PLATO and the glorious hopes that our cybernetic betters would teach us

The Friendly Orange Glow: The Untold Story of the PLATO System and the Dawn of Cyberculture (Brian Dear 2017; Pantheon) is about a mainframe-based computer system that died before most of today’s young coder heroes were born.

The book is kind of interesting, nonetheless, for what it reveals about human hopes that machines will take over our most onerous chores. We have a vacuum cleaner and a dishwasher, so why not a robot teacher for the youngsters that the government paid us to create?

Harvard professor B.F. Skinner arguably kicked off the idea of teaching by machine circa 1954:

Skinner’s machine is a plywood box some fifteen inches high, wide, and deep. The hinged top opens so that a scroll of pleated paper tape can be loaded inside. Typewritten onto the paper are arithmetic problems to be solved by the student; the answers to each problem are encoded as small holes, like in a player piano, punched in very specific places to denote a corresponding value. When the box top is shut, you can read the math problems, one at a time, through a small opening in the top of the box, showing a small, exposed area of the paper. You might see a math problem, such as “3 + 2,” through the viewing window on the top side of the box. To answer the problem, you manipulate a series of what Skinner called “sliders” that can be moved up or down through slits in the wood, serving as number scales. This particular math problem is looking for a one-digit answer, so only one slider need be moved into place. The correct slider to move is the one that causes numbers to appear in the hole under the “3 + 2.” Pull the slider toward yourself and the number increases up to 9. Push it away and it drops to zero. Once you’ve formed your answer you can attempt to turn a big black knob on the front side of the box. If the answer is correct, the knob turns freely and the scroll advances to the next problem. This is how you know your answer is correct. If the knob is locked and the scroll won’t advance, you know you’re wrong.

Down the river at MIT, Skinner’s ideas were not popular:

Instead of a machine teaching a student, Papert was in favor of children teaching machines, and in so doing, learning about mathematical concepts, not to mention gaining skills in computer programming. “I find Skinner somewhat of a contradiction,” Papert once confessed to this author, “because as a person he’s intellectually very rich and multi-sided and very literate and likes poetry and I think is a great person. When he thinks about children and education, there’s a lot of richness. The form in which it takes when it gets out into the world is extremely,” he said, pausing for a moment as if to choose the next word carefully, “pernicious. He has a very pernicious doctrine. The pernicious doctrine being that you can break up knowledge into fragments and guide children toward acquiring the knowledge like you might involve the behavior of a rat or a pigeon. I find that a contradiction. I find when you think of Skinner as a whole person, he’s so far away from this kind of thinking and practice of education, well, I’m full of wonderment that he isn’t the main critic of the way that his ideas are being used in the world.”

The idea of computerizing teaching was written about as early as 1958 in a paper by three IBM employees: Gustave Rath, Nancy S. Anderson, and R. C. Brainerd. They’d written a system for students on an IBM 650, but noted that it really needed to be time-shared for efficiency (bizarrely, after citing these IBMers for proposing time-sharing, the author of this book has fallen so deeply in love with his University of Illinois PLATO subjects that he attributes the invention of time-sharing to them, circa 1960 (Wikipedia says that the idea goes back to 1954 and John Backus, who was developing Fortran at the time for IBM. and that the first practical implementation was by John McCarthy, the inventor of Lisp (God’s own programming language), at MIT in 1959)).

The original PLATO system (“Programmed Logic for Automatic Teaching Operations”) ran on the tube-based ILLIAC I mainframe, with its roughly 5,000 bytes of RAM (compare to 4 GB on the latest smartphones, so that’s roughly 1 million times more memory!). Thus the very first implementation relied substantially on a computer-controlled slide projector within the “terminal”.

People forget just how capable early computer nerds were. The University of Illinois team behind PLATO was in theory mostly about delivering an educational experience, but they were not shy about building hardware, including plasma displays, or systems software, such as their own time-sharing operating system and their own computer language: TUTOR.

They also experimented with adding custom hardware for teaching:

“Alphabat,” designed to help young children learn the letters of the alphabet, was a new lesson authored by Amy Alpert (daughter of Dan Alpert), one of the high school kids who like Mike Walker worked on PLATO-related projects with Bitzer as mentor. “Kids who identified the correct letter on the screen were given an M& M,” says Mike Walker, “which was ejected by a contraption powered by a washing machine relay…. It was a bit too powerful and occasionally obliterated the piece of candy.”

