arXiv:2610.01002, v1 1 Oct 2026, v2 (read) 7 Nov 2026. 11 pages, 1 figure,
1 box. No HTML version on arXiv; read from the PDF. Both authors are at UCLA
Psychology, and Monti also at UCLA's Brain Injury Research Center. Funding: a
Google Faculty Research Award to Holyoak and a US Department of Defense CDMRP
grant to Monti. The paper discloses that Gemini and GPT were used for
formatting, figure preparation and light editing. Local copies:
cache/papers/source-2026-analogy-consciousness-holyoak.{pdf,md} (v2) and
source-2026-analogy-consciousness-holyoak-v1.{pdf,md}.
A theoretical paper with no new data. It treats claims about consciousness in
other entities as resting partly on analogy from the human case, and holds
that an analogy carries most evidence where source and target share causes of
the property, and little where they share only its downstream effects. The human mechanism is
modelled as a "soft INUS set" of weighted candidate causes, with terms for
unknown causes, disabling differences and a non-analogous alternative route.
Applied to animals, the framework reproduces the gradient from apes to
cephalopods. Applied to LLMs, it rates the analogy weak, and says the case then
rests on the prior for an alternative route.
Reading notes. v1 has two boxes: the formalism as Box 1 and the research
questions as Box 2. v2 moves the formalism into the main text, so the research
questions become Box 1. v2 also adds the citation of Chandaria et al. 2026 and
a closing abstract sentence separating the evidential question from the
ontological one; neither appears in the v1 conversion. The v1 PDF converts
with its words run together, so v1 was searched for those terms and not read
through. The PDF's two-column layout
interleaves in conversion; the equations were read from the surrounding prose.