Amara's law
Amara law · the hype cycle
We tend to overestimate the effect of a technology in the short run and underestimate it in the long run. Roy Amara, who ran the Institute for the Future, is credited with the formulation.
In practice
The thing that was going to change everything this year did not, and the conclusion drawn — that it was hype — is wrong in the other direction. Both errors are made by the same people about the same technology, a few years apart.
The common mistake
Using it as a prediction. It says the shape of the error, not the timing, and 'underestimated in the long run' is compatible with a decade of nothing happening and with never happening at all.
Roy Amara ran the Institute for the Future, and the line attributed to him describes a failure mode that repeats so exactly it is almost a genre: we overestimate a technology's effect in the short run and underestimate it in the long run.
Both halves are needed. Separately they are two ordinary complaints — that people are credulous, and that people are unimaginative. Together they describe one mechanism, which is worth having.
Why the short run gets overestimated
Because a working demonstration and a deployed system are separated by everything that is boring.
A new capability arrives as a demonstration. The demonstration is real. What is not yet real is the integration, the retraining, the regulation, the edge cases, the procurement cycle, the twenty adjacent systems that assume the old way, and the people whose jobs are organized around it. That work takes years, is unglamorous, and is invisible in the demonstration — so the forecast is made from the capability and the capability is the small part.
Why the long run gets underestimated
Because the estimate is made by imagining the new thing inserted into the current world, and the current world does not stay still.
The real long-run effect comes from what gets built once the technology is assumed. Cheap mobile computing did not mainly mean carrying a computer; it meant businesses that could not have existed otherwise. Those are second-order effects and they are structurally hard to forecast, because forecasting them means imagining things nobody has thought of yet.
And the adoption compounds. Each deployment makes the next one cheaper, better understood and more expected, which is a reinforcing feedback loop — slow, then not.
What it is actually good for
Not prediction. The law gives the shape of the error and no timing whatever, and any use of it that produces a date is misuse.
What it gives is two questions to ask of any forecast.
For a short-run claim: what has to be true besides the technology working? Usually a list of integration, regulation and habit, each of which has a duration somebody could estimate.
For a long-run dismissal: am I picturing this dropped into the world as it is? If the only way the technology matters is by replacing a current thing one-for-one, the estimate is missing the part that matters.
Where it fails
It is unfalsifiable as usually deployed, and this is the honest objection. Any technology that has not yet delivered can be defended with "long run", indefinitely. Plenty of technologies were overestimated in the short run and then simply did not matter — and they are not remembered, which makes the law look better than it is. That is survivorship bias operating on the evidence for a forecasting heuristic.
So it is worth stating the version that can be wrong: this technology will matter more in ten years than now, and here is the specific thing I expect to exist by then that does not exist today. Without the second clause it is a posture wearing the grammar of a forecast.
Concept web
Open the full webQuestions
What is Amara's law?
We tend to overestimate the effect of a technology in the short run and underestimate it in the long run. Roy Amara, who ran the Institute for the Future, is credited with the formulation.
Why do we overestimate technology in the short run?
Because a working demonstration and a deployed system are separated by integration, retraining, regulation, edge cases and the systems that assume the old way. All of that is invisible in the demonstration and is most of the work.
What is wrong with Amara's law?
As usually deployed it is unfalsifiable — any technology that has not delivered can be defended with 'long run' indefinitely. Plenty were overestimated and then simply did not matter, and those are not remembered.