1. AI Perception Distortion Phenomenon
AI can perform many ordinary human cognitive tasks in seconds that once took people minutes, hours, or even days. That sounds like an uncomplicated productivity gain. But something else is happening alongside it: as people become accustomed to near-instant answers, the benchmark itself begins to move.
A task that once took two hours may now take two minutes with AI. Eventually, those two minutes stop feeling extraordinary and become the new normal. Once that happens, even a short delay can begin to feel like inefficiency. AI may therefore be changing not only how quickly we work, but how we perceive the speed of work.
2. The Moving Benchmark
Human beings have always adjusted their expectations to the technology available to them. A letter once took days to travel. Email reduced that to minutes or seconds. Search engines reduced the time required to find information from hours in a library to seconds. Smartphones made communication almost instantaneous.
AI is doing something different because it is beginning to compress parts of cognitive work itself. Researching an idea, generating alternatives, explaining a concept, writing a draft, translating text or producing code may once have required substantial human effort. AI can often produce a useful first result almost immediately.
The psychological benchmark consequently changes. What was once fast becomes normal; what was normal becomes slow; and what is slow begins to feel inefficient.
That shift is easy to miss because the underlying historical comparison disappears from our minds. If a person spent an entire afternoon solving a problem twenty years ago, we would not necessarily have described that person as inefficient. Today, if an AI system produces a plausible solution in seconds, the same afternoon can suddenly look excessive.
But the task did not necessarily become intrinsically easier. Our reference point changed.
3. The Two-Minute Problem
Imagine an AI system that normally responds in ten seconds. One day it takes two minutes. Nothing about the human world has become slower; two minutes is still an extremely short period for many intellectual tasks.
Yet the user may experience the delay as a problem because the relevant comparison is no longer, “How long would a human take?” It has become, “How long does AI normally take?”
This is a subtle but important transition. The machine becomes the benchmark against which human time is judged.
Once that happens, we may begin to underestimate the amount of cognitive work that was previously required simply because we no longer experience that work at its original timescale.
4. We Forget What Thinking Used to Cost
Consider a simple research question. Before modern AI assistants, a person might have had to search through several sources, compare information, take notes, formulate a conclusion and then write it down. The process could consume hours.
Today, a person may ask an AI system the same question and receive a structured starting point almost immediately. That is a genuine gain.
But there is a psychological side effect: we stop feeling the cost of the underlying process.
The friction disappears. And friction had a useful property: it reminded us that thinking takes time.
When every intermediate step becomes invisible, the final answer can create the illusion that the underlying intellectual task was itself simple. This is where productivity and perception begin to separate.
5. The Hidden Struggle of AI
There is another difference between human and machine cognition that is easy to overlook: humans reveal their struggle.
When two people work together, one person might say that they are unsure, propose an approach, discover that it does not work, and then ask for another piece of information. The other person witnesses the process.
The struggle is visible. It keeps people inside the cognitive loop because they see uncertainty, iteration, mistakes and changing hypotheses. They understand that reaching the answer required work.
AI often presents a radically different interface. The user asks a question, the system disappears behind the interface, and an answer appears. The visible process can look like question → answer, even when considerably more computation and evaluation occurred between the two.
This creates a peculiar asymmetry:
Humans expose their struggle. AI often exposes its result.
6. The Effort Illusion
Suppose a human spends three hours solving a difficult problem and an AI produces a useful answer in thirty seconds. It is tempting to conclude that the AI solved it effortlessly.
But speed of output is not the same thing as absence of computation, uncertainty or difficulty. The interface hides almost everything between the input and the output.
This matters because humans tend to estimate effort from observable behaviour. Someone struggling over a problem looks as though they are working hard; someone producing an answer instantly does not.
AI therefore creates an unusual effort illusion: the less struggle we see, the easier the task appears.
This can affect how we perceive both AI and ourselves. We may overestimate how effortless AI cognition is while simultaneously underestimating the value of human cognition that remains necessary around it.
7. From Productivity Gain to Perception Distortion
The important point is not that AI is somehow making people imagine that it is fast. It really is fast at many tasks.
The distortion comes from what happens after repeated exposure to that speed.
Suppose the historical human baseline for a task is four hours and AI reduces it to ten minutes. That is a dramatic productivity improvement. But after hundreds of repetitions, the user’s psychological baseline may become ten minutes equals normal.
Now imagine a human colleague takes 45 minutes. The colleague has still completed the task in a fraction of the old baseline. Yet compared with the AI benchmark, the person appears slow.
That is the perception distortion. The denominator has changed.
8. AI Can Make Human Intelligence Look Slower Than It Is
This creates a particularly interesting problem for knowledge work.
