Search for: All records

Student engagement is a key predictor of academic achievement and is closely linked to career awareness, interest, and preparedness. Measuring student engagement during STEM learning is challenging for teachers, given the dynamic and ever-changing nature of these learning environments. Even when engagement data can be collected, leveraging this information to refine and personalize instruction requires significant experience and time. To address this, we are developing Scoutlier EngagEd, a digital teaching assistant that embeds in existing Learning Management Systems (LMS) to automatically and invisibly gather multidimensional data on student engagement and performance during STEM learning. These data are being leveraged to model student learning and generate insights that produce human-like, actionable recommendations through a Large Language Model (LLM) for teachers to improve STEM learning outcomes. 

Chart question answering (CQA) is a newly proposed visual question answering (VQA) task where an algorithm must answer questions about data visualizations, e.g. bar charts, pie charts, and line graphs. CQA requires capabilities that natural-image VQA algorithms lack: fine-grained measurements, optical character recognition, and handling out-of-vocabulary words in both questions and answers. Without modifications, state-of-the-art VQA algorithms perform poorly on this task. Here, we propose a novel CQA algorithm called parallel recurrent fusion of image and language (PReFIL). PReFIL first learns bimodal embeddings by fusing question and image features and then intelligently aggregates these learned embeddings to answer the given question. Despite its simplicity, PReFIL greatly surpasses state-of-the art systems and human baselines on both the FigureQA and DVQA datasets. Additionally, we demonstrate that PReFIL can be used to reconstruct tables by asking a series of questions about a chart. 

Research has shown that student achievement is influenced by their access to, or possession of, various forms of capital. These forms of capital include financial capital, academic capital (prior academic preparation and access to academic support services), cultural capital (the attitudes, knowledge, and behaviors related to education which students are exposed to by members of their family or community), and social capital (the resources students have access to as a result of being members of groups or networks). For community college students, many with high financial need and the first in their families to go to college (especially those from underrepresented minority groups), developing programs to increase access to these various forms of capital is critical to their success. This paper describes how a small federally designated Hispanic-serving community college has developed a scholarship program for financially needy community college students intending to transfer to a four-year institution to pursue a bachelor’s degree in a STEM field. Developed through a National Science Foundation Scholarships in Science, Technology, Engineering, and Mathematics (S-STEM) grant, the program involves a collaboration among STEM faculty, college staff, administrators, student organizations, and partners in industry, four-year institutions, local high schools, and professional organizations. In addition to providing financial support through the scholarships, student access to academic capital is increased through an intensive math review program, tutoring, study groups, supplemental instruction, and research internship opportunities. Access to cultural and social capital is increased by providing scholars with faculty mentors; engaging students with STEM faculty, university researchers, and industry professionals through field trips, summer internships, professional organizations, and student clubs; supporting student and faculty participation at professional conferences, and providing opportunities for students and their families to interact with faculty and staff. The paper details the development of the program, and its impact over the last five years on enhancing the success of STEM students as determined from data on student participation in various program activities, student attitudinal and self-efficacy surveys, and academic performance including persistence, retention, transfer and graduation. 

ABSTRACT Continuous, sub‐centennially resolved, paleo terrestrial records are rare from arid environments such as the Pacific south‐west United States. Here, we present a multi‐decadal to centennial resolution sediment core (Lake Elsinore, CA) to reconstruct late Wisconsin pluvials, droughts and vegetation. In general, the late Wisconsin is characterized by a wetter and colder climate than during the Holocene. Specifically, conditions between 32.3 and 24.9k cal a BP are characterized by large‐amplitude hydrologic and ecologic variability. Highlighting this period is a ∼2000‐year glacial mega‐drought (27.6–25.7k cal a BP) during which the lake shallowed (3.2–4.5 m depth). This period is approximately coeval with a Lake Manix regression and an increase in xeric vegetation in the San Bernardino Mountains (Baldwin Lake). The Local Last Glacial Maximum (LLGM) is bracketed between 23.3 and 19.7k cal a BP − a ∼3000‐year interval characterized by reduced run‐off (relative to the glacial), colder conditions and vegetative stability. Maximum sustained wetness follows the LLGM, beginning at 19.7 and peaking by 14.4k cal a BP. A two‐step decrease in runoff characterizes the Lateglacial to Holocene transition; however, the vegetation change is more complex, particularly at the beginning of the Younger Dryas chronozone. By 12.6–12.4k cal a BP, the climate achieved near Holocene conditions.