About

I am a cognitive scientist in the Department of Communication at UCLA. My research involves quantifying the dynamics of cognition, with a focus on human communication. The work I've been involved in has touched upon language's many levels of complexity: from how it evolved, to how we carry out brief conversations. I am also interested in a wide range of other topics, such as the interface between language and action, cognitive dynamics, and theoretical issues in cognitive science.

Research topics

• Computer Science
• Data Mining
• Machine Learning
• Psychology
• Artificial Intelligence
• Physics
• Programming language
• Social psychology
• Communication
• Engineering
• Geology
• Biology
• Cognitive psychology
• Human–computer interaction
• Neuroscience
• Computational science
• Theoretical computer science

Selected publications

• A Bayesian Dynamical System Model of Joint Action and Interpersonal Coordination

  ArXiv.org · 2025-09-10

  preprintOpen access

  Successful teamwork depends on interpersonal dynamics, the ways in which individuals coordinate, influence, and adapt to one another over time. Existing measures of interpersonal dynamics, such as CRQA, correlation, Granger causality, and transfer entropy, typically capture only a single dimension: either the synchrony/coordination or the direction of influence between individuals. What is missing is a psychologically meaningful representation that unifies these dimensions and varies systematically with behavior. We propose the "context matrix" as one such representation. The context matrix, modeled within a linear dynamical system, has psychologically interpretable entries specifying how much each individual's current behavior is attributable to their own versus every other group member's past behaviors. Critically, these entries can be distilled into summary features that represent synchrony and directional influence. Evidence for the context matrix as psychologically meaningful is provided in two steps. First, we develop a sequential Bayesian model that infers context matrices from timeseries data and show that it accurately recovers them in noisy simulations. Second, applying the model to human eyetracking data, we demonstrate that summary features of the inferred context matrices capture expected task-based differences in interpersonal dynamics (or lack thereof), predict task accuracy in psychologically reasonable ways, and show some correspondence with existing measures (CRQA and Granger causality). We conclude by situating the context matrix within a broader agenda for modeling interpersonal dynamics in joint action.

  PublisherOA PDFDOI

• Mapping the learning curves of deep learning networks

  PLoS Computational Biology · 2025-02-10 · 3 citations

  articleOpen accessSenior authorCorresponding

  There is an important challenge in systematically interpreting the internal representations of deep neural networks (DNNs). Existing techniques are often less effective for non-tabular tasks, or they primarily focus on qualitative, ad-hoc interpretations of models. In response, this study introduces a cognitive science-inspired, multi-dimensional quantification and visualization approach that captures two temporal dimensions of model learning: the "information-processing trajectory" and the "developmental trajectory." The former represents the influence of incoming signals on an agent's decision-making, while the latter conceptualizes the gradual improvement in an agent's performance throughout its lifespan. Tracking the learning curves of DNNs enables researchers to explicitly identify the model appropriateness of a given task, examine the properties of the underlying input signals, and assess the model's alignment (or lack thereof) with human learning experiences. To illustrate this method, we conducted 750 runs of simulations on two temporal tasks: gesture detection and sentence classification, showcasing its applicability across different types of deep learning tasks. Using four descriptive metrics to quantify the mapped learning curves-start, end - start, max, tmax-, we identified significant differences in learning patterns based on data sources and class distinctions (all p's < .0001), the prominent role of spatial semantics in gesture learning, and larger information gains in language learning. We highlight three key insights gained from mapping learning curves: non-monotonic progress, pairwise comparisons, and domain distinctions. We reflect on the theoretical implications of this method for cognitive processing, language models and representations from multiple modalities.

