This is the first issue of Reckonings, a weekly newsletter from The Quiet Cost.
Each week, I will gather a handful of the most important AI stories and studies from the previous seven days. I will explain what they actually found and ask what those findings could mean for each of us and for the society forming around the technology. We will pay attention to what AI can do, but the harder question is what repeated reliance on it may leave us less able or less willing to do for ourselves.
Each issue will also offer something practical. It might be a game, a Studio exercise, or a short writing prompt meant to keep one of those human capacities in use. These are small exercises, and some of them are fun. The things worth protecting should still have a place in ordinary life.
This week, three pieces of research circle the same question. A new review asks whether AI is broadly weakening cognition or allowing particular skills to fade. A series of experiments found that even ten minutes of assistance could leave people less persistent once the help disappeared. And a small Stanford study offers another possibility. When AI is used as rehearsal before a human conversation, it may strengthen the capacity it was built to support. Together, they draw a line worth watching between help that finishes the work and help that leaves us more capable of doing the next part ourselves.
THE HELP THAT LEAVES SOMETHING BEHIND
|
|
AI may not be making us stupid
A new review in Trends in Cognitive Sciences takes the alarm seriously and makes it smaller, which also makes it more useful.
The researchers distinguish the skills we acquire through practice from foundational abilities such as attention and working memory. The evidence is strong enough to worry that fully offloading work can interfere with learning and allow particular skills to decay. It is far less clear that AI is broadly diminishing our underlying cognitive abilities.
How we ask for help appears to matter. An AI that completes a problem removes the practice that problem once required. An AI that asks questions, offers feedback, or helps us find the next step can leave more of the work with us.
The problem begins when assistance removes the part we needed to do.
|
|
|
Ten minutes of help changed what happened afterward
An April preprint that received fresh attention this week reports a series of randomized trials involving 1,222 people. Participants worked on tasks that included mathematical reasoning and reading comprehension.
AI assistance improved performance while it was available. When the assistance disappeared, people who had received it performed worse on their own and were more likely to give up. The effects appeared after only a brief period of use, roughly ten minutes.
Within the AI-assisted group, people who said they asked for direct answers fared worse than those who said they used the tool for hints or clarification. That comparison was based on self-reported behavior rather than random assignment, so it is a signal, not proof.
Still, the study identifies something a completed task does not show us. A person can finish more and become less willing to continue.
|
|
|
A chatbot that sent people back to people
The third finding points in the other direction.
Researchers at Stanford built Noora, an AI program that let autistic adolescents and adults practice empathetic responses. Participants responded to short statements, received feedback, and repeated the exercise over four weeks. The researchers then looked for improvement where it mattered, during live conversation with another person.
In the final randomized sample of 30 people, the group using Noora showed a large improvement in appropriate empathetic responses during those conversations, while the waitlist group changed very little. The sample was small and the finding needs to be replicated.
But the design is worth noticing. The AI was not the relationship. It was practice before the relationship.
|
THE RECKONING
FOR THE INDIVIDUAL
Most of us are not going to choose between an AI-filled life and an AI-free one. We are going to make dozens of smaller decisions about where the tool enters the work. Before the first attempt or after it. As an answer or as a question. To finish a thought or to show us where our own thought broke.
Those choices feel procedural. Repeated long enough, they decide what we still practice.
After using AI, it may help to ask three questions. What did the tool do for me? What did I still have to do? What can I do now that the tool is gone?
If nothing remains except the completed task, the help may have spent something.
FOR SOCIETY
Schools and workplaces will be tempted to measure the work that gets finished. The essay was submitted. The analysis was completed. The client received the answer before lunch.
Those outputs cannot tell us whether judgment is being formed underneath them.
If we reward only immediate performance, we will build systems that quietly remove the conditions that produce expertise. Someday, the beginners using those systems will be asked to supervise the machines that helped them skip being beginners.
The cost will arrive late.
PRACTICE
THIS WEEK’S PRACTICE · THINK FIRST
The Quiet Cost Practice has a Studio called Think First. The exercise is simple by design. Before asking for assistance, make an attempt. Write down what you know, what you think, and where you are stuck. Then bring in the tool.
The point is sequence, not refusal.
Try the Studio, then play one of the four small daily games while you are there. Each one is a quick, enjoyable rep for a capacity that becomes easier to neglect when the friction disappears.
PLAY AND PRACTICE →
YOUR TURN
Choose one task you now give to AI that you once did yourself.
What part of you did the old version of that task exercise? How could you preserve that part without giving up the help?
Set a timer for five minutes. Write the answer before asking any machine to help.
PASS IT ON
If this Reckoning made you think of someone, forward it to them. If it was forwarded to you, join us at newsletter.thequietcost.ai.
Michael McNamara
