Browsing by Author "Villalobos, Esteban"

Now showing 1 - 5 of 5
  • No Thumbnail Available
    Item
    Can Feedback based on Predictive Data Improve Learners' Passing Rates in MOOCs? A Preliminary Analysis
    (Association for Computing Machinery, Inc., 2021) Pérez Sanagustín, Mar; Pérez Álvarez, Ronald Antonio; Maldonado Mahauad, Jorge Javier; Villalobos, Esteban; Hilliger, Isabel; Hernández Correa, Josefina María; Sapunar Opazo, Diego Andrés; Moreno-Marcos, Pedro Manuel; Muñoz-Merino, Pedro J.; Delgado Kloos, Carlos; Imaz, Jon
    This work in progress paper investigates if timely feedback increases learners' passing rate in a MOOC. An experiment conducted with 2,421 learners in the Coursera platform tests if weekly messages sent to groups of learners with the same probability of dropping out the course can improve retention. These messages can contain information about: (1) the average time spent in the course, or (2) the average time per learning session, or (3) the exercises performed, or (4) the video-lectures completed. Preliminary results show that the completion rate increased 12% with the intervention compared with data from 1,445 learners that participated in the same course in a previous session without the intervention. We discuss the limitations of these preliminary results and the future research derived from them.
  • No Thumbnail Available
    Item
    Fair Face Verification by Using Non-Sensitive Soft-Biometric Attributes
    (2022) Villalobos, Esteban; Mery, Domingo; Bowyer, Kevin
    Facial recognition has been shown to have different accuracy for different demographic groups. When setting a threshold to achieve a specific False Match Rate (FMR) on a mixed demographic impostor distribution, some demographic groups can experience a significantly worse FMR. To mitigate this, some authors have proposed to use demographic-specific thresholds. However, this can be impractical in an operational scenario, as it would either require users to report their demographic group or the system to predict the demographic group of each user. Both of these options can be deemed controversial since the demographic group is a sensitive attribute. Further, this approach requires listing the possible demographic groups, which can become controversial in itself. We show that a similar mitigation effect can be achieved using non-sensitive predicted soft-biometric attributes. These attributes are based on the appearance of the users (such as hairstyle, accessories, and facial geometry) rather than how the users self-identify. Our experiments use a set of 38 binary non-sensitive attributes from the MAAD-Face dataset. We report results on the Balanced Faces in the Wild dataset, which has a balanced number of identities by race and gender. We compare clustering-based and decision-tree-based strategies for selecting thresholds. We show that the proposed strategies can reduce differential outcomes in intersectional groups twice as effectively as using gender-specific thresholds and, in some cases, are also better than using race-specific thresholds.
  • Thumbnail Image
    Item
    Fair-face verification by using non-sensitive soft-biometric attributes
    (2021) Villalobos, Esteban; Mery Quiroz, Domingo; Pontificia Universidad Católica de Chile. Escuela de Ingeniería
    Los algoritmos de reconocimiento facial han demostrado tener diferencias en los resultados entre los distintos grupos demográficos. Incluso cuando se establece un umbral global para obtener una tasa de falsas coincidencias (FMR) específica para todo el sistema, algunos grupos demográficos pueden obtener resultados significativamente peores que los indicados. Para mitigar esto, algunos autores han propuesto utilizar umbrales específicos por grupos demográficos. Sin embargo, esto es poco práctico en un entorno operativo, ya que requeriría que los usuarios informaran de su grupo demográfico o lo predijeran en el sistema. Ambas opciones son controversiales, debido a que el dato del grupo demográfico es sensible. Además, en el caso de utilizar umbrales basados en un grupo racial, requiere enumerar exhaustivamente todas las razas posibles para el sistema. Demostramos que se puede conseguir un efecto de mitigación similar utilizando atributos biométricos blandos predecibles no sensibles. Se trata de atributos basados en la apariencia de los sujetos que no dependen de cómo se identifican los usuarios (como el peinado, los accesorios y la geometría facial). Utilizamos 38 atributos binarios no demográficos del conjunto de datos MAADFace. Presentamos los resultados en el conjunto de datos BFW, que tiene un número equilibrado de identidades por raza y género. Comparamos las estrategias basadas en la agrupación y en los árboles de decisión como formas de seleccionar estos umbrales. Demostramos que estas estrategias pueden reducir los resultados diferenciales en los grupos interseccionales con el doble de eficacia que el uso de umbrales específicos de género y, en algunos casos, también son mejores que el uso de umbrales específicos de raza.
  • Thumbnail Image
    Item
    Learning Inequalities Between Indigenous and Non-indigenous Children in Latin America: The Bridge Between School and Families to Promote Quality
    (Springer, 2023) Treviño, Ernesto; Villalobos, Esteban; Godoy Ossa, Felipe Andrés
    Latin America has been making progress in incorporating education policies with an intercultural bilingual focus in parallel with the struggles and demands of indigenous peoples, which international organizations such as the United Nations Educational, Scientific and Cultural Organization (UNESCO), among others, have helped to place on the global agenda.
  • No Thumbnail Available
    Item
    Student Attendance System in Crowded Classrooms Using a Smartphone Camera
    (2019) Mery Quiroz, Domingo Arturo; Mackenney, Ignacio; Villalobos, Esteban