  PublisherDOI

• Task goals constrain the alignment in eye-movements and speech during interpersonal coordination

  2025-03-25

  preprintOpen access

  Collaborative task performance is assumed to benefit from interpersonal coordination of behaviors between interacting individuals. Prominent views of language use and social behavior, including the Interactive Alignment Model (IAM; Pickering &amp;amp; Garrod, 2004), endorse this idea by building on tasks that require partners to monitor each other’s perspective (e.g., route planning) and positing that behavioral alignment enables task partners to converge conceptually. However, the role of alignment in tasks requiring complementarity (e.g., a “divide and conquer” strategy during joint visual search) has yet to be explored. We examine this question directly by manipulating task goals (route planning vs. visual search) as forty dyads work with ten trials involving subway maps while their eye movements and speech were co-registered. In five trials, dyads planned a route from an origin to a destination (route planning); in another five trials, they searched for landmarks sharing some feature (visual search). We used Cross Recurrence Quantification Analysis (CRQA) to examine the temporal relationships between partners' eye fixations and word sequences, generating measures that reveal both similarity and other dynamic relationships. Dyads exhibited more gaze alignment in route planning than visual search across a range of CRQA metrics. When examining the temporal evolution of gaze alignment, we found it to vary across the trial substantially, and its increase influenced accuracy differently over time across the two tasks. Specifically, in visual search, higher increases in alignment at the end of the trial were associated with accurate performance. When we turned to speech data, we found that dyads exhibited longer and more entropic word sequences in route planning but had lower overall word recurrence in that task. This finding suggests that the two modalities organize in a compensatory fashion to optimize distinct task goals. We suggest that these results support a theoretical framework that is more general than IAM yet has interactive alignment as an emergent consequence of how participants adapt to tasks. This framework emphasizes the dynamic adaptation of coordination strategies based on task demands. Altogether, task goals constrain how people coordinate their behavior and provide insights into how collaborating partners distribute their distinctive multimodal behavior.

  PublisherOA PDFDOI

• Teamwork as Linear Interpersonal Dynamics

  2025-09-12

  articleOpen access

  Successful teamwork depends on interpersonal dynamics, the ways in which individuals coordinate, influence, and adapt to one another over time. Existing measures of interpersonal dynamics, such as CRQA, correlation, Granger causality, and transfer entropy, typically capture only a single dimension: either the synchrony/coordination or the direction of influence between individuals. What is missing is a psychologically meaningful representation that unifies these dimensions and varies systematically with behavior. We propose the context matrix as one such representation. The context matrix is the transition matrix in a linear dynamical system, with entries specifying how much each individual’s current behavior is attributable to their own versus every other group member’s past behaviors. Its values can be distilled into psychologically interpretable summary features of synchrony and directional influence. Evidence for the context matrix as psychologically meaningful is provided in two steps. First, we develop a sequential Bayesian model that infers context matrices from timeseries data and show that it accurately recovers them in noisy simulations. Second, applying the model to human eyetracking data, we show that summary features of the inferred context matrices capture expected task-based differences in interpersonal dynamics (or lack thereof), predict task accuracy in psychologically reasonable ways, and show some correspondence with existing measures (CRQA and Granger causality). We conclude by situating the context matrix within a broader agenda for modeling interpersonal dynamics.

  PublisherDOI

• Teamwork as Linear Interpersonal Dynamics

  2025-09-11

  preprintOpen access

  Successful teamwork depends on interpersonal dynamics, the ways in which individuals coordinate, influence, and adapt to one another over time. Existing measures of interpersonal dynamics, such as CRQA, correlation, Granger causality, and transfer entropy, typically capture only a single dimension: either the synchrony/coordination or the direction of influence between individuals. What is missing is a psychologically meaningful representation that unifies these dimensions and varies systematically with behavior. We propose the context matrix as one such representation. The context matrix is the transition matrix in a linear dynamical system, with entries specifying how much each individual’s current behavior is attributable to their own versus every other group member’s past behaviors. Its values can be distilled into psychologically interpretable summary features of synchrony and directional influence. Evidence for the context matrix as psychologically meaningful is provided in two steps. First, we develop a sequential Bayesian model that infers context matrices from timeseries data and show that it accurately recovers them in noisy simulations. Second, applying the model to human eyetracking data, we show that summary features of the inferred context matrices capture expected task-based differences in interpersonal dynamics (or lack thereof), predict task accuracy in psychologically reasonable ways, and show some correspondence with existing measures (CRQA and Granger causality). We conclude by situating the context matrix within a broader agenda for modeling interpersonal dynamics.

