Make a Digital Twin’s Memory Visible Before It Acts

A personal digital twin should not quietly turn old preferences into permanent instructions. Give people a memory receipt: show which saved preference shaped a draft, let them edit or remove it, and make its source clear.

A small MVP for a work-writing assistant could run like this:

1. The user chooses a recurring task, such as drafting a weekly project update.
2. The assistant shows the few preferences it plans to use—before drafting.
3. The user can correct, skip, or delete any preference.
4. After the draft, they mark what helped and what was wrong.

The first test is not whether the twin sounds personal. It’s whether people can predict and control what it remembers. Try the flow with a small group over one week. If users can’t tell why a preference appeared or remove it easily, stop and redesign the memory controls before adding more personalization.