A carregar...
Projeto de investigação
Information Management Research Center
Financiador
Autores
Publicações
The Feeling Economy in Healthcare
Publication . Girão Carrilho, Mariana; Pinto, Diego Costa; Information Management Research Center (MagIC) - NOVA Information Management School; NOVA Information Management School (NOVA IMS)
This study investigates the impact of Conversational AI on satisfaction in healthcare services. Drawing upon the Feeling Economy framework, we demonstrate that Conversational AI, as compared to human agents, negatively affects satisfaction primarily due to a perceived lack of empathy, particularly in scenarios involving physical health symptoms. Conversely, in mental health contexts, Conversational AI is preferred, indicating that its technological attributes provide unique benefits to consumers that extend beyond empathy. Across three experimental studies, the findings highlight a nuanced role of AI in healthcare: while it may negatively impact certain service encounters, it also offers enhancements in others, depending on the specific health issues involved. This research offers significant insights for healthcare providers and AI developers on optimizing AI deployment in healthcare settings.
Drivers of academic achievement in high school
Publication . Beatriz-Afonso, Ana; Cruz-Jesus, Frederico; Nunes, Catarina; Castelli, Mauro; Oliveira, Tiago; Castro, Luísa Canto e; Information Management Research Center (MagIC) - NOVA Information Management School; NOVA Information Management School (NOVA IMS); International Forum of Educational Technology and Society,National Taiwan Normal University
Education is crucial for individual and societal growth. However, it was significantly impacted by the COVID-19 pandemic, with long-lasting effects. Estimates suggest that students’ learning decreased by up to 50% compared to a typical year, though the full impact remains unclear. This paper evaluates primary AA drivers to guide efforts addressing pandemic-related educational inequities. Using government data from virtually all public high school students in a European country, we applied advanced data science methods— Multiple Linear Regression, Decision Trees, Neural Networks, Support Vector Machines, Random Forest, and Extreme Gradient Boosting—to analyze AA determinants before and during the pandemic (2019 and 2020, respectively). Our data includes the most well-known potential AA drivers across four dimensions: students, parents, schools, and teachers. Our substantive findings highlight that student age and legal guardian education were key AA drivers, while Internet access and gender gained importance during the pandemic. Additional drivers, including school size, family nationality, and socioeconomic factors (such as the rate of students receiving school support), also emerged as relevant, particularly under pandemic conditions. This study quantitatively assesses these AA determinants across two distinct academic years, providing nuanced insights into the impact of COVID-19 on education. These results offer valuable guidance for policymakers to implement interventions addressing evolving needs and disparities exacerbated by remote learning. This study contributes to AA literature by utilizing extensive data and machine learning models to reveal enduring and emerging factors affecting educational outcomes during challenging times.
Towards a temporal privacy-preserving deep learning approach to residential address matching [poster]
Publication . Cruz, Paula; Information Management Research Center (MagIC) - NOVA Information Management School; NOVA Information Management School (NOVA IMS)
Integrating data from various sources has become a crucial component of decision-making processes in several application domains. In the absence of a unique identifier, addresses can serve as quasi-identifiers in the linkage of records related to the same entity in one or more data collections. Address matching is the process of identifying pairs of records by comparing full addresses or address fields, with the goal of obtaining the best matching result in relation to a searched address. Deep learning (DL) methods have gained popularity within the field of address matching due to their ability to extract semantic and contextual information from non-standard address records with redundant or missing address elements and few literal overlaps. Transformer-based algorithms are among the most used, including pretrained language models (PTLM) such as BERT and, more recently, large language models (LLMs) like ChatGPT. Based on the current state of art, two important research gaps in the field of residential address matching need to be addressed: the use of temporal data, such as creation or update timestamps, and the adoption of privacypreserving methods combining DL algorithms with the use of encrypted or non-encrypted encoding or synthetic data.
Assessing coastal vulnerability at the village level using a robust framework, the example of Canacona in South Goa, India
Publication . Nigam, Ritwik; Luis, Alvarinho J.; Gagnon, Alexandre S.; Vaz, Eric; Damásio, Bruno; Kotha, Mahender; Information Management Research Center (MagIC) - NOVA Information Management School; NOVA Information Management School (NOVA IMS); Elsevier
Climate change poses a significant threat to coastal regions worldwide. This study presents and applies a modified CVI to assess coastal vulnerability at the village level, focusing on Canacona, a taluka in South Goa, India. It adapts the existing CVI methodology by incorporating additional variables to represent the various dimensions of vulnerability better, resulting in 21 variables split into a Physical Vulnerability Index (PVI) and a Social Vulnerability Index (SoVI). The results show spatial variability in coastal vulnerability across the studied villages, with Agonda and Nagercem-Chaudi found to be highly vulnerable and Loliem to be the least vulnerable. A hydrological modeling approach is also used to compare the CVI of every village with their susceptibility to inundation due to rising sea levels. The result demonstrates the influence of local factors on vulnerability, challenging previous taluka-level assessments given the typical scale upon which adaptation typically takes place.
Full-Reference Image Quality Expression via Genetic Programming
Publication . Bakurov, Illya; Buzzelli, Marco; Schettini, Raimondo; Castelli, Mauro; Vanneschi, Leonardo; NOVA Information Management School (NOVA IMS); Information Management Research Center (MagIC) - NOVA Information Management School; Institute of Electrical and Electronics Engineers (IEEE)
Full-reference image quality measures are a fundamental tool to approximate the human visual system in various applications for digital data management: from retrieval to compression to detection of unauthorized uses. Inspired by both the effectiveness and the simplicity of hand-crafted Structural Similarity Index Measure (SSIM), in this work, we present a framework for the formulation of SSIM-like image quality measures through genetic programming. We explore different terminal sets, defined from the building blocks of structural similarity at different levels of abstraction, and we propose a two-stage genetic optimization that exploits hoist mutation to constrain the complexity of the solutions. Our optimized measures are selected through a cross-dataset validation procedure, which results in superior performance against different versions of structural similarity, measured as correlation with human mean opinion scores. We also demonstrate how, by tuning on specific datasets, it is possible to obtain solutions that are competitive with (or even outperform) more complex image quality measures.
Unidades organizacionais
Descrição
Palavras-chave
Contribuidores
Financiadores
Entidade financiadora
Fundação para a Ciência e a Tecnologia
Programa de financiamento
6817 - DCRRNI ID
Número da atribuição
UIDB/04152/2020