Events
Past Event
WED@NICO SEMINAR: Caterina Gratton, Northwestern Dept of Psychology "Functional Networks and Hubs in the Human Brain"
Northwestern Institute on Complex Systems (NICO)
12:00 PM
//
Lower Level, Chambers Hall
Details
Speaker:
Caterina Gratton, Assistant Professor, Department of Psychology, Weinberg College of Arts and Sciences, Northwestern University
Title:
Functional Networks and Hubs in the Human Brain
Abstract:
The human brain is organized into large-scale networks, or systems, of interacting brain regions. These interactions can be measured in living humans with functional Magnetic Resonance Imaging (fMRI), by measuring correlations in the patterns of activity between different regions. Increasingly sophisticated techniques enable the mapping of brain networks at unprecedented levels of detail, but many questions still remain. In this presentation, I will tackle three recent studies that we have undertaken to better understand human functional brain networks and their contributions to brain function. In the first study, we examine whether the topology of brain networks – specifically, the presence of connector hub regions – is important for brain function, by examining the consequences of damage to these regions. In the second study, we examine the variability in brain networks within and across subjects at different time-scales. Finally, I will present on very recent work, looking in detail at the characteristics of individual differences in brain networks. Jointly, these studies suggest that network topology has important implications for human brain function, and that measures of network organization are stable features that can be used to measure trait-like variability in brain organization.
Speaker Bio:
Dr. Gratton is currently an Assistant Professor in the Psychology Department at Northwestern University, with a secondary appointment in the Department of Neurology and affiliations with the NUIN and the Cognitive Science Program. Dr. Gratton received her B.S. from the University of Illinois in Psychology and Neuroscience and her Ph.D. in Neuroscience from the University of California, Berkeley, where she worked with Mark D’Esposito and Michael Silver. Afterward, she was a postdoctoral fellow with Steve Petersen at Washington University in St. Louis. Dr. Gratton is interested in large-scale brain networks and how they give rise to complex human behaviors. Her research program seeks to characterize how human brain networks are organized, how they contribute to the myriad goal-directed behaviors that are essential to our daily lives, and how these processes break down with damage and disease. In her work, she employs a variety of methodologies, including functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) to track the spatial and dynamic characteristics of brain activity, and transcranial magnetic stimulation (TMS), recordings from patient populations, and pharmacological manipulations to study perturbations of brain systems.
Live Stream:
Time
Wednesday, April 3, 2019 at 12:00 PM - 1:00 PM
Location
Lower Level, Chambers Hall Map
Contact
Calendar
Northwestern Institute on Complex Systems (NICO)
WED@NICO SEMINAR: Lightning Talks with NU Scholars!
Northwestern Institute on Complex Systems (NICO)
12:00 PM
//
Lower Level, Chambers Hall
Details
May 20th Speakers:
Yulin Yu, Postdoctoral Fellow, Kellogg School of Management
Feihong Xu, PhD Candidate, McCormick School of Engineering
Maalvika Bhat, PhD Student, School of Communication
Rochana Chaturvedi, Postdoctoral Fellow, Kellogg School of Management
NICO Lightning talks are open to Northwestern graduate students, postdoctoral fellows, and visiting scholars! If you are interested in signing up for a future session, please fill out this short survey.
Talk Titles and Abstracts:
Yulin Yu
Postdoctoral Fellow
Kellogg School of Management &
Northwestern Institute on Complex Systems
Human–AI Creative Pathways: How People and Machines Differ in Creative Strategy
Generative AI offers the promise of amplifying creativity by recombining knowledge at a scale far beyond human capacity, yet humans still hold key advantages in flexibility and contextual reasoning. To understand how each achieves novelty, we analyzed more than 5,000 responses to the Divergent Association Task from both humans and AI systems using network-based methods. We find that while individual humans use fewer and simpler conceptual categories than machines, the collective diversity of human ideas is substantially higher. Human creative pathways tend to follow a one-directional but highly unpredictable trajectory, whereas AI systems rely on repetitive, back-and-forth exploration patterns. Finally, both humans and machines show anchoring effects—early ideas shape later responses—but in opposite ways: humans anchor low, while machines anchor high.
