Mental skills are the abilities that help testers set goals, visualize a system, steer their inner voice and control stress. Together with concentration, motivation, confidence and arousal they raise the chance of testing in the flow. They matter more with AI, because a brain that lets a tool do the thinking loses connections that are hard to rebuild.
Key Takeaways
- Ard Kramer holds that testers can never say a system works, only that they tried everything and found nothing that did not work.
- Goal setting, visualization, self-talk and stress control are the four mental skills a tester can train, while concentration, motivation, confidence and arousal are the capacities already inside.
- The cue HAH, short for human, AI, human, puts a person at both ends of every AI loop, thinking first and testing the output afterwards.
- When AI takes over the thinking, use it or lose it applies to the brain. Unused connections disappear and are hard to rebuild.
- Walking around raises brain activity and improves thinking, which makes movement the simplest form of stress control before a hard testing task.
What Happens to the Brain When AI Does the Thinking?
Mental skills in testing start with a plain biological fact. The brain drops what it does not use. Ard Kramer frames the question of AI and the brain in evolutionary terms. Our brains are tuned to be as efficient as possible. Some people call that laziness; Ard prefers to call it being effective. If something else can do a task easily, the brain is happy to let it. With AI that becomes a risk. Type in a prompt, wait, burn a few tokens, take the output. No thinking happened, so no learning happened either.
Use it or lose it is the short version. The brain holds millions of connections, many of them in the subconscious, and using them is what keeps them alive. Hand the thinking over often enough and the connections thin out. Getting them back, Ard says, is a hell of a job. Psychology calls the habit of handing a mental task to an external aid cognitive offloading.
Ard does not argue against AI. He wants testers to use it in a way that keeps their own brains in the loop, and he has a model for that.
Thinking Starts Before the Prompt
Ard’s cue for working with AI is HAH: human, AI, human. The human starts the loop by thinking about what is wanted. AI does its part in the middle. The human closes the loop by checking the result. Both ends matter, and the first one is the one most people skip.
The front end is old craft under a new name. Knowing what you want before you prompt is the same discipline as good requirements engineering. Ard’s advice is to leave the keyboard alone at first. Write in a booklet, draw a mind map, talk to a colleague. A short exchange along the lines of “I’m thinking about this, what do you think?” often produces an angle nobody had looked at. Human heads are densely associative, and they make connections that are statistically unimportant to a model and therefore never show up in its output.
The back end is where testers should feel at home. AI output is software output. Testers check software output for a living, so there is no reason to treat a generated test plan or a generated piece of code differently. Ard puts it bluntly: we are testers, why don’t we test the output of AI? That also means there is still plenty of work to do.
Four Mental Capabilities Decide Whether You Are Ready to Perform
Ard draws his model from sports and splits it into two layers. The first layer is what you already have inside you, the mental capabilities. There are four of them, each with a question attached.
| Capability | The question it asks |
|---|---|
| Concentration | How long can you concentrate? |
| Motivation | How motivated are you, and is the motivation intrinsic? |
| Confidence | How confident do you feel, and can you deal with doubt? |
| Arousal | Do you have enough tension to be ready for the sprint? |
Confidence has a side note worth keeping. Confident people still feel doubt. What sets them apart is that they can handle it when it shows up.
Arousal is the one that varies most between individuals. Some athletes need a lot of it. Ard shows the haka of the All Blacks as the extreme case of a team working itself up before a match. Usain Bolt is the opposite picture, completely at ease on the start line, and he has won quite some medals that way. The point is to know your own level. When all four capabilities are high, you are ready to perform at your best. They are not always high, which is why the second layer exists.
Goal Setting in Testing Means Looking for Proof That It Does Not Work
The second layer is the set of mental skills that move the capabilities upward: goal setting, visualization, self-talk and stress control. Goal setting comes first, and in testing it changes what a test is for.
A goal in Ard’s sense is small and fully under your control. There is a piece of stoicism in it, reduced to one line, me and my task. I do my work as well as I can, independent of what anyone else does. “I’m going to test” fails that standard because testing is an activity, and an activity is no goal. The goal is the information you want out of the system. How much risk is left? Is it good enough? Did something else go wrong along the way? A payment interface makes a clean example. The interface matters because without it nobody can pay, so the test objective is information about whether paying works.
Then comes the turn. Do not look for confirmation that the interface works; one passing test will give you that and tell you very little. Form the hypothesis the other way round and go looking for the conditions under which the interface does not work. Ard calls this turning the world around, and the reason is confirmation bias. If you search for evidence that something works, you will find it. The honest report of a test effort therefore sounds like this. I have tried everything I could think of, and I have not found anything that did not work. Ard considers that the most valuable information testers provide.
Which Mental Skills Get Testers Into the Flow?
The remaining three mental skills work on the inside of the tester rather than on the test objective.
Visualization uses the full-size cinema everyone carries in their head. A mind map, even one that exists only mentally, lets you see what you have covered and what is still open. An architecture diagram does the same job. Ard’s favorite image is a soap opera. In a soap, people marry five times, disappear as if they were dead and come back later. An HR system has to survive the same plots. Someone drops out of the system; can they come back in, and what is the risk if they cannot? Those are interesting test cases, and a picture produced them.
Self-talk is the little voice that tells you to get a coffee first. Ard calls it procrastination, and the counter is a cue. A cue is a short phrase that pulls attention back to the task: interface, payment, important, now. HAH is a cue of the same kind, aimed at the moment you are about to open an AI tool. Stop, wait, think first.
Stress control works in both directions. Sometimes there is too much tension and it needs to come down; sometimes there is too little and it needs to go up. Ard’s main tool for both is movement. Walking around raises brain activity and improves thinking on its own, which is why so many people solve a problem on their feet and come back saying “I got it”.
When goals, visualization, self-talk and stress control are in place, the four capabilities climb. Add a really challenging assignment, and there is a good chance of hitting flow. Flow is the state where time disappears and the work carries itself, and most people only notice they were in it once they are out again. Ard’s goal for testers follows from that. Test in the flow.
The Zeigarnik Effect Lets an Open Task Work for You
Sometimes the best use of your mental powers is to wait. The Zeigarnik effect describes what happens when a task stays open in your head. The subconscious keeps working on it without you noticing. Ard’s own routine is an example. A proposal or a difficult email arrives late in the afternoon and he does not know how to respond. He reads it, leaves it alone, goes home and sleeps. The next morning the answer is there, and it is a better answer than the one he would have typed on the spot.
The effect has a downside as well. Too many open tasks produce stress, because every one of them keeps a process running. The remedy is to close tasks deliberately. Ard’s illustration is a waiter who takes an order for ten different beers and brings each one to the right person. Ask the same waiter at the end of the shift to repeat the order and he cannot. The task was closed, so the brain let it go.
How Do You Use AI as a Critical Partner?
AI tools, like social media feeds, are built to attract attention and keep you engaged. The follow-up questions at the end of every chat answer are part of that design. Ard’s defense is the same principle as before. Thinking comes first, the tool second.
Within that order, he suggests treating AI as a mate, a kind of colleague. Ask it questions. Better still, ask it to ask you questions. Ard uses the same trick in front of an audience. A question triggers the brain whether or not the person asking cares about the answer, and that is the effect you want from a tool as well. An AI that interrogates your test idea keeps you thinking.
The reason this matters is accountability. If you deliver something to your boss and the boss calls it garbage, you cannot say the AI did it. The result is yours either way.
“You are responsible. So always be aware of that. You can’t outsource your brain.”
(Ard Kramer)
