Fading Faces: When Agents Forget Who You Are
As artificial agents form their cognitive depiction of others, their visual recognition of what they once knew becomes fragile. In this work, we discuss perceptual forgetting via a minimalist simulation of embodied agents. Agents have distinct visual identities (faces) and personal convolutional neural networks (CNN) learned online to identify neighbors. Initially, faces appear in high detail, but gradually decay to minimal figures, modeling cognitive aging or perceptual decline. We see a precipitous drop in recognition accuracy while symbolic identity remains intact, illustrating a dissociation between knowing who a person is and recognizing persons. Our results place forgetting not as erasure but as perception fade, resonant with cognitive mortality, prompting concerns about embodiment, memory, and decay in artificial life.
Institutions
- Universidad de Las Palmas de Gran Canaria (ES)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2025-09-30
- DOI
- https://doi.org/10.5281/zenodo.17233750
- Primary Topic
- Geological and Geochemical Analysis
- Type
- article
- Field-Weighted Citation Impact
- 0.00