National Science Foundation was funding a handful of competitors for computer-based training:

Whereas PLATO was open, flexible, and devoid of any particular instructional theory limiting lesson authors to developing lessons that followed a particular design, TICCIT would reflect a single instructional theory burned not only into the software, but into the hardware as well, right in the keyboard with its special keys for RULE, EXAMPLE, PRACTICE. The culture of the TICCIT lab at BYU could not have been more different than CERL. For one thing, staff meetings were far more formal affairs, always starting with a prayer. Meticulous minutes were kept, typed up in memos, and filed away.

The fact that the projects were so radically different appealed to NSF. They already loved PLATO and had been familiar with Bitzer and Alpert for years. They recognized that TICCIT represented a wholly different approach, both at the scale (supporting a maximum of 128 color TV terminals running on a minicomputer, versus PLATO IV’s much heralded 4,096 terminals running on a supercomputer) and in terms of its instructional design model. “TICCIT was prepared,” says Arthur Melmed, “to demonstrate a certain kind of interaction in a relatively efficient way, and I thought that deserved a crack.” In 1972 both systems received roughly $ 5 million each from Congress. The race was on.

The idea of gamified learning was moderately successful on PLATO back in the 1970s:

Torpedo, another lesson focused on similar fractions problems, presented a situation where a student could play against other students or play against PLATO. The player operated a submarine deep in the ocean, above which swam occasional fish, octopi, and other creatures, and at the surface was a ship. The ocean surface served as the number line, this time horizontal, and the player needed to move their sub backward (by entering the desired negative amount, be it an integer or a fraction) or forward, and then the sub would fire a torpedo upward in an effort to hit the enemy ship. If some creature were in-between, it might get hit by the torpedo instead. The game resembled the popular video arcade game Space Invaders that would come out years later, although with Space Invaders there was no need to know anything about fractions; players simply moved a joystick left or right and fired away.

But it didn’t work in general:

December 1975 was the end of the first semester of the “demonstration year” for NSF and the Educational Testing Service (ETS). Testing revealed that the kids utilizing the elementary reading PLATO lessons were reading at a far lower level than the kids who were in the control classes. “The PLATO lessons were actually having a negative impact upon the kids,” says Yeager.

On the third hand, prisoners liked it:

The PCP project revealed interesting insights into the use of PLATO by its user population. “You couldn’t find,” says Siegel, “a more disadvantaged, disenfranchised, turned-off, uneducated group of people than the kind of people you found in prisons. And as you can imagine, survival in a prison means maintaining a kind of tough-guy image. If you are thought of as weak, bad things are likely to happen to you…. And so typically what happens is, a lot of people will not even opt for educational classes because that’s seen as weak. Or if you’re sort of required to sit in an educational class in a prison, you’re likely to try to misbehave or be the class clown, or act in some way that shows your disinterest. Not because you really are, but because you can’t afford to be wrong in front of your peers. So when the teacher says, ‘Where does the comma go in this sentence?’ and you don’t know the answer to that, you say something that indicates you don’t need to— that that’s irrelevant, or those aren’t the words that someone would say, but it would be the equivalent of that sort of blowing it off. You would blow off the task in front of your fellow classmates. That’s the sort of environment that we were stepping into.”

The inmates discovered not only the advantages of Self-Pacing and Immediate Feedback, but the fact that they were free— in the middle of a prison where there is no freedom or privacy to speak of— to learn, privately, at their own pace, and without fear of ridicule or threats of bodily harm or worse. The computer provided a way to learn that they were not used to. No tough-guy act was required, nor would PLATO have even known how tough a guy you were. You could answer a question and be told you were wrong and why you were wrong, and it was okay. You could answer a question and be told you were right, and that was okay too. PLATO provided a safe space for learning.

The original idea of PCP was to develop some courseware for inmates and then deploy it at a handful of Illinois prisons. The Department of Justice funding was expected to last only a few years. It had originally come from the administrations of Nixon and Gerald Ford. When Jimmy Carter was elected, the Democrats looked around for funding to cut, and the PCP project was put on the chopping block. “They saw these kinds of projects as pet projects of the Republican administration,” says Siegel, “and so there was a freeze on all of these projects. An indiscriminate freeze, I mean they didn’t look at ‘Is this a good project or a bad project’— it was a Republican project: bad idea. And we were shut down.”

Was all lost when the Great Father in Washington diverted the river of cash? No, the state of Illinois liked the system well enough to support it with state tax dollars for another 10+ years. (On the third hand, maybe this, plus pensions, contributed to Illinois being out of cash now!)

