Man-Computer Symbiosis

The situation

In 1960 a computer was a machine you submitted work to. You prepared a program and its data, handed them to an operator, and came back hours or days later for the results. The machine worked alone, on fully specified problems. The idea of thinking alongside a computer, posing a question, seeing an answer and adjusting, was almost unknown outside a few laboratories.

J. C. R. Licklider was a psychologist who specialised in hearing and had worked on human factors in air defence systems. He had used some of the few interactive computers then available, and he had begun to think that the relationship between people and computers could be different.

What it argues

He opens with a biological image. The fig tree is pollinated only by a particular insect, whose larvae live in the tree’s ovary. “The tree and the insect are thus heavily interdependent.” Neither can live without the other. Licklider proposes that people and computers could form a similar partnership, a symbiosis, and defines it carefully: “Man-computer symbiosis is a subclass of man-machine systems.” It differs from earlier systems in which the machine merely extended human muscle or replaced human tasks. In symbiosis, each partner does what it does best.

His hope is that “in not too many years, human brains and computing machines will be coupled together very tightly,” and that the partnership will think as no human brain has thought.

The paper’s most memorable passage is a small experiment. Licklider tracked how he spent his own working time. “About 85 per cent of my ‘thinking’ time was spent getting into a position to think,” he reports: finding information, plotting graphs, converting units, getting data into shape. Those activities were, in his words, essentially clerical or mechanical. The decisions that mattered took a small fraction of the time and depended on the preparation being done.

From this he draws the division of labour. People would set goals, formulate hypotheses, choose criteria and evaluate results. Computers would do the routine work of transforming data, calculating and storing, quickly enough to keep up with a person’s train of thought. To make this possible, he lists prerequisites: computer time-sharing, so that many people could interact with expensive machines at once; better memory; languages closer to how people think; and input and output devices such as displays and pens.

He is careful about the future. He cites an Air Force study estimating that it would be 1980 before artificial intelligence would let machines alone do much thinking of military significance. “That would leave, say, five years to develop man-computer symbiosis and 15 years to use it. The 15 may be 10 or 500, but those years should be intellectually the most creative and exciting in the history of mankind.”

The idea underneath

Licklider’s paper is a theory of what thinking is made of. He separates judgement from the preparation that makes judgement possible, and argues that machines should take over the preparation. This echoes Aristotle’s distinction, discussed under practical wisdom, between the deliberation about particulars that belongs to a wise person and the technical work that can be made routine.

It is also a theory of attention, close to Herbert Simon’s bounded rationality. If 85 per cent of thinking time goes to clerical work, the bottleneck of intellectual work is not intelligence but attention wasted on preparation. Freeing attention, not replacing judgement, is the goal.

How it was received and what followed

In 1962 Licklider became the first director of the Information Processing Techniques Office at the Advanced Research Projects Agency (ARPA) of the US Department of Defense. He used his budget to fund time-sharing projects, interactive graphics and research communities at universities, and he wrote memos imagining an “intergalactic network” of computers. His successors at the office funded the ARPANET, the forerunner of the internet. In 1968, with Robert Taylor, he published “The Computer as a Communication Device”, predicting that people would communicate more effectively through machines than face to face.

The interactive, networked computing described across this site, in the operating systems essay, the compute essay and others, owes a great deal to the research Licklider funded.

Reading it today as a decision-maker

The paper’s most practical lesson is in its method. Before arguing for a new kind of tool, Licklider measured how his own work was actually spent. That measurement made the case concrete. Many product decisions would benefit from the same discipline: observe where time actually goes before deciding what to automate.

Its second lesson is about division of labour, which has become urgent again with AI. Licklider’s framework asks which parts of a task are judgement and which are preparation, and assigns them accordingly. The word processors essay shows how unsettled that boundary has become when machines can write. Licklider’s “interim” period of partnership may be shorter, or longer, than anyone expected.

Sources and limitations

Quotations are from the copy of the paper hosted by MIT CSAIL and were checked against it. Biographical details and Licklider’s role at ARPA are drawn from Waldrop’s history. The connections to Aristotle and Simon are the author’s interpretation; Licklider does not cite them.