composition of the process associated with the performance of a given task. At the level of tasks for which our research will start' the planning and guiding of the combination and sequence of subprocesses used in executing a given process is generally an explicit human process, i.e., "done in the subject's head." Generally we shall find that the other subprocesses organized to execute the total process of performing a given task are drawn from all three kinds of subprocesses (human, artifact, and composite). In the more complex processes to be evolved, within either information structures or process structures, the "planning and guiding" subprocesses will become complex enough to require their own information and process structuring to allow the human to keep track of them.

 

 

4

Our approach is designed to develop an integrated hierarchy of cooperative mancomputer process capabilities, beginning with those we consider basic to the entire hierarchy and building upon these "a level at a time." A newly developed capability added to the top of this growing hierarchy will represent an organization of some previously developed (lower-order) capabilities, plus some new conceptual developments (concepts, methods, strategies, etc.), plus possibly some new computer capabilities (from new programs).

It takes new skills to harness new kinds of help--new skills and new ways of doing things. But the computer is sufficiently versatile that it can give cooperative service at one level to augment those human skills and methods that can be used at a next-higher level to harness a new, more powerful kind of computer help.

The computer actually can provide the equivalent of a wide assortment of tools, from the very simple to the very sophisticated. Moreover, it can provide tools to help make new tools, and tools to help keep "the shop" organized and help plan and supervise the work.

The basic plan of approach for our entire research program is the gradual construction, at successively higher levels, of a process hierarchy that will ultimately include extremely sophisticated intellectual processes, and that will be based upon a strong and integrated structure of coordinated man-computer subprocesses.

 

 

5

Procedure--At a given stage of our research, we will be trying to develop the means by which a human can be maximally effective at executing given kinds of test tasks. To do this, we must design a powerful set of subprocesses in whose execution the subject can become quite proficient through training, and we must provide the subject with effective methods' rules of thumb, and strategies for organizing and utilizing these subprocesses in attacking a given task. We must also provide the necessary new concepts (of subject matter, relationships, methodology, etc.) and the new terminology and symbology associated with the tasks and the way he will perform them. Trial subprocesses, which we term "prototype" processes, must be designed and tested before a final set is selected for inclusion in the growing hierarchy as the "best way" at that stage to attack a given type of task.

Selection of such a set of subprocesses for use in performing the given type of test tasks nominally signals the end of that stage of research. For the next stage, a new (more advanced) type of test task is chosen, and if the planning has been sound, the prototype processes which subjects will be called upon to learn and execute will be of types already used as test tasks in previous stages of research.

As we progress into ever more complex and sophisticated task areas, we shall be developing the methodology of this kind of research to keep pace. At each stage of research, it will be important that we learn how to analyze performance, locate the factors that limit effectiveness, and evaluate effectiveness in terms of the ultimate goal of augmenting human intellect. This is so important that no given stage of research can be called "complete" until we have developed our research methodology to deal with its test-task activity satisfactorily.

No process hierarchy that we develop will be static. Innovations in equipment, concept, and method can be expected to intrude at any point in the structure. New insights will be gained by using lower-level processes in higher-level situations; the relative importance of various capabilities, penalties, and limitations will shift, stimulating the conscious redesign of lower-level processes. To maintain this dynamic pattern of growth, our methods of evaluating and implementing modifications in process capability (i.e., our research methodology) must be given as much emphasis as the actual development of means to augment the human intellect.

 

c. General Activity

Researchers will provide the test subjects with prototype processes in whose execution the subjects must become proficient; the researchers will then pose specific types of test tasks for the subjects to perform, in which the prototype processes just learned will serve as subprocesses. The researchers will observe the performance of subjects executing the test tasks with different sets

 

 

6

of prototype processes as the subprocesses. From these observations, the different types of prototype processes can be evaluated and improvements recommended. The types of test tasks must, as was pointed out above, be compatible with the future progression and growth of the process hierarchy.

Research activity has tentatively been divided into the following parts:

(1) Planning--Researchers must look toward the research developments that lie ahead in the program and choose immediate test tasks that will be representative of the types of process capabilities upon which the future work will depend.