Feihong Xu
PhD Candidate
Engineering Sciences & Applied Mathematics
McCormick School of Engineering
A Well-Calibrated Model Similarity Measure for Arbitrary Neural Networks
Deep learning approaches have transformed biological and biomedical image analysis, but model opacity and fragility remain major obstacles to trustworthy use. One barrier is the lack of a well-calibrated measure of similarity across arbitrary neural networks trained with different architectures, checkpoints, random initializations, and training strategies. Existing notions of model similarity span functional and representational domains, often rely on heuristic assumptions, and are susceptible to spurious signals introduced by probing samples, making principled cross-model meta-analysis difficult. Here, we clarify prevailing notions of deep neural network similarity and benchmark their robustness under extensive out-of-distribution perturbations. We then introduce the Ahmad RV coefficient on chain weight matrices (wARV), a theoretically grounded weight-space similarity measure that combines chain-normalized weights with the RV coefficient. wARV is sample-agnostic, symmetric, computationally efficient, and better calibrated than current measures. Across benchmarks varying random initialization, training checkpoint, architecture, and training strategy, wARV more faithfully tracks functional similarity while avoiding confounding effects from probing data. Applying wARV to deep neural network models on both generic and medical image classification tasks, we uncover substantial learning heterogeneity and instability even among models with similar predictive performance.
Maalvika Bhat
PhD Student
Technology and Social Behavior
School of Communication &
McCormick School of Engineering
Scholars See Clickbait as a Greater Threat to Science Than to Their Own Work
As scientific research competes for attention in a media landscape driven by sensationalism, the risks of misrepresentation grow. This study examines whether academics, while widely recognizing clickbait as a threat to science broadly, tend to downplay its relevance to their own work. Surveying 5,603 U.S.-based researchers, we find a consistent perception gap between systemic and personal risk, one that varies by career stage and disciplinary context. Early-career scholars show a pronounced version of this asymmetry: they express heightened concern about clickbait’s harms to science while rating its relevance to their own work as comparatively lower, a pattern that leaves them most exposed at a stage when reputational stakes are highest.
Rochana Chaturvedi
Postdoctoral Fellow
Kellogg School of Management &
Northwestern Institute on Complex Systems
Who Gets the Callback? Generative Artificial Intelligence and Gender Bias
Large language models are increasingly embedded in hiring workflows, raising concerns about their potential to amplify societal biases — yet how these biases manifest within and across occupations, and the role of model 'personality' in shaping these biases, remains unexplored. We introduce a three-part attribution framework applied to 332,044 real-world job ads, measuring gender-based callback bias, associations of skills and traits with gendered stereotypes in LLMs, and the effect of simulated recruiter personas. We find that LLMs systematically favor men, especially in higher-wage roles, with their decisions tracking traditional gendered language cues in job postings. Notably, assigning a low-agreeableness persona reduces model bias, implicating sycophancy as a mechanism reinforcing societal stereotypes; at the same time, controversial personas trigger internal guardrails leading to more cautious and less-biased outputs. These findings highlight how alignment choices in AI-driven hiring systems shape bias, with important implications for fairness and diversity.
Location:
In person: Chambers Hall, 600 Foster Street, Lower Level
Remote option: https://northwestern.zoom.us/j/98031689779
About the Speaker Series:
Wednesdays@NICO is a vibrant weekly seminar series focusing broadly on the topics of complex systems, networks, and artificial intelligence. It brings together attendees ranging from graduate students to senior faculty who span all of the schools across Northwestern, from applied math to sociology to biology and every discipline in-between. Please visit: https://bit.ly/WedatNICO for information on future speakers.
Time
Wednesday, May 20, 2026 at 12:00 PM - 1:00 PM
Location
Lower Level, Chambers Hall Map
Contact
Calendar
Northwestern Institute on Complex Systems (NICO)