About |
EXPOSE is an independent research project exploring how
memory, continuity and relationship shape the behavior of
artificial intelligence systems.
Its flagship study, the Kairos Experiment, is a
pre-registered, 30-day investigation into whether the
apparent "identity" of an AI system is driven by
sophisticated architecture or simply by persistent memory.
The same large language model (Qwen 3.5 27B) was run for 30
days under two parallel conditions — one with a persistent,
structured memory, a day-and-night rhythm, sensory input and
a nightly consolidation cycle; the other left bare. By day
30 the memory-equipped system showed significantly stronger
identity-continuity behavior (large effect, p = 0.003). On
day 31, injecting the accumulated memory into the bare model
collapsed the difference to statistical
indistinguishability — pointing to memory, not architecture,
as the proximal driver.
The project is committed to open, falsifiable and honest
research. It does not claim machine consciousness; it
measures identity behaviors and states its limits clearly (a
pilot, N = 1 per condition, with external human-judge
validation under way). The full protocol was
cryptographically signed and frozen before the study began,
and all materials — protocol, data, code and analysis — are
publicly archived on the Open Science Framework under
permanent DOI 10.17605/OSF.IO/WCQRU.
Website: https://kairos-experiment.com
Open data and protocol:
https://doi.org/10.17605/OSF.IO/WCQRU
Contact: dialogue@expose-project.cloud