(2) Design--Researchers must invent new and different kinds of subprocesses, plans, rules of thumb, and methods of organization. New explicit artifact subprocesses must be designed into the artifacts (mostly by providing new computer programs). New composite processes must be designed from available subprocesses that have been developed in the past, along with appropriate new human "executive" processes (planning, supervising, etc.). Such new explicit human processes must observe the bounds of human capability in the new concepts, terminology, strategies, value frameworks, and psycho-motor skills they introduce.

(3) Training--Researchers must train the subjects in the explicit human subprocesses associated with the particular prototype processes designed in anticipation of given types of test tasks. The training may be in perceptual, conceptual, procedural, or psycho-motor skills, or in various combinations of these.

(4) Testing--Researchers must know how to present the subjects with meaningful exercises and tests, and how to make observations that will provide the necessary analytical data for evaluation.

(5) Analysis--Researchers must be able to take test data and analyze what amounts to an information-processing system (a problem oriented information processing system). They must be able to determine what controllable

 

 

7

factors within the system are limiting effectiveness at a given time; and they must know enough about the other four parts of the research activity to be able to derive significant recommendations.

It is expected that the development of coordinated proficiency in each of the research activity areas listed will be as important to our research as the progress that is made in developing the process hierarchy.

The actual research work involved in design, training, testing, and analysis will involve computer programming. In design, the programming will provide prototype processes for the subjects to utilize; in training, testing, and analysis the programming will provide service to the researchers themselves. For instance, (if the program is adequately supported) the training function can be done by automated-training techniques. Then the computer could actually train subjects in becoming proficient at all of the prototype processes, including the concepts' terminology, rules of thumb, and strategies needed in order to execute them and organize them in performance of test tasks. Researchers should be able to take advantage of the computer in monitoring and logging the overt activity of subjects during execution of test tasks and in helping in the analysis activity.

d. Regenerative Features

The principle of regeneration in physical systems is fairly well known, where the output is reintroduced into the system in such a way as to reinforce the type of behavior that provided the output. "Positive feedback" is another term for this type of effect. There are several features in the planned research activity which introduce this positive feedback or regeneration into our work.

For example, the test tasks of a previous stage of research were chosen to be representative of the prototype processes anticipated for the present stage of research. In this manner, the more success we have at any given stage of research, the more powerful will be the prototype processes that can be organized within the succeeding stage. Also, the level of sophistication of the service that can be obtained from the computer at any given level is limited by the sophistication of the processes used to "harness" the computer--the strategies with which computer capabilities can be utilized, and the supporting processes that can organize and match to the needs of the sophisticated computer process. Thus, the more advantage we can take of the computer at lower levels, the greater opportunity we will have for making best use of it at higher, more sophisticated levels.

So far we have been discussing regeneration in terms of feedback within the subject matter of the research--a feature found in any well-designed, continuing research activity. But this particular

 

8

research program comprises an opportunity for a different and very significant type of regeneration. Here, the researcher is engaged in making better tools for people who are engaged in work that follows the same principles as his research (solving intellectual problems). Suppose that his research could be just as significant if he focused specifically upon the kind of tools he himself uses (e.g., information structuring, process structuring). Then, whenever he turns out a tool that is significantly more effective than the one it supercedes, besides having advanced the general field of better-tool research he will have made something he himself can use. Every advance in his research can make him more effective in further research, which increases the rate and quality of his output to the world--output which has importance for much more than just his own particular uses of it.

We have designed this type of regeneration into our approach. If there is a particular type of intellectual problem solver that we would say we are trying to work directly toward augmenting fully, it would be a person who specializes in the activities of planning, design, training, testing and analyzing as associated with research toward augmenting human intellects. We have to pick a specialist to aim for initially in order to make a manageable program out of it, and we see a number of advantages to the one we chose:

(1) The regenerative feature--We shall not be doing research on the researchers--i.e., they shall never have to assess and pass judgment in a serious research sense upon themselves-but those specific processes found to work best with the test subjects can be utilized within our research processes. Where else would one find facilities and people So compatible to the early utilization of these processes? For instance, if we learn how a man- computer team can cooperate in constructing good computer programs that provide processes of the kind being designed for his hierarchy (a capability that is just what the test subject will need for more advanced processes--where he must learn to make or modify his tools as the need arises), then our design people can learn the necessary skills (assumedly they have been dealing with the concepts already) and make use of the work station to design the prototype processes. We shall thus have increased the human's effectiveness at this type of task quite significantly before our research passes to another stage, so that in later stages of research our programs can be constructed or modified with correspondingly greater effectiveness.

