Reproduce “The Umwelt Representation Hypothesis: Rethinking Universality” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Victoria Bosch, Rowan Sommers, Adrien Doerig, Tim C Kietzmann
This paper proposes the Umwelt Representation Hypothesis (URH) as an alternative to the prevailing "Universality" hypothesis, arguing that representations in both biological and artificial systems are locally optimal and shaped by a system's unique ecological constraints, or "Umwelt." It explains observed representational alignment as an overlap in Umwelts, rather than convergence to a universal optimum, and systematically re-interprets empirical evidence for structured differences.
Key insights
- Representational alignment between systems arises from an overlap in their respective Umwelts, explaining partial rather than universal convergence.
- Systematic differences in representations across diverse systems (species, individuals, ANNs) are not errors but rather adaptive, locally optimal specializations to distinct ecological constraints.
- The utility of ANN model comparison in neuroconnectionism shifts from identifying a single "best" universal model to mapping "clusters of alignment" within the vast space of ecological constraints to understand context-dependent relationships.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.