The curse of William Jevons

And the ghost of Maynard Keynes

If you are reading this, you probably read email regularly. Quite possibly you have received a prompt from artificial intelligence linked to your email platform offering to summarize the email you have received or to draft emails you are going to send. The premise — indeed the promise — is that AI will make dealing with your email more efficient. You will have more time to do other things.

It's occurred to me, and maybe to you, that this will never happen. Yes, we might get faster at reading and sending email. But we won't have more free time. Rather the new system will spawn more email and any savings will be lost.

Such has happened before. In the 19th century people in Britain worried that the country was running out of coal. An obvious response was to seek savings in the use of coal. In particular, more efficient steam engines, which were powered by coal, would extend the life of Britain's coal supply.

William Stanley Jevons thought this approach had things backwards. Jevons was an economist who had studied the matter historically. James Watt had greatly increased the efficiency of steam engines with the model he introduced in the 1760s. Had this reduced coal consumption? No. It had precisely the opposite effect. Watt’s efficiency reduced the price of deliverable steam power. In doing so it made the application of steam power more attractive. The efficiency gains in the use of a single engine were swamped by the many greater number of engines put into service.

The Jevons effect, as it came to be called, popped up wherever people looked for it. Thomas Edison's light bulb dramatically reduced the energy required to illuminate a room. But in doing so it made possible the illuminating of many rooms and other spaces, so that the overall energy expended on illumination increased. Times Square in the 1920s was the gaudiest illustration of the principle.

In the field of road construction the concept appears as induced demand. A new crosstown highway reduces the time cost of commuting for individual drivers. But this causes more people to drive, and the consequent congestion makes overall traffic worse.

John Maynard Keynes in 1930 predicted that the average work week for his grandchildren would be 15 hours. The greater efficiency of machines would make this possible. But it didn't happen. Instead of toiling fewer hours for the same amount of stuff, we toil the same number of hours for more stuff.

Many people have predicted that AI will put most of us out of work. What's likelier, if the past is any guide, is that AI will make us richer without reducing our work hours much or at all.

I and some of you reading this will remember when email was new. It promised to make correspondence faster and more efficient. Instead of having to find paper and envelope and stamp, writing or typing the letter and the address, and carrying it to a postal box, now we could correspond with a few keystrokes. It was hard to imagine what we would do with all our free time.

How naive we were. Today most of us spend more time on email than we ever spent on postal correspondence. The curse of Jevons is upon us.

Don’t expect AI to lift it.