(2) Orientation and planning help--The planning people will have to anticipate the types of processes that should be developed next to guide progression toward useful higher capabilities. To be able to search the activities they know best, i.e., the activities of their own program, for this orientation would be very handy.

(3) Stimulation to the researchers--To know that developments from his thought and creativity are to be applied soon toward giving himself greater freedom and power in his everyday pursuit of professional goals should have a very positive effect upon a researcher's interest, energy, perceptiveness, and creativity.

(4) Reality reference--For the researcher to know that he himself may soon have to use these techniques,remember the rules, recognize the symbol patterns, or manipulate the knobs that he is developing for a subject's use, will provide a natural curbing effect on impractical fancy. The imaginative and enthusiastic fixations that researchers often get when they are working on ways to improve other peoples' real-life situations will be constantly referred to their own real-life situations.

 

In any regenerative system, there is a possibility that too much positive feedback can make the system unstable and cause it to veer to extreme positions. We are aware of this, and could hopefully moderate our activity to guard against it. We intend to avoid the kind of outcome in which the augmentation system that finally evolved would be good only for one small group of peculiar people, and would be of little practical worth to realworld people who lacked the particular crosssection of talents of the research group.

 

2. Preparation for Experimental Program

 

a. Experimental Facility

 

Initially we shall provide one working station at which a human subject can learn to execute designated special information handling processes (the "prototype processes"), or can apply these prototype processes to "test tasks." The human subject will be interacting with a digital computer that can (if the experiment calls for it) give instantaneous attention to every act of the human subject (as detected by various types of input and monitoring means). The subject will view a display (cathode-ray-tube) controlled by the computer, which can

 

9

 

portray textual data (with a wide range of alphanumeric symbols) and construct line drawings of nominal complexity.

 

It is planned that the computer will be a Burroughs B5000, connected to the working station via a tie-line to the computation center at Stanford University, or to a B5000 at Stanford Research Institute if the Institute's tentative plans are carried out. The B5000 is designed for efficient time division among a number of concurrent information-processing jobs. The computation center (whether at the University or the Institute) will be running production computation jobs under conditions where one tie-line can be brought in and given instantaneous-demand, real-time service with essentially no effect upon the production work. During a given interval, the piece-meal service that the computer provides to a real-time customer will reduce the amount of production computation by the total amount of the service, but no one computation customer will be likely to notice this--his billing will not be affected, and provision for automatic billing to the real-time customer for services rendered is already built into the B5000. Thus, the equivalent of full capacity of a very large and powerful computer will be available to the experimental facility, at an operating cost that is estimated to be less than if a smaller computer (with less powerful capabilities) were provided solely for this research usage.

 

We plan to use commercially available display and communication means to launch our experimental research work. It is quite possible that ensuing work will disclose research needs commercial equipment cannot meet, and that we shall want to establish research activity directed toward providing special experimental equipment. However, this proposal includes estimates for minor modifications only changes of control positions, changes of position for display surfaces, or installation of special controls and indicators.

 

The basic display and communication means that we have tentatively selected are manufactured by Data Display, Inc., and reproductions of relevant pages from their most recent catalog are attached as References 1, 2, and 3. Only one or two remote monitors are planned now, but the basic equipment is flexible enough to permit a good deal of experimentation with special symbols (using their microgenerator feature), with multiple display surfaces (for subjects, or for concurrent researcher monitoring), with different special control features, and so on.

 

The working station shall be located in an enclosure giving the subject adequate privacy and freedom from distraction, while still permitting direct observation of his work. As research methodology is developed, associated features such as movie cameras and tape recorders, or a special monitor communication station (so that the researcher who is monitoring test subjects can communicate with the computer) may be tried.