  PublisherDOI

• Reconsidering intrapersonal communication through an interdisciplinary lens

  Frontiers in Psychology · 2025-08-11

  articleOpen accessSenior author

  Intrapersonal communication is a classification of phenomena such as self-talk and imagined interactions in which communicative messages are contained within a single individual and internal systems comprise both sender and receiver roles. Historically, the construct has met criticism with objections rooted in the notion that intrapersonal communication is simply a form of social cognition, but not all self-communicative behaviors have clear or empirically defined social cognitive connections. Self-directed behaviors, from inner speech to mindfulness, permit individuals to shape and interpret their experiences. Empirical examination of these phenomena would benefit from a unified framework. Relevant work is discussed supporting the perspective that intrapersonal communication constitutes a valuable interdisciplinary classification, including early research, functional and developmental approaches, and current connected methodologies and their limitations. A theoretical model is proposed that can guide understanding of the boundaries of intrapersonal communication by characterizing sender and receiver roles in the intrapersonal interaction based on active and inactive status.

  PublisherOA PDFDOI

• Task goals constrain the alignment in eye-movements and speech during interpersonal coordination

  2025-08-15

  articleOpen access

  Collaborative task performance is assumed to benefit from interpersonal coordination of behaviors between interacting individuals. Prominent views of language use and social behavior, including the Interactive Alignment Model (IAM; Pickering &amp;amp; Garrod, 2004), endorse this idea by building on tasks that require partners to monitor each other’s perspective (e.g., route planning) and positing that behavioral alignment enables task partners to converge conceptually. However, the role of alignment in tasks requiring complementarity (e.g., a “divide and conquer” strategy during joint visual search) has yet to be explored. We examine this question directly by manipulating task goals (route planning vs. visual search) as forty dyads work with ten trials involving subway maps while their eye movements and speech were co-registered. In five trials, dyads planned a route from an origin to a destination (route planning); in another five trials, they searched for landmarks sharing some feature (visual search). We used Cross Recurrence Quantification Analysis (CRQA) to examine the temporal relationships between partners' eye fixations and word sequences, generating measures that reveal both similarity and other dynamic relationships. Dyads exhibited more gaze alignment in route planning than visual search across a range of CRQA metrics. When examining the temporal evolution of gaze alignment, we found it to vary across the trial substantially, and its increase influenced accuracy differently over time across the two tasks. Specifically, in visual search, higher increases in alignment at the end of the trial were associated with accurate performance. When we turned to speech data, we found that dyads exhibited longer and more entropic word sequences in route planning but had lower overall word recurrence in that task. This finding suggests that the two modalities organize in a compensatory fashion to optimize distinct task goals. We suggest that these results support a theoretical framework that is more general than IAM yet has interactive alignment as an emergent consequence of how participants adapt to tasks. This framework emphasizes the dynamic adaptation of coordination strategies based on task demands. Altogether, task goals constrain how people coordinate their behavior and provide insights into how collaborating partners distribute their distinctive multimodal behavior.

  PublisherOA PDFDOI

• A Bayesian Dynamical System Model of Joint Action and Interpersonal Coordination

  2025-09-26

  articleOpen access

  Successful teamwork depends on interpersonal dynamics, the ways in which individuals coordinate, influence, and adapt to one another over time. Existing measures of interpersonal dynamics, such as CRQA, correlation, Granger causality, and transfer entropy, typically capture only a single dimension: either the synchrony/coordination or the direction of influence between individuals. What is missing is a psychologically meaningful representation that unifies these dimensions and varies systematically with behavior. We propose the context matrix as one such representation. The context matrix, modeled within a linear dynamical system, has psychologically interpretable entries specifying how much each individual’s current behavior is attributable to their own versus every other group member’s past behaviors. Critically, these entries can be distilled into summary features that represent synchrony and directional influence. Evidence for the context matrix as psychologically meaningful is provided in two steps. First, we develop a sequential Bayesian model that infers context matrices from timeseries data and show that it accurately recovers them in noisy simulations. Second, applying the model to human eyetracking data, we show that summary features of the inferred context matrices capture expected task-based differences in interpersonal dynamics (or lack thereof), predict task accuracy in psychologically reasonable ways, and show some correspondence with existing measures (CRQA and Granger causality). We conclude by situating the context matrix within a broader agenda for modeling interpersonal dynamics.

