Vinayaka Gude

Emotion recognition is fundamental to social intelligence. As artificial intelligence (AI) becomes increasingly common, its ability to recognize human emotions has become crucial. However, whether such systems can match or surpass human experts, particularly in multimodal large language models such as GPT-4o (MLLMs), remains largely unexplored. This study evaluates GPT-4o’s emotion recognition using the Reading the Mind in the Eyes Test (RMET) and its multiracial counterpart (MRMET) against human participants from low-, mid-, and high-performing groups. Results show that, on average, GPT-4o outperforms humans in identifying emotions across both tests. This trend persists across all performance groups. Yet, when we aggregate independent human decisions to simulate collective intelligence, human groups significantly surpass the performance of aggregated GPT-4o predictions, highlighting the wisdom of the crowd effect. Moreover, an augmented intelligence condition that combines human and GPT-4o predictions achieves greater accuracy than humans or GPT-4o alone. These results suggest that, while GPT-4o exhibits strong base-level emotion recognition, the collective intelligence of humans and the potential of human-AI collaboration offer the most promising path toward effective emotion recognition. We discuss the implications of this research, highlighting where and when AI or human-AI collaboration can be most beneficial for society.

This paper explores the impact of the Israel–Palestine conflict on the stock performance of U.S. companies and their public positions on the conflict. In an era where corporate positions on geopolitical issues are increasingly scrutinized, understanding the market implications of such statements is critical. This research aims to capture the complex, non-linear relationships between corporate actions, media coverage, and financial outcomes by integrating traditional statistical techniques with advanced machine learning models. To achieve this, we constructed a novel dataset combining public corporate announcements, media sentiment (including headline and article body tone), and philanthropic activities. Using both classification and regression models, we predicted whether companies had affiliations with Israel and then analyzed how these affiliations, combined with other features, affected their stock returns over a 30-day period. Among the models tested, ensemble learning methods such as stacking and boosting achieved the highest classification accuracy, while a Multi-Layer Perceptron (MLP) model proved most effective in forecasting abnormal stock returns. Our findings highlight the growing relevance of machine learning in financial forecasting, particularly in contexts shaped by geopolitical dynamics and public discourse. By demonstrating how sentiment and corporate stance influence investor behavior, this research offers valuable insights for investors, analysts, and corporate decision-makers navigating sensitive political landscapes.

The COVID-19 pandemic has impacted people’s lives in various cities across the United States. This study explored the factors influencing anticipated migration patterns as indicated by the house search on the Redfin platform in 132 United States cities before, during, and after the COVID-19 Pandemic. The research addresses a crucial gap in the existing literature on recent trends in migration patterns, factors influencing this, and the potential impact of the pandemic. We employed a multivariate regression model to study overall migration and additional models to examine yearly data, identifying key factors that drive people to relocate. We identified that regions with better healthcare services, lower income inequality, and lower unemployment rates increasingly attract residents. Cities with greater racial and ethnic diversity, as well as high school completion rates, were also appealing until 2020 when the importance of education in migration decisions was overshadowed by housing affordability issues. Meanwhile, proximity to larger cities played a significant role, with people favoring smaller cities nearby for better growth opportunities. Our findings suggest that developing cities with smaller populations are becoming preferred destinations due to their mix of affordability, job opportunities, and better living conditions. These insights are invaluable for policymakers, city planners, and community organizations, helping guide urban development strategies and address challenges such as income inequality, neighborhood segregation, and healthcare accessibility.

This study investigates the influence of social media presence and conflict response on the stock returns during the Russia–Ukraine war. We examined the long-term impacts regarding social media presence, response time, action taken, and industry affiliation using a sample of 174 firms in 10 industrial sectors. The results highlight that response time and actions significantly impacted stock returns in both the short- term and long-term. Conversely, social media presence marginally affected response decisions, but did not significantly affect stock returns.

Asphyxiation associated with metabolic acidosis is one of the common causes of fetal deaths. The paper aims to develop a feature extraction and prediction algorithm capable of identifying most of the features in the SISPORTO software package and late and variable decelerations. The resulting features were used for classification based on umbilical cord pH data. The algorithms developed here were used to predict cord pH levels. The prediction system assists the obstetricians in assessing the state of the fetus better than the category methods, as only about 30% of the patients in the pathological category suffer from acidosis, while the majority of acidotic babies were in the suspect category, which is considered lower risk. By predicting the direct indicator of acidosis, umbilical cord pH, this work demonstrates a methodology, which uses fetal heart rate and uterine activity, to identify acidosis. This paper introduces a forecasting model based on deep learning to predict heart rate and uterine contractions, integrated with the classification algorithm, resulting in a robust tool for predictive fetal monitoring. The hybrid algorithm resulted in a model capable of providing future conditions of the fetus, which obstetricians can use for diagnosis and planning interventions. The ensemble classification algorithm had a test accuracy of 85% (n = 24) in predicting fetal acidosis on the features extracted from the cardiotocography data. When integrated with the classification model, the results from the prediction model (long short-term memory network) can effectively identify fetal acidosis 2 or 4 min in the future.

Covid-19 has impacted the lives of people across the world with deaths and unprecedented economic damage. Countries have employed various restrictions and lockdowns to slow down the rate of its spread with varying degrees of success. This research aims to propose an optimal strategy for dealing with a pandemic taking the deaths and economy into account. A complete lockdown until vaccination is not suitable as it can destroy the economy, whereas having no restrictions would result in more Covid-19 cases. Therefore, there is a need for a dynamic model which can propose a suitable strategy depending on the economic and health situation. This paper discusses an approach involving a systems dynamics model for evaluating deaths and hospitals and a fuzzy inference system for deciding the strategy for the next time period based on pre-defined rules. We estimated Gross Domestic Product (GDP) as a sum of government spending, investment, consumption, and spending. The resulting hybrid framework aims to attain a balance between health and economy during a pandemic. The results from a 30-week simulation indicate that the model has 2.9 million $ in GDP higher than complete lockdown and 21 fewer deaths compared to a scenario with no restrictions. The model can be used for the decision-making of restriction policies by configuring the fuzzy rules and membership functions. The paper also discusses the possibility of introducing virus variants in the model.

Floods are a complex phenomenon that are difficult to predict because of their non-linear and dynamic nature. Therefore, flood prediction has been a key research topic in the field of hydrology. Various researchers have approached this problem using different techniques ranging from physical models to image processing, but the accuracy and time steps are not sufficient for all applications. This study explores deep learning techniques for predicting gauge height and evaluating the associated uncertainty. Gauge height data for the Meramec River in Valley Park, Missouri was used to develop and validate the model. It was found that the deep learning model was more accurate than the physical and statistical models currently in use while providing information in 15 minute increments rather than six hour increments. It was also found that the use of data sub-selection for regularization in deep learning is preferred to dropout. These results make it possible to provide more accurate and timely flood prediction for a wide variety of applications, including transportation systems.