Education: Programed Learning

Can machines replace teachers? Probably not. But if they can help teachers to teach, and do it effectively, they may at least help solve the teacher shortage.

The teaching machines’ stoutest advocates feel that they can do much more than that. They argue that “programed learning,” which is what they feed into teaching machines, can not only improve and speed up teaching, but can greatly increase the pupil’s ability to grasp and retain. Programed learning is being tested on thousands of youngsters across the nation. Already it promises the first real innovation in teaching since the invention of movable type in the 15th century.

Programing automates teaching by: ¶ Breaking information into small, sequential steps that can be exhibited one by one in a machine (or page by page in a book). The program writer is compelled to use the utmost logic and clarity.

¶ Questioning the student at each step, riveting his attention and rewarding him —immediately and continually—with the satisfaction of being right. This experience of fast “reinforcement” sets the lesson in the student’s mind.

The notion that learning should be hard gets little backing from psychologists. The Roman schoolmaster Quintilian, 1,900 years ago, said: “A student should strive for victory, yes, but it must be arranged that he gains it. In this way, let us draw forth his powers with both praise and rewards.” Such views lead psychologists to take a dim view of modern education.

They see attentiveness being killed by the dull or incomprehensible textbook, and by the lock-step teaching that bores the bright and overtaxes the dull. The principle of “praise and rewards” is diluted by everything from the sarcastic teacher to the delayed exam grade; teachers have in fact commonly used the threat of failure to induce learning.

Pigeons Playing Pingpong. The new boom in programed learning goes back to 1954 and takes as its father Harvard’s eminent Behavioral Psychologist Burrhus Frederic Skinner. By “conditioning” experiments. Skinner had produced such laboratory oddities as pigeons playing pingpong. Pigeons are hardly bright, but Skinner made them smart by one-step-at-a-time teaching, immediately “reinforcing” each correct response with a grain of corn. Soon the pigeons blithely pecked a ball back and forth across a small table.

Applying the same method to human education, Skinner argued that any subject with a logical structure can be “taught in half the time with half the effort.” Skinner first did it himself in 1957 by programing a Harvard course in human behavior. Now the best-known programing approach, Skinner’s linear method breaks a subject into small “frames” with write-in answer blanks, followed by correct answers (see box).

To control the process, Skinner insisted on a machine. On his prototype machine, the size of a portable record player, the student pulls a lever to make a frame appear in the window. He ponders, writes his answer, pulls the lever again. The answer moves under glass (to prevent his changing it) and the correct answer appears. As the Skinnerian student clicks along, he concentrates fully on each item, advancing only when he is ready to answer. If he gets spring fever he may stop work, but at least he misses nothing, as he would in class. If he wants to soar ahead, he can.

To strengthen reinforcement, the student is fed cues such as the first or last letter of the answer. Gradually, the cues are withdrawn so the student can stand on his own. Medical students see labels on an anatomy diagram disappear in successive frames. Third-grade spellers may learn “manufacture” in six frames: 1) Manufacture means to make or build. Chair factories manufacture chairs. Copy the word here. — ™™™™™™™™™™™™™™™™™the word here: 2) Part of the word is like part of the word factory. Both parts come from an old word meaning to make or build. manu ™™™™™™™ ure 3) Part of the word is like part of the word manual. Both parts come from an old word for hand. Many things used to be made by hand.

™™™™™™™™facture

4) The same letter goes in both spaces: m ™ nuf ™ cture

5) The same letter goes in both spaces:

man ™ fact ™ re

6) Chair factories ™ ™ ™ ™ ™ ™ ™ ™ ™ ™ ™ The student is not supposed to get wrong answers; Skinner argues that though the student must be continually tested, he must be kept generally right. In fact, too many wrong answers prove that the program must be rewritten: Skinner aims at an ideal error rate of only about 5%. The fact that it can be tested and improved is supposed to be one of programing’s assets: it is far more flexible than a textbook.

But do tiny steps really lead anywhere? A clue comes from comparing the first simple frames of a typical program with later ones. Here is frame No. 3 of Mathematician Lewis D. Eigen’s Sets, Relations, and Functions, recently published for junior high schools by New York’s nonprofit Center for Programed Instruction:

“The set of whole numbers from o to 9 is made up of 10 numbers. They are 0, 1, 2, 3, 4, 5, 6, 7, 8, — Simple. But here is frame 89, which an average student might reach just one hour later:

” (7) ⊂ (1, 3, 5, 7, 9)’ means that (7) is a — of(1,3,5,7,9). Algebra Speedup. Theintervening 85 frames obviously carried the student a long way.Eigen and Teacher P. Kenneth Komoski did some pioneer math programing at Manhattan’s Collegiate School, one of the oldest U.S. boys’ schools, and in one case 74 students completed in two weeks a highly abstract algebra course that used to take more than two months. A programed course in logic at Hamilton College cut class time by one-third. At Columbia University, one student wrote a perfect.final exam after doing one term’s math homework in four hours and 20 minutes. In Roanoke, Va., last year, 34 eighth-graders finished a one-year algebra course in six months, working 50 minutes a day without homework; 41% of them tested at ninth-grade level or better.

