From Idea to Publication-Ready Research

Won 1st in Category at Washington State Science & Engineering Fair, Published in: Curieux Academic Journal​

This research investigated how deep-learning systems interpret human emotions across different cultural groups. Two ResNet-18–based convolutional neural networks were trained on separate datasets: one made up of South Asian faces and the other primarily Caucasian (FER-2013). Testing revealed clear cultural bias—South Asian-trained models tended to label Caucasian expressions as “happy,” while Caucasian-trained models often identified South Asian expressions as “sad” or “angry.” After enriching each dataset with a small percentage of the other group’s images, accuracy across demographics improved significantly without reducing performance on the original group. The results show that modest diversification of training data can help reduce bias and improve fairness in emotion recognition systems used in sensitive domains like healthcare, law enforcement, and education.

Published in: Curieux Academic Journal​

Emerging tools for genetic selection raise difficult ethical and social questions. Advocates see opportunities to lessen human suffering by reducing inherited conditions, while critics warn about reinforcing inequality and creating narrow definitions of genetic “normalcy.” Lessons from the history of eugenics illustrate the risks of exclusion and discrimination, and debates about disability highlight that quality of life is not easily defined. Access to genetic technologies is also shaped by socioeconomic inequality, adding another layer of concern. Responsible progress in this field depends on strong safeguards, inclusive decision-making, and a commitment to respecting human diversity.