MA
M. Al Nahadi
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
2 records found
1
"Girls are bad at math, but are better than boys in linguistic subjects!" Such stereotypes, potentially manifesting through various sources, can impact children's development and negatively affect their academic performance. In this work, we study the presence of stereotypes among the Top-10 suggestions of Recommender Algorithms. In particular, we conduct an empirical exploration using an extensive suite of Recommender Algorithms on two well-known datasets from different domains: MovieLens-1M (movies) and Goodreads (books). We aim to assess the presence of gender, race, and religion stereotypes. We utilize three stereotype detection models, based on machine learning and Large Language models, and we leverage performance metrics to contextualize the Recommender Algorithms and stereotype prominence metrics to measure the extent of their presence in the recommendations. Outcomes from this work evidence that stereotypes are not equally prominent across all Recommender Algorithms, with certain content-based and deep-learning models showing higher tendencies to recommend stereotypical content to children. Findings emerging from our exploration result in several implications for researchers and practitioners to consider when designing and deploying Recommender Algorithms, especially when children are also interacting with these systems. Furthermore, this work presents a blueprint in which stereotype detection can be expanded to other domains, other types of stereotypes, and other demographic user groups.
...
"Girls are bad at math, but are better than boys in linguistic subjects!" Such stereotypes, potentially manifesting through various sources, can impact children's development and negatively affect their academic performance. In this work, we study the presence of stereotypes among the Top-10 suggestions of Recommender Algorithms. In particular, we conduct an empirical exploration using an extensive suite of Recommender Algorithms on two well-known datasets from different domains: MovieLens-1M (movies) and Goodreads (books). We aim to assess the presence of gender, race, and religion stereotypes. We utilize three stereotype detection models, based on machine learning and Large Language models, and we leverage performance metrics to contextualize the Recommender Algorithms and stereotype prominence metrics to measure the extent of their presence in the recommendations. Outcomes from this work evidence that stereotypes are not equally prominent across all Recommender Algorithms, with certain content-based and deep-learning models showing higher tendencies to recommend stereotypical content to children. Findings emerging from our exploration result in several implications for researchers and practitioners to consider when designing and deploying Recommender Algorithms, especially when children are also interacting with these systems. Furthermore, this work presents a blueprint in which stereotype detection can be expanded to other domains, other types of stereotypes, and other demographic user groups.
Bachelor thesis
(2019)
-
Wessel Turk, Frank Ooijevaar, Robin Faber, Murtadha Al Nahadi, Neil Yorke-Smith, Otto Visser, Joost Verdoorn, Bas Hintemann
This report concludes the project 'Predictive Learning Analytics' for FeedbackFruits. It includes all the work that has been done in a ten week span at FeedbackFruits. FeedbackFruits is a company inspired to help teachers shape learning activities and spark students' active thinking, by creating an online platform on which teachers can create interactive learning activities. The goal for this project was to extend the analytics tools for interactive presentations that are offered to teachers who use the platform. This hopefully results in giving them more insight in the participation of the class, understanding of the course material and the effort that students put in their school work.
...
This report concludes the project 'Predictive Learning Analytics' for FeedbackFruits. It includes all the work that has been done in a ten week span at FeedbackFruits. FeedbackFruits is a company inspired to help teachers shape learning activities and spark students' active thinking, by creating an online platform on which teachers can create interactive learning activities. The goal for this project was to extend the analytics tools for interactive presentations that are offered to teachers who use the platform. This hopefully results in giving them more insight in the participation of the class, understanding of the course material and the effort that students put in their school work.