Does programed learning stick? Pro-gramers so far have precious little evidence. But in Roanoke this month, 25 of the original 34 guinea pigs were tested again, averaged 90% or better of their first scores. Psychologist Allen Calvin of Hollins College, who ran the experiment, called the results “truly striking.”

Linear Y. Branching. Skinnerians have proved something, but not to the satisfaction of a rival school of anti-behavioral programers led by Psychologist Norman A. Crowder of U.S. Industries’ Western Design & Electronics division in Santa Barbara, Calif. While Skinner deplores multiple-choice questions because they contain “plausible” errors that students may remember, Crowder bases his whole approach on multiple choices. Instead of small steps, Crowder programs big chunks of information followed by a question with alternate answers. Choosing a right answer wins the student an advanced frame; a wrong answer sends him to a remedial frame with an explanation of his error and perhaps a chiding comment (“Come, come now”).

In Crowder’s lively “scrambled books,” which Doubleday publishes as TutorTexts, the reader starts on the first page, is then sent scurrying to widely scattered pages throughout the book. This is the “branching” technique, which Crowder developed as an Air Force psychologist while tutoring technicians in troubleshooting on 8-47 bombsights. Crowder believes that his method is better fitted to individuals than Skinner’s somewhat Orwellian linear system. Crowder’s method is demonstrably effective in such problem-solving areas as labor-management relations.

Skinnerians already use forms of branching, such as speeding up programs for the bright by using “diagnostic” frames (“If you got the last six questions right, skip the next 30″). Though Skinner’s method may seem applicable only to fact learning (spelling, vocabulary, scientific terms), it” has already been used for everything from chess and chemistry to history, navigation and philosophy. It is also possible to program the new “concept” approaches in math and physics. The only real limitation is the programer’s ability to analyze precisely how to define and achieve mastery of a subject. This seems to exclude creative arts and personal morality, but programers are an optimistic lot.

Machines or Books? There are three leading models of machines. Simplest is Grolier-Teaching Machines Inc.’s Min/ Max ($20): the student slides pieces of paper through it with a pencil eraser. Rheem Califone’s Didak 501 ($157.50) follows Skinner’s original design, with the programing on paper tape. Crowder designed Western Design’s new AutoTutor Mark II ($1,250), a highly sophisticated branching device with up to 5,000 frames of microfilm. Eastman Kodak is well launched on a microfilm device, capable of handling different programs, that would sell to public schools for about $100.

The fact that cheap machines are inflexible and flexible machines are costly leads many a programer to holler go slow. Ken Komoski wants to wait until machines “can be completely flexible” through the addition of sound, motion-picture and automatic scoring devices.” Others propose simply to let programed books do the job. Skinnerian books make turning a page to find an answer and a new frame the equivalent of switching frames on a machine. That permits easy cheating, but book programers argue that interesting programing eliminates the desire to peek ahead. Encyclopaedia Britannica Films’ big programing division uses nothing but books, employing a plastic mask to reveal frames one by one.

Programs remain the key problem: machines are useless without them. Tested programs of full-year courses cost as much as $75,000 to produce, and they are still scarce. Apart from Crowder’s branching school, the leading program makers include Albuquerque’s Teaching Machines Inc. (the programing affiliate of Grolier); Encyclopaedia Britannica Films, which is programing an entire high school curriculum; Manhattan’s Basic Systems Inc., which is testing programs for underdeveloped countries; and Manhattan’s Carnegie-and Ford-financed Center for Programed Instruction, which grew out of the project at Collegiate School, and is now writing and testing programs for public schools.

The Future: Incalculable. As knowledge grows more complex, as automation demands retraining of workers, programing is due for heavy use. Already it is used widely in industry—for example, to teach drug salesmen their wares faster. In schools, it promises to free teachers to tackle things that they have been forced to neglect. Programed spelling, for example, frees the teacher for creative composition (such as writing poetry) that cannot be programed. At present, programing is no full substitute for classroom work. For one thing, two hours of it is enough for anybody at one stretch. But the potential effect is incalculable. Conceivably, programing might change school design and the entire social structure of U.S. youth.

All of this argues for school administrators to try out good programs—soon, if cautiously—remembering that a bad program is no better than a bad textbook.

They should also remember that talk about machines replacing teachers does not square with probable events. Programing may reduce the teacher shortage, but raise the need for truly skilled teachers.

As Programer Komoski puts it: “Any teacher who can be replaced by a machine deserves to be replaced.”