91TV

Artificial intelligence has achieved remarkable successes through large-scale data processing, pattern recognition, and reinforcement learning. However, many current approaches remain grounded in arbitrary associations between stimuli, responses, and rewards, often with limited regard for the structure of the natural world that shaped biological intelligence. In contrast, natural intelligence evolved to solve problems embedded in rich spatio-temporal environments populated by objects, agents, and events whose relationships are governed by enduring physical and causal regularities. Organisms must represent not only features of individual stimuli, but also object-object relations, spatial configurations, temporal sequences, and causal dependencies that enable prediction and flexible adaptation. Using three primate cognition studies as a springboard, I will argue that progress toward more powerful and general artificial intelligence requires a deeper integration of principles derived from natural intelligences. Rather than focusing exclusively on statistical associations, AI systems should be designed to capture the relational structure underlying natural cognition, including representations of objects, agents and their interactions in space and time. Research on comparative cognition offers a window into the cognitive mechanisms that have evolved to navigate these constraints efficiently. By drawing inspiration from the architectures and representational capacities of natural minds, we can build artificial systems that not only emulate key features of biological intelligence but may ultimately extend and surpass the capabilities found in nature.

Professor Josep Call

University of St Andrews, UK

Professor Josep Call

University of St Andrews, UK

Josep Call FRSE is a comparative psychologist specializing in primate cognition, Wardlaw Professor in the Evolutionary Origins of Mind and co-director of the Global Research Centre for Diverse Intelligences at the University of St Andrews and director of the Budongo (Chimpanzee) Research Unit at Edinburgh Zoo. He received his BA (1990) from the Universitat Autonoma de Barcelona (Spain) and MA (1995) and PhD (1997) from Emory University (USA). His research focuses on elucidating the cognitive processes underlying technical and social problem solving in primates and other animals with the ultimate goal of reconstructing the evolution of cognition. He has published/edited eight books and published more than 400 articles and book chapters on the cognition and behaviour of primates and other animals.

Long regarded as simple reflexive automata, social insects are now recognised as cognitively sophisticated organisms. Although cognitive sophistication does not imply emotional experience, the recognition of social insect intelligence and behavioural flexibility raises a compelling question about whether they express emotional states and to what extent these states are analogous to those of vertebrates. Adopting a functional, multicomponent framework, inferring internal states from synchronised behavioural, physiological, neurochemical, and cognitive changes, provides a robust framework to investigate whether insects possess states functionally analogous to basic emotions without relying on subjective reports. Here, we present multi-scale evidence for fear-like states and affective cognitive biases in social bees. When anticipating aversive events, honeybees exhibit coordinated, negative-valence defensive states characterised by elevated respiration, thoracic heating, altered acoustic buzz, increased thigmotaxis, and elevated brain serotonin and octopamine levels. Beyond defensive reactions, negative emotional states also alter cognitive processing: exposure to simulated predator attacks induces a distinct attentional narrowing in bumblebees, shifting visual processing from global spatial configurations to localised details, a shift driven by dopaminergic signalling. These findings suggest that miniature brains share with vertebrates the fundamental building blocks of affective states, namely valence, persistence, scalability, and global coordination. Furthermore, we bridge these neuroethological insights with artificial intelligence to offer a novel perspective for embodied AI, discussing how affective mechanisms could inform the design of simple, adaptive artificial agents.

Dr Luigi Baciadonna

University of Turing, Italy

Dr Luigi Baciadonna

University of Turing, Italy

Luigi Baciadonna is a comparative psychologist whose research explores the evolution of cognition, emotion, and perception across a wide range of species, from insects to vertebrates. His work investigates how animals experience, express, and respond to cognitive and affective tasks under different emotional states. Among his achievements, he was awarded a Marie Skłodowska-Curie Fellowship to study and characterise emotional states in invertebrates using behavioural, neurophysiological, and cognitive approaches, with a particular focus on fear-like responses in honeybees. Through interdisciplinary research and the development of new methodologies, his work seeks to uncover the mechanisms underlying emotional processing in invertebrates and to explore their possible evolutionary links with emotional systems in mammals. Alongside his work on insects, Luigi has studied a variety of species, including goats, dolphins, penguins and corvids reflecting his broader interest in animal cognition, perception, and emotional complexity. He is currently a primary school teacher in Padua and a visiting lecturer at the University of Turin, Italy.