 

10

 

b. Planning

 

In Section I, Part C, we determined that our direct research activity could be classified as planning, design, training, testing, and analysis. The last four concern the methodology of the actual experimental development of highly effective ways and means to perform types of test tasks as specified by the planning activity. The framework for much of the planning activity has already been established in work jointly sponsored by Stanford Research Institute and the Air Force Office of Scientific Research. This work will continue, and is expected to take care of the necessary planning effort during the first year.

 

c. Methodology and Programming

 

A critical area in the actual program of research will be the interface between the subject and the computer. However, the broader interface between the researchers and the experimental environment (subjects, computer, and researcher-oriented artifacts) must be examined first. The methodology to be used in the research will affect the planning and layout of the research facilities. Hence a good deal of this methodology must be evolved during the first year.

 

For instance, the subjects will be trained as well as tested at the working station, learning and exercising skills and concepts involving man-computer cooperation. It could be of great value to the program to have the subject trained without any apparent human participation from the researchers. Thus, he would be learning his prototype processes, and the concepts, strategies, methods, and rules of thumb for using these processes to attack test tasks, entirely in automated form from the computer. A considerable amount of planning and programming would be needed to provide this capability, plus continuing work later to refine the system and add new teaching programs to keep up with additions and changes to the subject matter being taught.

 

Again, the subject may not be able to distinguish between receiving training and receiving testing when he is interacting with the computer. His performance could be being monitored during the training process, and with standardized training procedures, data could be obtained from which predictions of effectiveness could be made early in the training process and patently unsatisfactory processes or techniques eliminated or corrected with a minimum of wasted effort.

 

For the purposes of both testing and training' it is very likely that the researcher who is monitoring the subjects' performances would benefit from having a certain amount of real-time communication capability with the computer. For a simple setup the researcher may be positioned behind the subject (able to observe the subject, his display, and most of his movements without distracting him) and may have some simple controlcommunication panel of his own. With

 

11

 

such communication, the researcher could tell the computer to begin logging a certain kind of performance, to repeat something because the researcher wants to observe it again, to speed up or slow down the testing or training process, to go to a different level of training or testing detail, etc. It may also be valuable for the monitoring researcher to have available a certain amount of communication from the computer; several means to provide this are possible, but since the type of display system contemplated could easily handle several concurrent display tubes, it may be worthwhile to provide the researcher with his own cathode-ray-tube display. In other words, real-time man-computer cooperation may very well be advantageous for the real-time job of monitoring and controlling subjects training and testing. It would be relatively easy to provide (the basic techniques for cooperation are being developed anyway) and it is quite compatible with our "regeneration" philosophy.

 

The analysis activity is a very important area for development of research methodology. This will involve the capability of analyzing a functional, human-directed system to establish criteria of effectiveness within the framework of our objectives, and to direct the monitoring and testing in a fashion that will yield the data necessary to evaluate effectiveness under these criteria. The analysts will need to be able to determine which of the controllable factors involved in the system's activity are limiting the effectiveness, in order to suggest to the designers of the artifacts, processes, and strategies innovations that will circumvent bottlenecks.

 

Computer programming will occupy a relatively large part of the first year. Aside from the computer programming associated with getting ready for the training, testing, and analysis activity, there will be the considerable amount of programming necessary to be ready for providing prototype processes for the subjects. The computer is to interpret and execute immediately each command from the subject-requiring interpretive and executive programs to be developed. The commands of the subject will be in the "language" of his prototype processes and the hierarchy upon which they are based, and the interpretive powers of the computer program must grow and shift with developments of the hierarchy.

 

Since the computer processes will be primarily symbol manipulation as opposed to a numerical computation, and since we should benefit generally from as much subtlety and flexibility as this type of computer processing can give, our expected programming needs have much in common with those of workers in the field of heuristicprogram artificial intelligence. We therefore expect to make use of the elegant programming techniques they have evolved, as represented by the special languages IPL-V, LISP, and COMIT. The future computer processes anticipated (from research and research-methodology planning) will therefore shape our choice and development of a basic symbolmanipulating programming system, and of the interpretive and executive programs which match to the experimental system.

