Invited Speakers

How should scientists carve up the cognitive domain to generate good predictions, explanations, and models of cognition? Based on join work with the philosopher and developmental psychologist Mariel Goddu, I argue that cognitive categories should be constructed the same way that biological categories are: in terms of homology. I will make use of a recent account of Character Identity Mechanisms (DiFrisco, Wagner and Love 2020) to make sense of the notion of “cognitive homology.” The consequence of this notion is that brain structures and the organism’s ongoing interactions with the environment turn out to be crucial for individuating cognitive homologies, and thus for individuating cognitive capacities.

Body perceptions are crucial for individuals’ motor, social, and emotional functioning. Importantly, neuroscientific research shows that body perceptions are continually updated through sensorimotor information. This talk will showcase our group’s research on how sensory feedback, particularly sound related to one’s body and actions, can modify body perception, leading to Body Transformation Experiences. I will discuss how these findings contribute to the design of innovative body-centered technologies to address people’s needs and support behavior change. Additionally, beyond such practical applications, these technologies serve as valuable tools for examining multisensory influences on body perception. Our ERC-funded project, BODYinTRANSIT, aims to establish a framework for individualized sensorial manipulation of body perceptions with long-lasting effects in everyday use contexts. The framework stands on four scientific pillars to induce, measure, support, personalize, and preserve body transformations: neuroscience of multisensory body perception; data modeling of the links between body perception, behavior, and emotion; wearable-based embodied multisensory interaction design; and field studies in real-life contexts with diverse user groups. Finally, I will identify challenges and opportunities in this research field.

Human autonomy is a pillar of contemporary ethics and politics, particularly in liberal democracies like the UK, as well as in biomedical ethics. Psychological research robustly shows that a personal sense of autonomy is essential to wellbeing and sustained motivation. In technology design, such felt autonomy also underpins user adoption, engagement, and satisfaction. But today, human autonomy is coming under new threat by AI-driven technologies. Meanwhile, current AI research and policy, questions of safety, fairness, or explainability have received far more attention than how AI may impact autonomy – let alone how to design AI in an autonomy-supporting fashion. In this talk I will describe a vision of a socio-technical future where evidence-based and legitimate design and regulatory guidelines ensure that algorithmic environments safeguard and support human autonomy.

The integration of action and language knowledge and skills is a pivotal element in the realm of human intelligence and stands as one of the most compelling challenges in scientific inquiry. In my presentation I will review the body of evidence and insights collected by attempting to design learning robots capable of understanding and using language and operating in a physical environment. I will particularly highlight the contribution of foundational models and the integration of passive observational learning and active embodied learning modalities. Furthermore, I will examine the merits of learning methods that foster the simultaneous development of diverse competencies indirectly by focusing on the optimization of a single learning objective. 

What can we learn about the structure of individual minds, human or artificial, using large-scale social data, such as the textual or visual data flowing through search engines and social media platforms? In this keynote, I present a diverse range of studies showing that large-scale social data can reveal striking insights into the mind, ranging from the structure of embodied cognition to the psychological biases that drive the formation of stereotypes. I will give special attention to presenting the results of a study we recently published in Nature which demonstrates how combining large-scale image and text data from online sources, analyzed via artificial intelligence, can reveal the latent multimodal structure of gender stereotypes. I will then share ongoing work that builds on these results by revealing the multimodal structure of intersectional stereotypes (e.g., gendered ageism) not only in human minds, but also in the judgments and associations formed by generative AI. Importantly, I will emphasize that big data and artificial intelligence are useful not only for testing existing theories about cognitive structure, but also for discovering and testing new theories. As an example, I will discuss ongoing work that harnesses this suite of algorithmic methodologies to unveil deep connections between the representational structure of gender and the concreteness and abstractness of concepts across domains, using visual and textual data, as well as behavioral outputs from AI. Opportunities for further advancing the integration of computer science, cognitive science, and cultural sociology will be discussed. 

In the early days of classical cognitive science, when the mind was often likened to a computer, cognitive theories developed largely without concern for the ‘hardware’—the brain. Neuroscience was seen as irrelevant to psychological inquiry. However, this began to change in the 1990s with the rise of functional Magnetic Resonance Imaging (fMRI). Neuroscience started to play a significant role in shaping psychological theories, as researchers sought to map specific cognitive functions onto corresponding neural structures. Some proposed that an ideal neurocognitive theory would feature a perfect one-to-one mapping between functions and structures. However, such precise mappings have proven elusive. Instead of neat pairings, we find complex, many-to-many relationships. This raises an important question: how can we reconcile the ideal of one-to-one mappings with the current, entangled status of our knowledge? In this presentation, I will explore four (non-mutually exclusive) approaches that may help us refine our neuro-inspired Cognitive Ontology: (a) We may have chosen the wrong structures or functions, and a one-to-one mapping might be found with the correct selections; (b) The one-to-one mapping might be unattainable, and a probabilistic mapping could be a more realistic goal; (c) It’s possible that the one-to-one mapping exists, but our concepts of ‘functions’ and ‘structures’ need to be redefined; (d) One-to-one mappings may exist, but they might need to be contextualized to specific circumstances.