There has been a lot of discussion lately about whether AI is weakening human cognitive skills. I think that concern is real. I’ve felt it in my own work.
Like most powerful tools, AI does not simply make us stronger or weaker. It changes us. It alters what we need to know, what we bother to practice, and the habits of mind that shape daily life.
How I Use AI at Work
One of the ways I’ve started using AI is through “model context protocol” (MCP) servers integrated with a chatbot. These servers enable it to search our company work tickets and internal technical documentation for context to help improve the quality of chatbot responses.
One especially useful application has been in reviewing software updates and other technical changes before they are deployed to our production systems.
At my company, proposed changes go before a review board for approval. Part of my job is to read change tickets and provide the review board with my opinion about whether they are safe to deploy. Now that the chatbot has access to our documentation and tickets from previous work, I’ve been asking it to help. I give it prompts such as, “Review ticket #12345 and identify any security, stability, or performance impacts.”
The results can be impressive. The chatbot scans the ticket in a fraction of the time that it would take me to do the work. It checks linked work items and documentation and produces a detailed, well-structured assessment. Often it notices issues that I might not have noticed on my own.
I also find AI helpful for drafting things like work memos or task tickets, where no one cares if the document has a “personal touch.” They just want it to be clear and well-organized. I can give the bot a summary of what I want said with links to supporting documents, and it fleshes out the details.
I’ve also used it to analyze error logs and other system data. On one occasion, we had a system outage that stumped a dozen of our top technical engineers for more than eight hours. In desperation, I fed a batch of error logs to the chatbot, and it figured out how to fix the problem in less than five minutes.
That is the upside.
The Three Problems
The downside is that this convenience comes with some hazards.
The first problem is obvious: the chatbot sometimes makes mistakes. It can misunderstand context, overstate a risk, miss an important nuance, or simply invent something that is not true. That means I cannot treat its output as authoritative, no matter how polished it sounds.
The second problem is more subtle. Sometimes the chatbot’s review is so good, or at least so seemingly well written, that I feel tempted to copy and paste its assessment without fully reviewing and understanding it myself. That temptation is not surprising. After all, the whole point of having an AI assistant is to save time and effort. It saves labor, but also invites me to surrender attention.
The third problem is verbosity. The chatbot often says too much. It may generate several pages of commentary, with repetition and unnecessary elaboration, when the necessary point could be made in one or two crisp sentences.
The Discipline I’ve Adopted
To keep those problems under control, I’ve adopted a simple rule: I always read and mentally review the change ticket myself before I ask the chatbot for its assessment.
That forces me to engage directly with the proposed change and reduces the risk that I will become lazy about understanding it. It also gives me a baseline for spotting places where the chatbot may be mistaken.
After I get its review, I edit it aggressively before sharing any comments in the ticket. If the bot produces a long and windy analysis, I may boil it down to a sentence or two. I use the chatbot’s output as raw material, not as a substitute for my own judgment.
So far, that seems like the healthiest way to use it: first think for myself, then compare my thinking with the machine’s, then rewrite the result in my own words.
A Mansion Full of Delegated Choices
What fascinates me most about the way the chatbot changes my own thinking is the temptation it creates for me to stop checking the work carefully. That impulse reminds me of an experience I had during my college days.
A grad student friend of mine once house-sat for a wealthy professor while he was away on vacation. The professor lived in a beautiful home with a swimming pool and other luxuries, but my friend, who was very observant, noticed something revealing about the place. The home reflected the habits of a man who employed hired help.
The professor taught philosophy and had a study with bookshelves including rare books by major philosophers. However, the valuable books were arranged carelessly, as if he himself had not handled them much. Other details suggested that someone else did the shopping, cooking, and other domestic chores. The household functioned well, but everything was vaguely disorganized in ways that we could notice.
I think something similar happens when we delegate work to AI. The work gets done, but our relationship to it changes. We lose some of the hands-on familiarity that we would get from doing it ourselves.
Progress Always Costs Something
The rise of AI is not entirely different from past technologies. Human progress has always involved trade-offs.
When human beings learned to preserve knowledge in writing, memory changed. People in pre-literate societies often had astonishing memories compared with the memories of literate people today. They remembered things because they had to. If knowledge could not be stored externally, it had to be carried in the mind. As writing became common, people could afford to “get lazy” about remembering because they could look up the facts that they needed from books and ledgers.
That trade-off was worth it. Whatever we lost in memory was more than offset by the advantages of preserving knowledge and sharing it across time and distance. But it was still a trade-off.
What Socrates Feared
As far back as ancient Greece, people were already worrying about this.
In Plato’s Phaedrus, Socrates warned that writing would weaken memory and create only the illusion of wisdom. Written words, he argued, were dead things compared to spoken speech. They could not answer questions or defend themselves the way a living mind could.
Socrates was a smart guy, and he was right about part of this. Writing really did weaken certain habits of memory, but it also made possible a vastly more durable and expansive culture of thought.
We may be entering a similar moment now.
What We Gain and What We Lose
Similar patterns shows up in other domains of human activity. Musicians who cannot read sheet music may develop strong skills at learning and remembering by ear. The Beatles famously did not read music. Paul McCartney once remarked that because they couldn’t write down their music, they had to make them memorable so they could remember the tunes when it came time to perform them.
Human brains are plastic. We adapt to the demands that life places on us. Our ancestors needed to remember landscapes, seasons, and oral traditions in extraordinary detail because survival required it. My own generation learned slide rules and paper-and-pencil arithmetic because we did not yet have electronic calculators in our pockets. Younger people today develop different skills because different tools are available to them.
As artificial intelligence takes on tasks that once required human effort, it seems inevitable that human cognition will also change. Some skills we now consider basic may atrophy. Others may become more important. Perhaps the future will place less value on rote recall and first-draft composition, and more value on judgment, verification, synthesis, and knowing when not to trust a machine.
What we will lose, and what we will gain, remains to be seen. The question is whether we will notice the loss when convenience feels like progress.
Leave a Reply