Senior Projects 2023

Author: Bina Aaronson
Title: Impacts of Sleep Health and ADHD Symptoms on Adolescents’ Reward and Loss Anticipation
Advisor: Claudia Lugo-Candelas

Abstract:
Poor sleep health has significant cognitive consequences, such as impairing the ability to competently anticipate and respond to incentives. However, it is unclear how poor sleep health is associated with reward and loss anticipation in adolescents, as well as if this association varies as a function of ADHD symptoms. Using data from the Adolescent Brain Cognitive Development (ABCD) study, an open-access study of child development, the current study analyzed adolescents' brain activity during the anticipation of reward or loss during a Monetary Incentive Delay (MID) MRI task, as well as sleep health data measured via the fitbit. Analyses revealed that sleep duration and sleep midpoint did not have an effect on all adolescents’ reward or loss anticipation. However, we documented atypical neural activity for adolescents with ADHD symptoms, where they either showed deactivation during anticipation of loss and reward or less activation than their typically developing (TD) peers. Greater sleep duration and earlier sleep midpoint were related to even greater deactivation. These findings suggest that adolescents with ADHD symptoms seem to anticipate loss and reward differently, and that sleep health may have more of an impact on the reward processing of adolescents with ADHD than that of TD adolescents. Such results indicate that there may be a link between sleep, ADHD symptoms, and reward and loss anticipation.

Author: Samuel Hutchinson
Title: Classifying Online Misinformation with Cognitively-Informed Features from Transformer Language Models
Advisor: Christopher Baldassano

Abstract:
Online misinformation has crept into the public consciousness through discourse on topics ranging from presidential elections to the COVID-19 pandemic. Misinformation-containing statements spread both farther and faster than true statements on social networks, suggesting the need for an interpretable, algorithmic flagging mechanism. This investigation endeavors to devise such an algorithm based on cognitively-plausible explanations for misinformation’s virality: informational novelty and emotional valence. I use next-word prediction error measures from a GPT-2 model fine-tuned on true news stories from Reuters to assess the novelty component of this problem and a pre-trained RoBERTa-based sentiment classifier for the emotional-valence component. To create a classification model, I calculated the joint distribution of these errors and sentiments over a subset of true news stories and misinformation-containing stories from the ISOT Fake News dataset. I then used this joint distribution to predict the likelihood that unseen news stories contain misinformation. This model classifies news stories with around 79.4% accuracy, furthering prior work showing similarity in next-word prediction between human readers and generative models like GPT-2. These results also indicate that online misinformation may be classifiable through computable and cognitively-interpretable natural-language metrics.

Author: Corinna Jones
Title: Imposter Phenomenon and the Fear of Success
Advisor: Lisa Son
Research Assistants: Katerina Millner, Emma Ziessler, and Liana Eisler

Abstract:
The purpose of this study was to see if there is an association between the imposter phenomenon and the fear of success. The study involved a participant and a confederate taking a critical thinking test followed by an acknowledgment of success from the confederate to the participant. Participants were then given the option of choosing the difficulty of their next test. We found that the imposter phenomenon does have an association with a fear of success. Additionally, we found variances based on ethnicity that may have implications for the understanding of the imposter phenomenon.

Author: Jane Mok
Title: Modeling Curiosity as a Positive Feedback Loop: The Role of Prediction Errors in Energizing Epistemic Curiosity in Learning
Advisors: Daphna Shohamy, Yaniv Abir

Abstract:
Most curiosity research has so far focused on a single learning experience, and there is an absence in the literature in characterizing how prediction errors—gaps in information—can pique curiosity and sustain it during learning over separate learning experiences. Through measuring how much someone is willing to wait for answers to trivia questions, and the difference in the anticipation and subsequent satisfaction of the information, we ask: “what is the pattern in which past positive epistemic experiences energize us to be more curious, and negative outcomes inhibit?” Modeling these valence patterns provide important insights that allow us to better understand and harness curiosity for effective learning, thus improving academic and life outcomes.