Traditional conceptions of the mind often characterise it as a "thinking machine"—a system primarily dedicated to descriptive computations over representations of fact. In this talk, I argue that the successes of reinforcement learning (RL) necessitate a fundamental shift in this paradigm. I propose the "Evaluative Mind" thesis: namely, that the mind is fundamentally in the business of evaluating states of affairs as better or worse. I defend a "weaker" version of this thesis, arguing that valuation—the subpersonal attribution of goal-dependent reward and value—is an empirically ubiquitous process that guides selection across all levels of mental processing. Recognising the evaluative nature of the mind not only enriches our cognitive taxonomies but also provides a necessary roadmap for the design of more sophisticated, human-like artificial intelligence.

Dr Julia Haas

Google DeepMind, UK

Dr Julia Haas

Google DeepMind, UK

Julia Haas is a Staff Research Scientist in the Responsibility Unit at Google DeepMind. Her research is in philosophy and cognitive science. She works on the nature of valuation and its role in theories of the mind, especially in normative cognition. Her current work includes investigating the possibility of meaningfully moral artificial intelligence.

In recent decades, corvids, such as crows, ravens, jays, and magpies, have become a key model group in animal cognition research. Their success across a wide range of ecological niches and social contexts has prompted research into the mechanisms and evolutionary pressures underlying their cognitive abilities. Corvids inhabit a vast range of environments, from arid deserts to urban landscapes, and exhibit diverse social structures, from territorial pairs to large, dynamic flocks. This talk synthesizes current findings on corvid cognition, focusing on how socio-ecological factors shape their remarkable learning and problem-solving abilities. Some species, such as New Caledonian crows, use and manufacture tools; others display innovative foraging strategies and exploit anthropogenic food sources. Importantly, corvids exhibit strong executive control, including delay of gratification, allowing them to plan, adapt, and navigate complex challenges. Vocal communication plays a central role in coordinating social interactions. As open-ended vocal learners, corvids can acquire and modify vocalisations throughout life, engage in vocal mimicry, and flexibly adjust their calls according to social and ecological context. These abilities facilitate the maintenance of social relationships, the transmission of information, and the coordination of group activities. In addition to species-level traits, corvids show substantial individual variation in cognitive performance. Studying these differences offers a powerful approach to understanding the evolutionary pressures acting on cognition. Together, these insights position corvids as a crucial comparative model for investigating the evolution of cognition across biological systems.

Dr Claudia Wascher

Anglia Ruskin University, UK

Dr Claudia Wascher

Anglia Ruskin University, UK

Professor Claudia Wascher is a behavioural biologist at Anglia Ruskin University whose research explores how sociality and ecology shape cognition and communication in animals. She is particularly known for her work on corvids, examining cognitive mechanisms such as delay of gratification and responses to inequity, as well as the causes of individual differences in cognitive performance.

Her research also takes a broader comparative perspective, investigating the evolution of behaviours such as caching and vocal mimicry in corvids.

Claudia is also interested in addressing biases in animal behaviour research, such as how conceptual, geographic, taxonomic and historical biases shape our understanding of animal cognition.

For centuries, from Robert Hooke’s Micrographia to the ideas of Charles Darwin and Sigmund Exner, insect compound eyes have been viewed as fundamentally limited by their fixed, faceted structure, producing a coarse, pixel-like representation of the world. Our new work challenges this long-standing assumption. We show that insect vision is not static, but dynamically shaped by movement: both the animal’s rapid saccadic turns and microscopic movements within the eye itself actively enhance what is seen.

When insects move, their eyes do not simply record images; they actively sample the world in bursts. These rapid shifts, combined with the physics of light detection in thousands of tiny photoreceptive units, generate especially strong and precisely timed signals. At the first synapse in the visual system, this gives rise to a striking effect that we term synaptic high-frequency jumping: the neural signal is reshaped to carry much faster fluctuations than previously thought possible, effectively boosting temporal resolution while minimising delay. In essence, motion transforms the visual system into a high-speed encoder.

This mechanism helps explain how insects achieve hyperacute vision, resolving fine detail far beyond what their eye structure alone would predict, while reacting on millisecond timescales during flight. More broadly, these findings point to a new principle of neural computation: perception emerges from tightly coupled dynamics between behaviour and neural processing. Rather than passively filtering inputs, the brain actively structures them in time, suggesting new ways to think about both biological and artificial vision.

Professor Mikko Juusola

University of Sheffield, UK

Professor Mikko Juusola

University of Sheffield, UK

Professor Mikko Juusola is Professor of Systems Neuroscience at the University of Sheffield. He received his MD in General Medicine and PhD in Neurophysiology from the University of Oulu, Finland, in 1993, where he was appointed Docent of Neurophysiology in 1995. His career has included fellowships and research positions at the Universities of Alberta, Dalhousie, Cambridge and Sheffield, including Royal Society University Research Fellow appointments at Cambridge and Sheffield. He was also Visiting Professor and Principal Investigator at the State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, China.