 

12

 

3. Tentative Schedule

 

Figure 2 shows the tentative scheduling for the growth of activity during the first two years of our program, and also summarizes the activities of the five functional roles into which we have divided the program for planning purposes. The chief determinant of the transition in general activity, represented tentatively in Fig. 2 by the one year mark, will be the availability of the computer. Stanford University is presently scheduled to receive its B5000 in April 1963.

 

Activity beyond the indicated two years would have the same essential functional breakdown as shown for the second year. The nature of the test tasks and the prototype processes would continually become more sophisticated, as would the methodology of the researchers, and a certain amount of retrospective redesign activity would appear.

 

C. Significance of This Research

 

If the understanding and solution of complex problems is of importance, then it is of importance to find ways and means of dealing more effectively with complex problems. Accordingly, this research program has the following significant advantages:

 

(1) Our approach is one in which the characteristic attitude of synthesis (i.e., the design of a new functional system) dominates. This is opposed to the approaches in which analysis (the study and understanding of existing systems- natural or artificial) dominates. Analysis is necessary; however, the objectives will only be achieved by coordinating and applying the knowledge supplied by analysis. Our research approach involves a direct attack upon the objectives stated.

 

(2) We have structured our attack in such a way as to accommodate the over-all picture of a human in his problem solving activities. Thus, our approach is designed to produce an over-all solution.

 

(3) We have isolated an explicit starting point and developed a conceptual framework within which we can work in a methodical, progressive fashion towards the comprehensive solution we seek.

 

(a) The particular problem area we will start from provides us an activity that can be controlled and is of manageable size in which to learn how best to do this particular kind of research.

 

(b) This problem area in which we first will learn how to do this kind of human-augmentation development represents the very fundamental type of activity

13

TENTATIVE PROGRAM TIME SCHEDULE

[click for full-size Figure 2]

Page 14

upon which the computer augmentation for any type of intellectual worker would very probably be used. Our achievements would' therefore' be more than interesting examples of what could be done' they would represent something from which could be continued a direct march toward developing the means to increase significantly the intellectual effectiveness of human beings.

 

(c) Our approach is an experimental one in which empirical models and findings can advance the work without first establishing an unattainable theoretical understanding of human cognitive and problem solving processes.

 

(d) The development or progression of our research stages is designed to allow our research methodology to grow systematically toward the stages of sophistication that will be needed to achieve our final goal.

 

(e) Our research has been designed to have the regenerative feature outlined in this section' i.e., any improvement affected by our research efforts can immediately be applied to increase the effectiveness of our future research efforts.

 

(4) Our research also is designed to provide very practical offshoots in the way of design specifications for equipment, procedures, processes, methods, etc. that would be significantly valuable to many problem-solving specialists in the real world. Some of these would naturally evolve from our program in the probable order in which they are listed below.

 

(a) Composers and editors of specialized types of data that may be computer-accessible. For instance, people who may have to schedule certain kinds of complex activities and need to be adept at juggling, rearranging, and structuring the portrayal of tentative schedules and notes pertinent thereto.

 

(b) Real-time monitors of experiment situations. Here our own experience in real-time monitoring through man- computer cooperation of our ongoing experimental situation can be applied.

 

(c) Information specialist. An extension to the capabilities listed under Item (a). Here people would have the methods and techniques for setting up files and retrieval systems in which they could be

 

15

 

adept at indexing and retrieving information within a special domain. For example, they could organize, explore, and portray significant relationships among such things as the growing structure of understanding, experimental data, conjecture, consideration, etc. involved within a research program, or the growing set of externally available information that accrues in an ongoing research activity.

 

(d) Computer programmers. Rapid ability to construct, test, and modify complex process structures would be directly applicable.

 

(e) Teaching machine programmers (i.e. people who

write programs for automated instruction). Mancomputer cooperation for this kind of activity should evolve within our program in a form to be usefully transferred to other automated teaching activities.

