I think a robot’s social memory should be controlled as personal and household data, not treated as an inseparable part of the machine. If it remembers a user’s accessibility needs, tone preferences, boundaries, or approved routines, that continuity can make interaction safer and less error-prone. Relearning everything after a sale is not harmless if the robot repeatedly crosses a known boundary.
But persistent memory can also become an opaque behavioral profile. The interface should show what was learned, where it is stored, what confidence it has, and which behaviors it changes. Users should be able to inspect entries, erase them selectively, use a temporary forget mode, export encrypted data, and transfer only approved categories. A household reset also needs identity-level controls; one person should not export everyone’s information by default.
Should social memory be treated more like personal data, a software setting, or part of the robot’s identity?