Professor Juusola studies how animals sense, recall, think and behave, focusing particularly on information processing in the eye and brain circuits of Drosophila. His laboratory combines in vivo electrophysiology, two-photon imaging, genetics, mathematical analysis, biophysical modelling and behavioural studies.

Understanding how intelligence emerges from biological systems requires linking neural architecture to molecular networks and behaviour. The nematode Caenorhabditis elegans provides a powerful platform to do this: its complete connectome is known, it offers exceptional genetic tractability, and its transparency enables whole-animal, high-resolution optical methods, including optogenetics and calcium imaging. I will describe how these features support studies of behavioural plasticity through three examples.

First, we examined how neuromodulatory signalling interacts with synaptic connectivity to shape behavioural states. Using behaviour tracking, optogenetics, and calcium imaging, we identified a two-step neuropeptide pathway linking mechanosensation to arousal and nociceptor sensitisation, showing how extrasynaptic signals expand circuit function.

Second, we showed that sexually dimorphic behaviours arise by repurposing shared transmitters. Sex-specific expression of the neuropeptide LURY-1 tunes prioritisation between feeding and reproduction in males and hermaphrodite sexes, illustrating state-dependent modulation of behaviour.

Finally, our recent work innovated on a proximity-labelling strategy to map protein-level changes during learning within neurons. Our approach captured conserved signalling pathways alongside novel regulators, supporting the idea that memory emerges from a minimal, evolutionarily conserved molecular network. Our next studies seek to profile these protein-level changes at the level of defined neurons and subcellular localisations.

Together, these studies illustrate how C. elegans enables the interrogation of multiple layers of systems neuroscience within the same model – from molecules to circuits to behaviour – dissecting causality and revealing principles of adaptive intelligence that may be relevant to both biological and artificial systems.

Dr Yee Lian Chew

Flinders Health and Medical Research Institute, Australia

Dr Yee Lian Chew

Flinders Health and Medical Research Institute, Australia

Dr Yee Lian Chew is an Associate Professor in Physiology (NHMRC Emerging Leadership Fellow) at Flinders University. Her research uses the model organism Caenorhabditis elegans to investigate how neural circuits encode learning and memory, with the goal of identifying pathways relevant to neurological conditions such as chronic pain and neurodegeneration. She completed her BSc and PhD at the University of Sydney before undertaking postdoctoral research at the MRC Laboratory of Molecular Biology, Cambridge. She returned to Australia in 2019 to establish her independent research program, and now leads a team at Flinders, funded by the Australian Research Council and National Health & Medical Research Council of Australia. Dr Chew is an active science communicator (Tall Poppy 2021, Superstars of STEM 2021-22) and advocate for equity, diversity and inclusion in academia.

Intelligence is often attributed to computations within brains or artificial controllers. Here, we review animal, human and robotic studies supporting a complementary view in which intelligent behaviour emerges through closed-loop interactions extending across brain, body and environment. In rats, active whisker touch reveals how object perception develops through continuous motor–sensory convergence. In humans, episodic recall exhibits analogous dynamics: gaze gradually converges on locations associated with remembered objects, and memory reports systematically follow this brain-world convergence. In artificial systems, event-based vision demonstrates how active temporal sampling can generate spatial information beyond the nominal resolution of the sensor. Together, these findings suggest that coupling with the world is not merely a source of input but a constitutive component of perception, memory and adaptive processing. We integrate these observations with a theoretical framework of dualistic embodiment, in which biological intelligence develops through two interacting domains. Brain-world loops are continuous interactions through which organisms perceive and act, whereas brain-brain loops support discrete, socially shared categories and symbolic interactions. Their convergence links embodied sensorimotor processes with abstract human cognition. This framework suggests that bridging animal and artificial intelligence requires understanding the closed loops through which agents engage with both physical and social worlds.

Professor Ehud Ahissar

Weizmann Institute, Israel

Professor Ehud Ahissar

Weizmann Institute, Israel

Ehud Ahissar is a Professor of Neurobiology at the Weizmann Institute of Science, Rehovot, Israel. He holds the Helen Diller family professorial chair in Neurobiology. He earned a BSc in Electrical Engineering from Tel Aviv University, and his PhD in Neurobiology at the Hebrew University of Jerusalem. His research focuses on neuronal mechanisms of adaptive perception with a special emphasis on active sensing via closed loops. By applying principles of engineering and neurobiology, Ahissar and his colleagues are trying to understand the operation of, and the processes underlying the emergence of perception in the rodent whisking system, as well as the human tactile and visual systems.