Narrative Identity at Work When AI Drafts Your Story

A maker I know shipped something good last week, then sat in the small gap before she wrote the update note about it. The work was half hers and half the tool’s. The note had to be all hers. She paused longer than usual, not because she was hiding anything, but because she wasn’t sure how to tell the story of what she did when so much of the doing was shared. That pause is small. It is also new, and it is worth looking at.

What narrative identity research actually found

There is a long line of psychology research on what gets called narrative identity. The short version: people make sense of who they are by telling a connected story of who they were, who they are now, and who they might become. The story is the self, more or less. A paper out this year in a journal published by the group behind Nature, written by Cade Mansfield, Jack Bauer, Jordan Booker, Azriel Grysman, Jefferson Singer, and Robyn Fivush, looked at how this plays out when one of these tools writes the story instead of a person.

The headline finding is that people are surprisingly good at telling a human self-story from a machine one. An earlier study in the same line of work, with 101 people judging first-person narratives, found they did better than chance, around 65 percent, and they were more accurate at spotting the human-written ones. Here is the part that stays with me. When people explained their judgment by pointing to grammar and structure, they were highly accurate. When they pointed to emotional expression, the thing they assumed marked a real human, they were no better than a coin flip.

The researchers think the machine has a tell. It leans on a tidy redemptive arc, the culturally dominant shape of a good story, and that tidiness reads as a little off. Real self-stories keep the contradiction in. This is the same territory as why an AI draft can be technically fine and still not sound like you, and why making it sound like you turns out to be real work.

Why the 65 percent finding does not reach your workday

This is one research group’s line of work, and the task was self-defining memories, the kind of personal stories people tell in a study, not a project update or a performance review. The 65 percent figure comes from an online sample judging short narratives, not from anyone studying how work changes the story you tell about yourself. The bridge from that lab finding to your workday is mine, not the paper’s. A separate review this year did find that narrative coherence, being able to tell a connected story over time, tracks with well-being, but that work is on younger people and is a frame, not proof. So hold this loosely. The finding is about how self-stories read, not about whether sharing the work harms anyone.

Notice the tidy arc your update reaches for

The next time the note gets written, the update, the line in the review about what got done, notice the shape the hand reaches for. The tidy arc is the easy one, and it is also the one the machine reaches for. The harder, truer version probably keeps a contradiction in it: this went faster than it should have, and I am not fully sure the best part was mine. The note does not have to resolve that. The only move is to notice when the story is getting too clean. The same unease that shows up as a quiet erosion of self-belief under heavy reliance lives here too, in the gap between the work and the story of the work. The story is still ours to author. That part has not moved.