written prior to completion of IP in 1974. Development of Sandin Image Processor 1971-1974.
"The document you see before you is an early draft of a grant proposal, or more probably, a series of grant proposals. It is also a relatively incomvenient (but at least on paper) forum for some of my thinking about man-machine interactions and man-machine education."
WHY EDUCATIONAL MACHINES AT ALL
The reason for using educational machines is similar to the reasons for using machines elsewhere i.e. to do something that is already being done cheaper or better or to do things that cannot be done presently. The public educational system is incredibly labor intensive. The average capitalization of the public school system is $10/per participant while a modern communication company may have a capitalization of $20,000/per worker. The hope is, of course, that by the development of, and, the investment in the appropriate educational machines, then, the whole system will become economically efficient.
In the transfer of the content of courses previously taught by humans to machines there is much implicit and incidental learning which is not transferred. Iin addition, there are new kinds of information generated by the technology itself
which needs to be learned. It is in these areas that many criticisms of teaching machines are levied and where ruch work is needed.
The advent of generalized machines is quite recent. In the past a tool was developed for (or developed by doing) a single or small set of tasks. The greater part of the industrial system is patterned after this. A car factory not only just makes cars, but just makes one kind of car, and has to shut down for extended periods to change to new yearly models. It apparently takes a long time to learn how to use generalized machines in a generalized way.
A tool which is used in a fairly generalized way is the pencil. It is used to write novels, sketch pictures, take notes, doodle, write text books, draft precise plans, write letters, and more. Television has the same potential generality, and can do things corresponding to novels, sketches, notes, doodles textbooks etc., but is now only used to do novels and an occasional textbook (with the notable exception of real news-time and sports coverage.
Probably the most impressive generalized machine in our culture is the general purpose digital computer. The programs in modern computers change from one task to a completely disrelated task at 1/60th second intervals. The machine can be drawing the floor plan of a house, solving engineering calculations writing a report, and translating a language, all at apparently the same time. This, however, is deceptive generality. From the user's point of view the computer is a set of special purpose tools. The user cannot typically come to the computer with a generalized problem and use generalized methods to solve it. He must call on a series of highly specific routines to deal with parts of the problem.
A generalized device is not simply one which can solve a large range of' problems but which has a high probability of solving tasks for which it was not specifically designed. In the case of computers it must be re-prograinmed to deal with problems significantly different than the ones for which the program was originally designed.
For ecological and efficiency reasons, generalized machines and processes will become more necessary and more used. It is necessary for people to learn how to use generalized machines and processes.
In addition to learning how to use generalized machines it is necessary that work be done in deciding how our generalized machines of the future are to behave. How can they be designed to enhance our consciousness? How can they be designed to be less rigid and less demanding on us? How can they be organized to be true adjuncts to our consciousness? It is common parlance to talk about how dehumanizing our machines are. Although high technology is probably necessary to solve many human problems now confronting us, there is a large anti-technological attitude on the part of young people. Computers and electronic communications technology are the magic wands and crystal balls of the future. Experimentation must be done to decide how these wands and balls are to operate and interface.
"Eyes have They, But They See Not", is the title of an article in which Rudolf Arnheim discusses the problem a generation has that has lost touch with its senses. In most schools (with the exception of art, architecture, and design schools) the emphasis on linguistic behavior is so strong that sensory and kinesthetic behavior is relegated to, at best, a pleasant pastime, and, at worst as interfering with higher pursuits. Perception is not simply the transmission of outside events inside to be linguistically processed by the brain but a complex process in itself. Scientists strongly rely on visual representation and physical models to aid them in their work. Many discoveries begin with the perception of visual correlation of previously uncorrelated data. Einstein wrote that his initial grasp of the theory of relativity was a kinesthetic image, a certain feeling that lie would get through his body. Problem solving as taught at the Bauhaus Institute of Design in Chicago and many other design schools strongly depends on the recording of perception. The point here to be .made is not just that students must have more perceptual exercises but also, that our machines must have much more elaborate input output devices, and perceptually organized structures, to be efficiently used by humans and to contribute to a perceptually enriched culture.
MAN MACHINE INTERFACE
The man machine interface is where human flesh hits steel. It has been little improved since the typewriter. In implementing most activities, man, is strongly dependent upon feedback. For instance, in a simple task of walking across the room and turning on a light switch, one does not measure his present position relative to the light switch and then execute a straight line motion based on that measurement. But instead, one constantly measures his position relative to the light switch and constantly makes small corrections according to his directions. His ability to execute the task accurately does not strongly depend on the stiffness of his muscles, the texture of the floor, or the weight of the coat lie is wearing, but it does depend on the position that the feedback forms from his actions. If man did not use corrective feedback in muscular control, lie would be affected by the seemingly irrelevant factors mentioned in the example.
The feedback control model (cybernetic model) can be extended to a great range of human behavior. A property which remains important throughout this extension is that the character of the feedback strongly determines the charactor of the behavior. In the design of machines and educational environments, the charactor of the feedback is critically important.
In most educational environments, including the use of conventional teaching machines, the feedback is delayed, not very specific, (i.e. correct, incorrect and usually applied only upon completion of a task, i.e. grade on a paper. This kind of feedback is very sparse when compared to the feedback associated with driving a car, walking through a forest, or conversing with someone.
An essential feature of feedback control systems is that it has to be able to vary its action in order to optimize the action under feedback. In other words, in order to find out if a particular variable is optimized it is necessary to wiggle the-value of some action to see if and in what direction corrections are necessary.
In terms of man machines interfaces the machine must be able to sense variations on the part of the man to be able to return usable feedback. The more precise and detailed t he interfacial connection the more precise and detailed the feedback, and, hence, the more precise and detailed the original action can be.
Development of more perceptually oriented interfaces, and interfaces capable of recognizing and producing nuances of action are necessary to the development of more effective educational machines.