AI Psychosis

An AI assistant can offer an explanation, a draft, and a second opinion in the space of a few minutes. It is easy to ask for another version before deciding what was wrong with the first one.

The risk is a workflow with plenty of conversation and very few completed decisions. More responses can feel like progress because something is always appearing on the screen.

A simple boundary helps: decide what the tool is being asked to contribute before opening it. Are you exploring options, checking an argument, improving wording, or looking for a potential mistake? Each purpose needs a different kind of response.

For an important choice, write your current view first. Include the constraints and the evidence that would change it. Then use the assistant to challenge that view. This makes it easier to notice whether the conversation adds information or merely changes the tone of your existing opinion.

Afterward, move the result into the place where work happens: a document, a task, a code change, or a calendar entry. Record the decision and the reason for it. A chat transcript is useful context, but it is not always a good system of record.

Some questions are better answered through direct observation or a conversation with the person affected. A model cannot replace information you have not collected.

Useful assistance should leave you better able to act. If a session ends with a clear next step and a reason for taking it, the conversation has done its job.