  PublisherOA PDFDOI

• Antisemitic and Islamophobic hate speech precedes a decrease in lexico-semantic diversity in comment threads online

  Humanities and Social Sciences Communications · 2025-04-02

  articleOpen accessSenior author

  Abstract Studies of Antisemitic and Islamophobic hate speech (AHS and IHS) demonstrate that they severely impact the psychological and social well-being of Jewish and Muslim communities. However, work to date has not adequately addressed the effect that the introduction of AHS/IHS has on subsequent expression in groups that tolerate hate speech online. We thus do not know how influential AHS and IHS are. The current study attempts to address this gap in the literature directly by providing an information-theoretic account of what happens when social media users on the website Reddit vary the intensity of Islamophobic and/or Antisemitic sentiment in their comments. We find that the more overtly Antisemitic or Islamophobic the comment, the easier it is to recover the ideas expressed in that comment from subsequent comments. In other words, comments that rank high for AHS or IHS appear to impose a strong bottleneck on the lexico-semantic diversity of subsequent conversations. This effect was strengthened after the onset of the conflict in Gaza and Israel on October 7th, 2023. Our results offer a step toward investigating how information transmission is altered due to the effects of particular kinds of HS, and have direct implications for organizations with a vested interest in content moderation.

  PublisherOA PDFDOI

• Bridging Cognition and Communication: Identifying Opportunities for Cross-Disciplinary Connections Using Scientometric Techniques

  2025-06-02

  preprintOpen access1st authorCorresponding

  Casual intuition suggests that human communication and cognition would have robust connections across their relevant phenomena: Communication involves minds, and minds are often driven to communicate. Nevertheless, these domains are now represented by very distinct domains of scholarship. This paper explores ways to bridge them. We propose a bidirectional theoretical perspective on the relationship between communication and cognitive science, and support this perspective with an analysis of over 15,000 titles and abstracts in published work in these two disciplines. Using semantic analysis inspired by scientometrics, we argue for a rapprochement between these fields. Specifically, we articulate numerous promising avenues of overlap and mutual influence that may hold between them, such as connecting sociocultural and media issues in communication with core linguistic and mechanistic processes in cognitive science. We conclude with a schema for future theoretical and empirical work aligning communicative and cognitive processes.

  PublisherOA PDFDOI

Recent grants

• Collaborative Research: Dynamics of Interpersonal Coordination and Embodied Communication

  NSF · $50k · 2009–2012

• Collaborative Research: Action Dynamics as an Index of Learning and Generalization

  NSF · $20k · 2011–2012

• Collaborative Research: Dynamics of Interpersonal Coordination and Embodied Communication

  NSF · $40k · 2011–2013

• Collaborative Research: Action Dynamics as an Index of Learning and Generalization

  NSF · $206k · 2007–2012

Frequent coauthors

• Jonathan B. Freeman

  Columbia University

  31 shared

• Nicholas D. Duran

  Arizona State University

  29 shared

• Thomas A. Farmer

  29 shared

• Sarah E. Anderson

  The Ohio State University

  26 shared

• Daniel C. Richardson

  University College London

  22 shared

• Alexia Galati

  22 shared

• Michael J. Spivey

  21 shared

• Alexandra Paxton

  University of Connecticut

  21 shared

Labs

• Co-Mind LabPI

  Conducts research on the cognitive science of communication, focusing on the dynamic coordination of communication and cognition in time.

Awards & honors

• Professor & Mark Allen Itkin Centennial Chair in Communicati…