Control Data Corporation, the vendor of the mainframe on which PLATO had been running, took over the project and began to commercialize everything. The director of the lab at University of Illinois struck a deal with CDC where they would pay big royalties on the hardware and systems software where he had been a co-inventor, but small royalties for courseware that other folks had written. The course authors were enraged, but began fighting over the scraps nonetheless.

The effect the CDC-Illinois deal had on CERL was significant. “It became like independent little companies inside the laboratory,” recalls former CERL staffer Lezlie Fillman, “who were protecting their product, because of this business that they might be able to get royalties from the stuff that they develop. So the math project didn’t speak to the reading project, didn’t talk to PCP, didn’t talk to community colleges, and there was very, very little sharing of the ideas that these groups were coming up with and I’m sure it was because all of a sudden people said, ‘Well, gee, if I develop this and put my name on it as an author, then I’m going to get royalties from it.’ ” People stopped cooperating, she says. “And there became a lot of infighting inside the lab about who was going to get at a particular concept first. Who was going to be able to put their product out there and begin to rack up student contact hours, which is where the royalties came from. Rather than being an all-lab push it became instantly factionalized.”

Not everyone was fighting all the time…

The CERL laboratory itself, intense as it was as a work environment, with people spending much of their lives there, became a site of numerous hookups, affairs, and romances. One former staffer recalls coming to work early on a Saturday morning and finding two other staffers going at it on the sofa in the lounge. More than a few CERL staffers were married to each other. Every now and then, such couples would divorce, then sometimes the two parties would remarry other staffers who had also divorced. [see Illinois family law]

Fifteen years into the project, nobody can agree on how computers should be used to teach.

In mid-September 1975, Robert J. Seidel, an experimental psychology PhD … hosted along with the National Science Foundation a conference at the Airlie Conference Center. Seidel’s sixty invited participants included Bitzer; Suppes; Papert; Bunderson; Stolurow; J. C. R. Licklider, Marvin Minsky from MIT; Fred Brooks, author of The Mythical Man-Month; John Shoch from Xerox PARC; and Andrew Molnar from NSF. The purpose of the conference was to “identify key developments in computer and communications technology that are expected to have major impact by 1985” in the field of education.

Seidel had high hopes for the conference, including that the attendees would reach consensus on the need to include some fundamental, psychologically sound, practical, instructional theories into the designs of systems of the future. His aims were not achieved. “My problem with all of these folks,” says Seidel, “they’re all messiahs!” The luminaries in the computer-based education field were all visionaries, each respective vision different from the others in important ways. The common thread seemed to be ego. … “When I talked to Minsky, he had his view of artificial intelligence and that was the answer for everything. And Papert had the answer in LOGO for everything. Don had the answer in PLATO for everything. It was his approach, and he wouldn’t leave that plasma terminal either.” Dave Liddle, who worked on the plasma panel project at Owens-Illinois and later joined Xerox PARC, shared similar opinions of Papert and Bitzer, the latter whom he measured at 700 “milliPaperts,” explaining that “if the unit of arrogance is one Papert, okay, he’s about a point-seven instead of Seymour.”

The conference proceedings include a transcription of a discussion:

[Don BITZER, the father of PLATO]: I don’t think teachers, the parents, or the federal government are nearly as excited about improving education as they have led us to believe.

[Seymour PAPERT]:… Nevertheless, it might be true that a thing called education is possible, and some of the people believe in it, and it will be installed in the world.

BITZER: My feeling is this will progress because of interest for other than educational, direct educational purposes. It is by meeting these needs that you will be able to get back to education, which is the reason for all of this. But if we don’t include these needs in some way we just won’t make it, we will just become obsolete. We have to be very careful how we plan to get from where we are today to where we are going.

[Larry STOLUROW, a psychology professor]: This sounds like a much more radical change in position than I understood you to say before. What you are now saying, if I am reading it correctly, is that the direct instructional purposes of a computer-based educational system is [sic] the least important component of what you see the total mix to achieve this flexibility. And that the other things will carry the instructional component along.

BITZER: It depends on whose viewpoint you are looking at. If I were the parent buying the terminal, I would be looking at it on another basis. What is it going to do for me? What will it do for my wife, for my family, for my vacation, for my pay, all of these other things? Incidentally, if it happens to teach, that’s great. From my standpoint I happen to think the most important impact it will have is making literacy commonplace in inner-city schools. But I just don’t think that the rest of the world does.

Readers: What do we think about why progress in computer-aided education has been so slow? The hardware is literally 1 million times better than when the PLATO folks started working on this.

More: Read The Friendly Orange Glow: The Untold Story of the PLATO System and the Dawn of Cyberculture