A human may spend twenty minutes thinking carefully about a question while AI produces five possible answers in ten seconds. The human now appears slow. But the five answers may still require human judgment: which one is correct, which assumptions are wrong, what information is missing, what matters in the real world, and which consequences were overlooked?
AI has compressed generation. It has not necessarily compressed every part of judgment.
This distinction may become increasingly important as AI systems become more capable. The danger is not simply that humans will become less productive. It is that humans may begin to evaluate themselves using a machine’s output speed for tasks where human cognition operates under fundamentally different constraints.
9. The Benchmark Moves Upward
Every major technology changes expectations. Once calculators became common, mental arithmetic became less important. Once search engines became ubiquitous, memorising large quantities of factual information became less necessary.
AI may do something similar to a much larger portion of cognitive work. But there is a difference in scale: AI does not merely provide a tool for one narrow operation. It can participate in writing, coding, analysis, research, brainstorming, explanation and decision support.
That means the benchmark can move across many forms of knowledge work simultaneously.
Yesterday, the question was often whether you could do something. Today, we increasingly ask whether you can do it quickly. The next psychological step may be to ask why it has not already been done.
That is where perception becomes particularly interesting.
10. The Paradox of Infinite Productivity
There is also a deeper question. If AI continuously increases the amount of cognitive work that can be produced per hour, does human perception of productivity rise at the same rate?
Not necessarily.
Humans still have limited attention. They still have limited time. They still have to read, evaluate, decide, communicate, build physical things and live in the physical world.
AI can generate ten reports, but a person still has to decide which report matters. AI can generate a hundred ideas, but attention has not increased one hundredfold. AI can write thousands of pages, but human readers do not suddenly acquire thousands of hours to consume them.
So there is a strange possibility: AI may make the supply of cognition abundant while making our perception of insufficient cognition even stronger.
The machine gives us more, and we become dissatisfied with how little we can process.
11. The Human Sense of “Enough”
Before AI, the amount of effort required to complete a task imposed a natural stopping point.
If writing a report took two days, finishing the report was an achievement. If researching ten sources took an afternoon, ten sources might have been enough.
AI changes the economics of continuation. Another version can be generated in seconds. Another explanation can be requested immediately. Another comparison costs almost nothing. Another idea is one prompt away.
The marginal cost of asking for more approaches toward zero changes the meaning of done.
When additional cognition becomes nearly free, the human may begin to feel that stopping is premature. Perhaps there is a better version, another angle, or another refinement.
The technology that makes completion easier can therefore also make satisfaction with completion harder.
12. The Human May Need a New Cognitive Benchmark
The answer is not to pretend that AI is slower than it is. Its speed is real, and its productivity benefits can be enormous.
The challenge is to distinguish three different things: machine speed, which is how quickly a system can generate or transform information; human processing speed, which is how quickly a person can understand, evaluate and act on that information; and real-world completion speed, which is how quickly the actual objective can be achieved.
These are not the same.
AI may generate a business plan in thirty seconds, but the business may still take months to build. AI may generate code in seconds, but the software may still require testing, deployment and maintenance. AI may produce ten hypotheses instantly, but reality still has to determine which hypothesis survives contact with the world.
The output can be instantaneous while the consequence remains slow. For a related discussion of what information availability does—and does not—solve, see When AI Has the Information, What Still Requires a Human?.
13. The Strange Future of Visible and Invisible Work
Perhaps the most important adjustment will be psychological.
Humans have historically understood work partly by watching people struggle to perform it. AI removes much of that visible struggle.
The future workplace may therefore contain an increasing amount of hidden cognitive work. A system may perform thousands of operations without showing them. An agent may explore alternatives without presenting every failed path. A model may produce an answer without exposing the uncertainty that preceded it.
Humans will consequently need better ways to distinguish fast from easy, polished from certain, automated from understood, and instantaneous from complete.
That may become a new form of AI literacy.
14. AI Is Changing the Clock in Our Heads
The deepest effect of AI may ultimately be neither automation nor productivity. It may be calibration.
AI changes the internal clock against which we judge intellectual work. Once we become accustomed to instantaneous assistance, human timescales can feel strangely slow. Once we become accustomed to effortless-looking machine output, human struggle can look disproportionately expensive. Once we can generate almost unlimited alternatives, stopping can feel like failure.
And once the machine becomes the benchmark, we may forget that the benchmark itself is new.
AI has not merely accelerated cognition. It may be changing our perception of what cognition is supposed to look like.
That is the AI Perception Distortion Phenomenon.
Next article
Are AI Productivity Gains Worth It?