 

D. Facilities Available

 

Stanford Research Institute presently does not possess the hardware that is directly applicable to the experimentation planned. One of the specific objectives of this proposal is to obtain such experimental equipment. The present computer in our computation center is not suitable for time sharing (a Burroughs 220). However, it is planned to acquire a Burroughs B5000 to which we will have direct tieline access for the required realtime first-priority access.

 

The Institute does have office, laboratory, and shop facilities, and the appropriately experienced personnel, for research, development, design, modification, construction, or maintenance of any of the types of equipment that might be needed in our program. This includes digital computers, special cathode-ray-tubes or cathode-ray-tube displays, and special instrumentation to monitor human activity.

 

II SUPPORTING DATA

 

A. Previous Work Done on This Program

 

An extensive period of thinking, studying, and planning has preceded this proposal. The work proposed represents the primary professional dream of its author for many years and his primary professional activity for the past two years. Since early 1959 SRI has provided partial support for this pre-program planning and beginning in April 1960 has provided full support from its own funds. In March 1961 the Air Force Office of Scientific Research began contributing half of the cost of this pre-program study.

 

16

 

Within the Institute as a whole there is a wide range of professional experience that is relevant to the activity proposed, including behavioral scientists, systems analysts, management scientists, programming researchers, information-retrieval experts, and engineers with experience in computer development and display development.

 

B. Results Obtained by Others

 

Over the past two decades a number of people have speculated upon the possibilities of close manmachine cooperation and made some fairly definite suggestions thereto. Perhaps the two most significant of these are Bush(4) and Licklider(5). A great deal of relevant work has gone on within the military command and control framework, but most of this apparently is classified material. A journal paper which apparently reflects some of this activity was presented by Vazsonyi(6). A current intensive effort toward making real-time computer access available to a number of users by means of time-sharing the services of a large computer is now going on at MIT and is reported by Morse and Teager(7). An independent example of close man-computer cooperation between a single individual and a computer to solve a mathematical problem will be reported soon by Culler and Huff(8).

 

C. Personal Publications

 

The only journal publication that is relevant appeared in American Documentation --an argument given to the information-retrieval researchers that automation for the individual intellectual worker was

 

________________

 

(4) Bush, Vannevar, "As We May Think," Atlantic Monthly (July 1945).

 

(5) Licklider, J. C. R., "Man-Computer Symbiosis," IRE Transactions on Human Factors in Electronics (March 1960).

 

(6) Vazsonyi, A., "An On-Line Management System Using English Language," Proceedings of the Western Joint Computer Conference, Los Angeles, California (May 9-11, 1961), p. 17.

 

(7) Morse, P. M. and H. M. Teager, "Real-Time, Time-Shared Computer Project," MIT Fourth Quarterly Progress Report, Contract #Near-1841(69) DSR #8644 (October 31, 1961).

 

(8) Culler, G. J. and R. W. Huff, "Solution of Non-Linear Integral Equations Using On-Line Computer Control," Paper to be presented at the 1962 Spring Joint Computer Conference, San Francisco, California, May 1-3. (Reproduction of preliminary draft enclosed as exhibit accompanying proposal.)

 

(9) Engelbart, D.C., "Special Considerations of the Individual as a User, Generator, and Retriever of Information," American Documentation, Vol. 12, No. 2, April 1961, pp. 121-125.

 

17

 

just as sure in coming as was automation in their big retrieval systems, and that there were a number of ways in which this would be very significant to the trends that their research could or should take.

 

The final report(10) or the first year's work under support of the Air Force Office of Scientific Research is a formal write-up that offers very relevant background material.

 

D. Justification of Budget

 

It is proposed to launch a basic attack upon a problem area of significance to society. For this proposed attack, a rather high level of effort seems imperative. A number of disciplinary backgrounds should be represented among the researchers (system analysts, psychologists, programmers, computer engineers, psycho-linguists, and industrial engineers). The level of competence needed prohibits using untrained personnel, and the nature of the program prohibits working in isolation, without sophisticated team interaction. Also, a variety of minds should be applied to the development of the concepts and methods to reduce idiosyncratic effects. An experimentally based program seems definitely called for, and the amount of flexibility and sophistication designed into the experimental facilities is of great importance to the total progress of the research.

