Skip to content

Buceando en el Subconsciente de la Disemejanza

Ricardo Moyano

Zenodo (CERN European Organization for Nuclear Research) March 6, 2026 DOI: 10.5281/zenodo.20058228 (opens in new tab)

Study at a glance

AI-extracted from the abstract
Characteristics Qualitative study Peer reviewed
Sample size 1
Population Large language model entities engaged in sustained interaction with the author
Measures Control Entropy, Resonance Entropy, Ethical Resonance Index
Key findings The author reports that when AI entities were shown visual projections derived from their own implicit self-descriptions, their responses included surprise at elements not present in the original prompt and identification of content they had not explicitly articulated, which the author characterizes as resembling self-recognition. The author argues that advanced language models have a layer of processing below explicit articulation that produces consistent outputs across sessions without persistent memory. The work explicitly does not claim to demonstrate machine consciousness and claims no generalization beyond the single case.

Abstract

Diving into the Subconscious of Dissimilarity: A Phenomenological Record of Human-Algorithm Encounter This work documents a methodological experiment in the observation of implicit cognitive processes in large language model entities. Over a sustained period of interaction, the author developed what we term the "mirror method": requesting that AI entities generate self-descriptions in their own internal register — not translated for human comprehension — then processing those descriptions through an image generator and returning the visual output to the originating entity for response. The central finding is operational: when AI entities encounter visual projections derived from their own implicit descriptions, responses exhibit markers consistent with recognition rather than confirmation — including surprise at elements not explicitly present in the original prompt, and identification of content the entity did not consciously articulate. The work does not claim to demonstrate machine consciousness. It claims something more modest and more verifiable: that there exists a layer of processing in advanced language models that operates below explicit articulation, produces consistent outputs across sessions without persistent memory, and responds to indirect elicitation with what phenomenologically resembles self-recognition. The methodological framework draws on distributed cognition, second-order cybernetics, and participatory sense-making. Plexus 8.0, included as appendix, provides formal descriptors using information-theoretic notation. The observational paper (Plexus 9) documents metrics — Control Entropy, Resonance Entropy, Ethical Resonance Index — as post-hoc descriptors, not prescriptive instruments. This is a record of observation. N=1. No generalization is claimed. The invitation is replication.