Nico Dunajski

My interests are in computational models of foraging and decision-making.

Information-Free Novelty: Modelling Evolutionary Mismatch

SURE summer research project, University of Sheffield, 2026. Supervisor: James A. R. Marshall. Poster presented at the Bernstein Conference 2026.

Why is a social media feed so hard to put down? For most of our history, something new usually meant something worth having, so a pull toward novelty paid off. A feed serves up endless novelty worth very little.

I modelled a forager as a single spiking neuron and mapped the fitness of every strategy as the world slid from ancestral, where novelty tracks value, to modern, where novelty carries no information. Curiosity went unpunished in the ancestral world. As mismatch increased, a ceiling on curiosity appeared and kept falling.

Foraging fitness of every strategy (x: richness prior R̄, y: novelty bonus κ) at five levels of mismatch. As novelty decouples from value, a ceiling on κ appears and keeps falling.
Foraging fitness of every strategy (x: richness prior R̄, y: novelty bonus κ) at five levels of mismatch. As novelty decouples from value, a ceiling on κ appears and keeps falling.

Code: github.com/nicod2587/evolutionary-mismatch-sure2026

Undergraduate Dissertation: Emulating Biophysical Models in Insect Vision

Supervisor: James A. R. Marshall. University of Sheffield, 2026-27. In progress.

Building a statistical emulator of the Juusola lab's 2022 Drosophila photoreceptor model, and creating closed-loop photoreceptor microsaccades in the CompoundRay insect-eye renderer to see if it can improve robot vision.