Author: Isobel Moure
Title: Neural Discrimination of Syntactic Structures: Connecting Theory to Brainwaves
Advisors: Christos Papadimitriou, Tony Ro, Tatiana Aloi Emmanouil

Abstract:
Do our theories of syntax match what we can measure in the brain? This project sought evidence that the mind discriminates between different syntactic structures, replicating and extending the findings of the Ding et al. 2016 study. Participants, monitored with an EEG cap, were presented with synthetic speech sentences of different classes that would hierarchically merge at different rates, according to English phrase structure. A story class was added to test if sentences that were part of a larger narrative were processed differently. A Support Vector Machine was trained on the data and used to attempt to decode the sentence class from the EEG data. It was found that on the majority of conditions, the classifier obtained an accuracy higher than chance. These data suggest hierarchical processing of constituent structure and provide further evidence for the Assemblies hypothesis as well as support the previous findings from Ding et al.

Author: Katherinne Rabanal
Title: The Association of the Interpersonal Environment and Memory Complaints: An Exploration of the Complex Interplay of Social, Psychological, and Cognitive Factors
Advisors: Emily Hokett, Jennifer J. Manly

Abstract:
Community subjective social status (CSSS) is an individual’s evaluation of their position concerning others in their community. CSSS is related to health and wellness, independent of objective socioeconomic status or other objective factors. Memory complaints indicate mild cognitive impairment, highly associated with high depressive symptoms. The current study aims to determine the relationship between community subjective social position and memory complaints. Previous research highlights how memory complaints indicate objective cognitive function, the pivotal factors of stress, and decreased motivation leading to memory difficulties. Participants in this study are from a longitudinal study (the Offspring study) and are the adult children of the Washington Heights Irving Columbia Aging Study participants; most participants identify as Hispanic/Latinx. CSSS is collected from a 10-ring MacArthur Network community ladder referring to an individual’s community. Outcomes were self-reported scores from a questionnaire and interviews from the baseline data. We hypothesize that people with lower subjective community social positions and low self-esteem will show more subjective cognitive function complaints. Hierarchical multiple regression analysis was performed, using self-esteem and depressive symptoms total scores along with other demographic characteristics. Results provide preliminary evidence for CSSS as a predictor of memory complaints. Potential projects and the implementation of a follow-up group to test memory complaints over time and community-centered care programs are suggested.

Author: Samantha Seltzer
Title: Mechanisms of Language Evolution: Evidence from Handshapes of Emergent Sign Languages in Nicaragua and São Tomé & Príncipe
Advisor: Ann Senghas
Co-Advisor: Michele Miozzo

Abstract:
Languages change and evolve over time and across generations, subject to the constraints of both anatomical and cognitive factors. In natural sign languages, there exists a standard corpus of basic handshapes observed across mature languages (Miozzo 2022). In this study, we sought to understand the nature of this evolution and whether the impact of biological constraints pushes handshapes towards or away from cross-linguistic homogeneity. We compared the variety of handshapes used in signs produced by the cohort who created Nicaraguan Sign Language (LSN), whose emergence began in the 1970s, to the handshapes of a second cohort who learned LSN from them. We then compared these LSN handshapes to those of an even more recently emerging sign language from São Tomé and Principe (LGSTP), whose evolution began in the early 2000s, and finally to the mature language corpus. In this way, we were able to compare the effects of language use over time to transmission across age cohorts. The data suggests that emergent handshapes approach those of mature sign languages and that these changes occur in the transmission of language across cohorts, via learning, rather than through language use over time.

Author: Ariel Chuhe Wu
Title: Exploring the Use of Virtual Reality as a Visualization Tool for Sub-micro Chemistry Knowledge
Advisor: Elliot Matthew Hu-Au

Abstract:
Molecular Chemistry poses a challenge for many K-12 students due to its abstract nature. Understanding topics like atomic structures and molecular behaviors requires strong visual-spatial skills, and students often rely on visual tools to aid in the visualization process. The use of virtual reality (VR) has been proposed as an educational tool to facilitate learning and visualization. However, concerns have been raised about its effectiveness compared to traditional methods, as well as its potential to distract students. This study aims to address these concerns by comparing the effectiveness of traditional pedagogical tools, such as ball and stick manipulatives, with VR in teaching students about molecular structures. A between-subject experimental study was conducted to measure the learning gains of students who received a molecular lesson in either a ball and stick format or a VR format. The study involved 80 students (N = 80), aged 11-18 years old, from a local community center. The participants completed a pretest, received the molecular lesson in one of the two formats, and then took a post-test. The results indicate that learning with VR is as effective as learning with physical models, and both VR and physical manipulatives demonstrated significant improvement in three out of the four drawing categories: molecule shape, atom quantity, and relative sizes. This study provides evidence that VR is an effective tool for teaching abstract chemistry concepts and can serve as a valuable substitute or complement to traditional methods.