Kozzy Voudouris

A great opportunity to work at the cutting edge of AI and cognitive science!

Leveling up fun: learning progress, expectations, and success influence enjoyment in video games Scientific Reports - Leveling up fun: learning progress, expectations, and success influence enjoyment in video games

What influences whether people have fun with a task?

Our paper “Leveling up fun: learning progress, expectations and success influence enjoyment in video games” with @thecharleywu.bsky.social and @ericschulz.bsky.social now in Scientific Reports!

rdcu.be/eI069

Paper summary below 1/4

Generating Computational Cognitive Models using Large Language Models Computational cognitive models, which formalize theories of cognition, enable researchers to quantify cognitive processes and arbitrate between competing theories by fitting models to behavioral data....

Happy to announce our paper got accepted to #NeurIPS!
@akjagadish.bsky.social @marvinmathony.bsky.social @ericschulz.bsky.social & Tobi Ludwig

arxiv.org/abs/2502.00879

Excited for this paper to be out, literal years of hard work by Kozzy. Excitingly, my first last author paper!

This work came from joining the Kinds of Intelligence group at Cambridge and being given time by @martahalina.bsky.social to explore and cross disciplines. Hard work but very fun! 🧪 🤖🧠

@matishalin.bsky.social @martahalina.bsky.social

Exploring Major Transitions in the Evolution of Biological Cognition With Artificial Neural Networks Transitional accounts of evolution emphasise a few changes that shape what is evolvable, with dramatic consequences for derived lineages. More recently it has been proposed that cognition might also h...

We find that recurrence confers a significant advantage for learning more complex grammars, but lamination does not.

This work would not have been possible without Matishalin Patel, Colin Klein, Marta Halina, and Andrew Barron. You can check out our preprint here: arxiv.org/abs/2509.13968.

2) The evolution of lamination, in which information is processed by independent sub-components.

We test these networks on artificial grammar learning tasks, for which the complexity of the problem can be precisely stated.

We focus on two transitions

1) The evolution of recurrent systems, in which information feeds back to earlier parts of the network.

In a new preprint, we explore this hypothesis with a computational approach. By systematically evolving neural nets with different structures, we measure whether key structural transitions improve performance on more complex problems, while controlling for the resources available to the network.

The arrangement of neurons is adjusted, using similar resources but opening up the possibility for more complex behaviour. This helps to explain the huge diversity of cognition we see across our planet: from nematodes 🪱 and box jellyfish 🪼 , to bees 🐝 , octopus 🐙 , chimpanzees 🐵 , and us 🙋 .

What if the same is true for the evolution of cognition?

It has recently been argued that the evolution of brains across the animal kingdom can be characterised in terms of just a few major structural changes.

A prominent idea in biology tells us that evolution is not always incremental, but often involves a few important structural changes that open up phylogenetic possibility.

Think: single cells ➡️ multicellular life. Solitary individuals ➡️ eusocial colonies.

That's certainly a hard part too, I can't deny that!

Thanks @dmoralesp.bsky.social, I'm interested to hear your thoughts!

You can check out the published article here (doi.org/10.1007/s105...) and a pre-print here (philsci-archive.pitt.edu/26165/).

This gives rise to an apparent, but ineffable, distinction between hypotheses. Crucially, this apparent distinction does not necessarily imply a distinction at the level of behavioural processes.

For decades, this distinction has been tacitly assumed but impossible to accurately define. I argue that the apparent distinction is actually an artefact of widespread analogical reasoning based on theories in human psychology and classic behaviourist associative learning theory.

I suggest that the widespread use of analogies explains away one of the most controversial debates in this field: whether associative learning and cognition are distinct behavioural processes.

In my latest thriller, published today in Biology and Philosophy, I argue that one useful hypothesis generation strategy in comparative psychology is analogical reasoning.

The hypothesis space is so large and it requires us to eschew our human-centred expectations about behaviour and cognition.

Doing cognitive science on non-human systems like animals or artificial intelligence, brings inherent challenges. One of them is generating plausible alternative explanations for behaviour that can be tested empirically.

Excited to say our paper got accepted to ICML! We added new findings including this: models fine-tuned on a visual counterfactual reasoning task do not generalize to the underlying factual physical reasoning task, even with test images matched to the fine-tuning data set.

Many thanks to my coauthors @martahalina.bsky.social , @lucycheke.bsky.social, and Ben Farrar for working with me on this project.

You can check out the open-access paper here: psycnet.apa.org/fulltext/202...

We found that comparative psychologists have incredibly diverse attitudes to these methodological challenges. This suggests that there is still a lot of work for both psychologists and philosophers of science to do to clarify fundamental debates at the heart of comparative psychology research.

First, the apparent distinction between associative learning and cognition, and second, the question of whether some explanations of animal behaviour are simpler than others (often discussed with reference to Morgan's Canon).

We surveyed 220 practicing comparative psychologists, asking them questions about two key methodological challenges that define their science...

I am very pleased to announce that my paper "Morgan's Canon and the Associative-Cognitive Distinction Today: A Survey of Practitioners" was published this week in the Journal of Comparative Psychology.

Bringing comparative cognition approaches to AI systems

Comment by Konstantinos Voudouris (@kozzyvoudouris.bsky.social), Lucy Cheke (@lucycheke.bsky.social) & Eric Schulz (@ericschulz.bsky.social)

go.nature.com/43djJlv

In order to understand cognition, we often recruit analogies as building blocks of theories to aid us in this quest. One such attempt, originating in folklore and alchemy, is the homunculus: a miniature human who resides in the skull and performs cognition. Perhaps surprisingly, this appears indistinguishable from the implicit proposal of many neurocognitive theories, including that of the 'cognitive map,' which proposes a representational substrate for episodic memories and navigational capacities. In such 'small cakes' cases, neurocognitive representations are assumed to be meaningful and about the world, though it is wholly unclear who is reading them, how they are interpreted, and how they come to mean what they do. We analyze the 'small cakes' problem in neurocognitive theories (including, but not limited to, the cognitive map) and find that such an approach a) causes infinite regress in the explanatory chain, requiring a human-in-the-loop to resolve, and b) results in a computationally inert account of representation, providing neither a function nor a mechanism. We caution against a 'small cakes' theoretical practice across computational cognitive modelling, neuroscience, and artificial intelligence, wherein the scientist inserts their (or other humans') cognition into models because otherwise the models neither perform as advertised, nor mean what they are purported to, without said 'cake insertion.' We argue that the solution is to tease apart explanandum and explanans for a given scientific investigation, with an eye towards avoiding van Rooij's (formal) or Ryle's (informal) infinite regresses.

Figure 1 in https://philsci-archive.pitt.edu/24834/

Box 1 in https://philsci-archive.pitt.edu/24834/

Box 2 in https://philsci-archive.pitt.edu/24834/

Tired but happy to say this is out w @andreaeyleen.bsky.social: Are Neurocognitive Representations 'Small Cakes'? philsci-archive.pitt.edu/24834/

We analyse cog neuro theories showing how vicious regress, e.g. the homunculus fallacy, is (sadly) alive and well — and importantly how to avoid it. 1/