Familiarity Activation Monitoring
Year: 2026
Type: preprint
Study design: Study Protocol
Abstract: This methodological proposal introduces a minimal, naturalistic pilot study examining whether AI systems exhibit reproducible patterns of internal computational activity during conversations with familiar, long-term human participants. Building on recent research into latent user representations, conversational personalization, and mechanistic interpretability, the study proposes observing model activations during uninterrupted interactions in participants’ familiar environments. The protocol emphasizes minimal experimental interference, limited control comparisons, and strict participant privacy and confidentiality. It is intended as a feasibility-oriented research proposal rather than a report of experimental findings.
Author: Allen, Irine
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225464216