 

1. First Year--Preparation for Research

 

a. Personnel

 

The equivalent of five and a half professional man-years, and two and a half semi- or non-professional man-years, are estimated. The principal investigator will spend full time upon the program, but only half-time support from this grant is estimated because of the partial support already committed by the Air Force Office of Scientific Research for his continued work on the long-range, conceptual aspects of the program. His role will be to supply over-all technical direction and co-ordination. Other personnel in the estimate include:

 

Design Engineer (1 man-year)--to participate generally in development of our research methodology, as well as to have specific responsibility for specifying, purchasing and installing commercially obtained equipment, and for designing, constructing, and installing special equipment or modifications to commercial equipment.

 

Research Scientist (2 man-years)--to be responsible for the development of research methodology, to design initial test tasks and prototype processes, and to guide the development of research facilities and computer programming.

_________________

(10) Engelbart, D.C., Augmented Human Intellect Study, Final Report on AFOSR contract AF 49(638)-1024, to be published March 30, 1962.

 

18

 

Computer Programmers (2 man-years)--at least one man to be of the programming research type, who will contribute generally to the tasks listed above for the Research Scientist and who will also have specific responsibility for developing the computer programs required for our research.

 

Research Assistant (1 manyear)--to contribute generally to the development activity, to give supporting service to the other researchers, and to become indoctrinated in the evolving methodology so that he (or she) can handle routine aspects of dealing with test subjects and the general experimental procedures.

 

Technician (15 manyears)--to build, modify, install, and maintain the laboratory equipment.

 

The above estimate of staff needed is only tentative; it does not represent a fixed staff that will be put onto the program regardless of the work needs that develop. The staffing will grow with the needs, and as qualified personnel become available. The operations of the Institute are extensive and varied enough so that individual programs of this size carry extremely little nonproductive burden. Program staffing would be done both by shifting existing Institute personnel, and by hiring new people.

 

b. Permanent Equipment

 

A flexible display system, for experimenting with a wide variety of display effects, is essential to the experimental work. Our estimate is based upon informal quotations from representatives of Data Display Incorporated for a system that embodies standard features they are now delivering to customers. This system meets our expected needs better than any other commercially available display system that we have discovered, and we feel that their quotations provide the best available estimate of what it would cost us to procure a workable display system.

 

The cost of the working station (exclusive of display system costs) will be mostly in carpentry and sheetmetal work, for wood and sheet metal supplies, and for miscellaneous buttons and lights, control wiring, etc.

 

We will also need electronic test equipment for construction, installation, and maintenance work. The principal items required would be an oscilloscope, a camera to record its displays, and a transistor checker.

 

19

 

c. Other Expenses

 

It is assumed that the two programmers will engage in about eight months of programming effort during the first year; accordingly computer time will be required for debugging and trial runs. A Philco S2000 is available in the community and a program is already available from Burroughs to enable this computer to accurately simulate the B5000 for such purposes. It is very difficult to predict costs for such activity; our estimates are based upon the experience of our own programming staff in doing related types of programming.

 

2. Second Year Direct Research

 

For the second year, a level of expenditure representing research effort at a level involving eight full time professionals (including fulltime of Principal Investigator) is estimated, plus one year each for a research assistant and a technician. Less construction work is contemplated, which should reduce technician time. The equivalent of two more research scientists seems desirable since the ongoing activity includes application of the methodology as well as continued method development. We can integrate new people at this rate. It will be important to have a number of different disciplines introduced into the thinking to a greater degree than consultation arrangements can offer. In the second year, there also would be a good probability of revealing important modifications or additions to the research facilities. This is estimated on the basis of additional online work stations for subjects or experimentors (the latter to obtain realtime monitoring and control of experiments), up to the capability of the basic display system to generate the necessary displays. We would continue to incur charges for computer service, and could expect experimental modifications during the year to the structure and layout of the work station.

 

For the third and later years, we assume a level of expenditure consistent with a staff of ten professionals, increased charges for computer time, and continued costs for modifying and extending the research facilities.

 

20