Rutgers Center for Cognitive Science
Videos footage from RuCCS Colloquium Talks can be found on the RuCCS YouTube Channel.
Details to follow
Abstract 1 (Dr. Jardine): The ability to learn language is a fundamental part of being human, but researchers do not yet understand the mechanisms that underlie this ability. However, linguistic generalizations are not arbitrary, and are in fact guided by computational and structural restrictions related to memory and perception. Importantly, such restrictions can be expressed with computationally simple forms of automata. In this talk, I discuss how these computational restrictions can be leveraged to provide a modular, interpretable, and small-data approach to tackle the problem of language acquisition. As a case study, I discuss a solution to the problem of simultaneously learning the mental representations of words and their morphological and phonological patterning. In this way, we can better understand, qualitatively and computationally, the mechanisms which underlie the human capacities for both constructing mental representations of words and learning the morpho-phonological patterns governing them.
Abstract 2 (Dr. Semenova): Machine learning is increasingly deployed in high-stakes domains like healthcare, lending, and criminal justice. In these fields, it is crucial that human decision-makers can actually understand how an algorithm arrives at its conclusions. However, there is a widespread assumption that we must face a trade-off: to get the highest accuracy, we must rely on complex, uninterpretable "black-box" models. Because finding optimal, perfectly tailored simple models is computationally difficult, researchers often don't even undertake the effort to look for them. This talk asks a practical question: do we really have to sacrifice interpretability for accuracy? I will explore the conditions under which simple, interpretable models—like basic rule lists or decision trees—perform just as well as complex black boxes. By looking at the mechanics of how data is generated and the choices analysts make during the learning process, I will show that highly accurate, easy-to-understand models exist in practice far more often than we might expect.
Abstract: Children have powerful cognitive capacities that enable efficient social learning, but also acquire many beliefs that are false or over-generalized. The present talk proposes that this is partly because, although even young children distinguish between knowledgeable and ignorant sources, they do not always successfully integrate and preserve this epistemic information in memory in ways that appropriately constrain subsequent belief and generalization.
Synopsis: Public discussions of artificial intelligence often swing between enthusiasm and alarm: AI is described as a transformative technology that may expand human capabilities, or as a threat to work, creativity, education, truth, and even humanity itself. But evaluating many of these claims requires us to think carefully about concepts that cognitive scientists have studied for decades: intelligence, learning, reasoning, representation, understanding, memory, and agency.
Abstract 1 (Dr. Feldman): Compositionality---the way people combine concepts to form more complex concepts---has often been identified as a point of difficulty for AI. In this talk I'll present results comparing human and machine performance in on a common platform, the "Game of Hidden Rules" (GOHR). The GOHR is a simple rule-discovery game in which a player (human or AI) tries to classify objects into categories based on an unknown rule that they must infer by trial and error. In a series of experiments, we compared human and AI (RL) performance on "compositional" rule sequences, in which later rules were constructed from combinations of earlier ones. For example, if the learner first solves a rule involving a color match, and then one involving shape, does that make it easier to solve a later rule involving both color and shape? That is, does simpler learning "transfer" to composite learning? The results show that human and AI simple-to-complex transfer are almost completely uncorrelated, suggesting that contemporary AI does not yet effectively reflect the way that humans learn from prior experience.
Abstract 2 (Dr. Schellenberg) : Our capacity for self-reference allows us to introspect on our thoughts and preferences, engage in metacognition, and evaluate our progress on projects while reconfiguring our approach as needed. Despite being central to human intelligence, it has received marginal attention in developing artificial intelligence. There are burgeoning signs of an artificial equivalent of our capacity for self-reference. To illustrate, so-called self-attention mechanisms in Large Language Models, meta-reinforcement learning modules, and second-guessing can each be argued to be examples of artificial self-reference. However, in each case, the capacity is rigid in scope compared to our human capacity, and it developed as a side effect of pursuing other goals. To date, no explicit attention has been directed at artificially replicating this central element of human intelligence. The paper argues that a general capacity for self-reference is key to bringing AI to the next level
.
What Works and What Doesn’t: Things to Know to Land an Internship
Speaker: Elizabeth Langemak, Associate Director of Campus Programs, Verizon
In this talk, we will discuss how insights into the computational underpinnings of psychotic symptoms can serve as a bridge across levels of description and across time as psychosis develops.
Abstract: Brood parasitic birds represent several evolutionary and ontogenetic dilemmas for researchers: how do parasites recognize their own species when raised by a different one and why do hosts accept parasitic eggs and chicks instead of rejecting them.
Abstract: The best future for science on children's learning will be one that embraces both descriptive research and great experiments. We can all look forward to the new hypotheses that emerge from this union, which will position us to create new models of learning, growth, and change.
Abstract: Humans regularly reason about the belief states of others. Where does this capacity originate? Competing hypotheses include biologically evolved endowments, social interaction, and exposure to language. Recent
Please arrive on time as parking is very limited and students arrive promptly at 1:00 PM.
