I am a Learning and Data Scientist, analyzing complex data and helping create personalized and engaging learning experiences for kids and find ways to share information about kids' learning abilities and performance with their educators and families. I do this by executing computational simulations of learner behavior, applying latent trait estimation algorithms, and working cross-functionally with clients, engineers, and our product group in order to ensure that partner company products meet expectations.
I have a strong background in Cognitive Science, Neuroscience and Linguistics, and extensive experience with experimental research methods, advanced statistical modeling (mixed models, latent trait models, nonlinear data analysis), and programming in R and Python.
I completed my PhD studying how the motor system of the brain and our bodies contribute to perception, cognition, and empathy.
Current cognitive assessments for young children are labor-intensive on the part of educators and prone to subjective biases, limiting the utility of the data these assessments could generate. As a learning scientist, I designed gamified cognitive tasks to measure student learning enabling our client to release a digital assessment platform for early education language skills. This platform empowers educators to assess young children more quickly and accurately, while improving the overall engagement with learning by including fun stories and immersive environments. My primary role on this project was to identify and translate current assessment paradigms into accessible digital designs and engage with a crossfunctional team of engineers, product designers, and client executives to implement the platform.
Gamified testing tools need to be calibrated for precision, but for adaptive tools, collecting the data to calibrate is a time-consuming and expensive process. While calibration is essential, the overall flow and engagement of the experience must also be maintained. To address these challenges, I developed a computational framework for simulation of learner behavior using R. The framework enables us to measure the performance of content recommendation algorithms, which can reduce the number of questions needed for assessments and improve the overall accuracy. In addition, we can apply constraints to external parameters to improve the continuity and engagement of the experience.
Mental simulation is ubiquitous, yet definitions of what simulation means vary from researcher to researcher and even more so from research domain to research domain. Furthermore, a consensus on the underlying mechanisms of mental simulation still evade us. I am interested in understanding what simulation as a process shares across domains such as music listening, language processing, action observation, and sensorimotor prediction.
Language is highly embodied and situated in the environments in which it is used. Researchers have shown that neural regions corresponding to actual perception/motor acts are also active while processing language about these perceptions and motor acts. Yet we have few experimental methods that can actually demonstrate the process by which concepts become embodied during learning.
I designed and executed a series of experiments investigating how the affordances available while learning new words influence how those words are grounded in our bodies and in corresponding neural areas responsible for actions. Using mixed-effects models to account for nested variation within and between subjects, I demonstrated that participants rapidly acquire new embodied word associations in an immersive virtual reality environment.
Many experiments have shown that motor regions of the brain are active during perception of others' actions, as well as sensory prediction of spatiotemporal information like rising and falling pitch. I designed and conducted experiments using Transcranial Magnetic Stimulation (TMS) to study how motor areas of the brain learn to predict during these sensorimotor tasks. TMS is a tool for studying the brain that allows researchers to probe the neural activity in the cortical motor network, which measures the strength of sensorimotor associations and can act as a rough measure of current "readiness-to-act". I collected multimodal demographic, behavioral, and neural data, developed filtering and preprocessing analytic processes, and applied statistical models including mixed-effects, repeated-measures ANOVA, and power analyses.
While a lot of work in the linguistic domain looks at embodied language, most theories of motor control don’t account for effects of language on motor tasks. Another VR task developed by myself and a few collaborators looks at motor learning in a 3-dimensional virtual world and examines the effect that linguistic labels has on motor learning. Other work I’ve been involved in has shown particularly interesting relationships between motor learning improvement and improvement in learning a linguistic label.
Using NLP to classify and predict product ratings using app store reviews, and to predict assessment item parameters using question data
Predicting placement satisfaction for PhD Philosophy graduates using clustering methods as well as multilevel modelingUsing mouse-tracking methodology to capture online uncertainy in a word learning paradigm Combining neuroimaging findings using meta-analysis to better understand the relationship between music and the neural motor network Using TMS to down-regulate activation in targeted cortical regions to determine functional contributions of that region to behavior