Search for: All records

Deep-sea hydrothermal vents inject dissolved and particulate metals, dissolved gasses, and biological matter into the water column, creating plumes several hundred meters above the seafloor that can be traced thousands of kilometers. To understand the impact of these plumes, rosettes equipped with sample bottles and in situ instruments, e.g., for turbidity, oxidation-reduction potential, and temperature, have been key tools for collecting water column fluid for informative ex situ analysis. However, deploying rosettes strategically in distal (>1km) plume-derived fluids is difficult when plume material is entrained rapidly with background water and transported by complicated bathymetric, internal, and/or tidal currents. This problem is exacerbated when the controlling dynamics are also poorly constrained (e.g., no persistent monitoring, few historical data) and data collected while in the field to estimate or compensate for these dynamics are only available to be analyzed hours or days following an asset deployment. Autonomous underwater vehicles (AUVs) equipped with equivalent in situ instruments to rosettes excel at exploration missions and creating highly-resolved maps at different spatial scales. Utilization of AUVs for hydrothermal plume charting and strategic sampling with rosettes is at a techno-scientific frontier that requires new data transmission and visualization interfaces for supporting real-time evidence-based operational decisions made at sea. We formulated a method for monitoring in situ water properties while an AUV is underway that (1) builds situational awareness of deep fluid mass distributions, (2) allows scientists-in-the-loop to rapidly identify fluid distribution patterns that inform adaptations to AUV missions or deployments of other assets, like rosettes, for targeted sample collection, and (3) supports robust formulation of working hypotheses of plume dynamics for in-field investigation. We will present a description of the method with preliminary results from cruise AT50-15 (Juan de Fuca Ridge, 2023) using AUV Sentry and discuss how supervised autonomy will improve ocean robotics for future science missions. 

Hospitalization of patients with chronic diseases poses a significant burden on the healthcare system. Frequent hospitalization can be partially attributed to the failure of healthcare providers to engage effectively with their patients. Recently, patient portals have become popular as information technology (IT) platforms that provide patients with online access to their medical records and help them engage effectively with healthcare providers. Despite the popularity of these portals, there is a paucity of research on the impact of patient–provider engagement on patients’ health outcomes. Drawing on the theory of effective use, we examine the association between portal use and the incidence of subsequent patient hospitalizations, based on a unique, longitudinal dataset of patients’ portal use, across a 12-year period at a large academic medical center in North Texas. Our results indicate that portal use is associated with improvements in patient health outcomes along multiple dimensions, including the frequency of hospital and emergency visits, readmission risk, and length of stay. This is one of the first studies to conduct a large-scale, longitudinal analysis of a health IT system and its effect on individual level health outcomes. Our results highlight the need for technologies that can improve patient–provider engagement and improve overall health outcomes for chronic disease management. 

Interactive learning environments facilitate learning by providing hints to fill the gaps in the understanding of a concept. Studies suggest that hints are not used optimally by learners. Either they are used unnecessarily or not used at all. It has been shown that learning outcomes can be improved by providing hints when needed. An effective hinttaking prediction model can be used by a learning environment to make adaptive decisions on whether to withhold or provide hints. Past work on student behavior modeling has focused extensively on the task of modeling a learner’s state of knowledge over time, referred to as knowledge tracing. The other aspects of a learner’s behavior such as tendency to use hints has garnered limited attention. Past knowledge tracing models either ignore the questions where a hint was taken or label hints taken as an incorrect response. We propose a multi-task memory-augmented deep learning model to jointly predict the hint-taking and the knowledge tracing task. The model incorporates the effect of past responses as well as hints taken on both the tasks. We apply the model on two datasets – ASSISTments 2009-10 skill builder dataset and Junyi Academy Math Practicing Log. The results show that deep learning models efficiently leverage the sequential information present in a learner’s responses. The proposed model significantly out-performs the past work on hint prediction by at least 12% points. Moreover, we demonstrate that jointly modeling the two tasks improves performance consistently across the tasks and the datasets, albeit by a small amount.