Trustworthy AI Through Transparency, Generalization, and Fairness

Abstract

Machine learning has demonstrated its power in a wide range of real-world applications. However, for many serious tasks today, its ability still needs to surpass humans. Oftentimes, it loses users' trust due to noise or adversarial attacks. This work enhances the trustworthiness of deep neural network (DNN) models from three perspectives: transparency, generalization, and fairness. Our unique approaches show that enhancing the performance of one perspective does not have to conflict with the other two as a trade-off. Specifically, we introduce ConceptX for linguistic interpretation of the neural representation, deep reinforcement-learning enhanced data selection, and fair decision-making based on model interpretation. These methods provide up to 30% better prediction accuracy using about half of the training data on the LM task. Furthermore, when excluding personal (biased) features from the training of prediction tasks, our model achieves similar accuracy holding a promise of fairness and reliability.

Biography

Jia Xu is an assistant professor at the Stevens Institute of Technology, and previously, she was a faculty member and Ph.D. advisor at Tsinghua University. Her research interests are Machine Learning and Natural Language Processing (NLP), focusing on highly competitive AI systems. She has more than 40 papers and regularly publishes in mainstream venues in NLP and machine learning (e.g., AAAI, ICML, ACL, EMNLP, NAACL) with 1220 citations. Professor Xu holds a Diploma from TU-Berlin and a Doctorate degree from RWTH Aachen University in Germany. During this time, she had industrial Internships at IBM in Watson and Microsoft Research (MSR) Redmond. Professor Xu has a unique record of winning over ten NLP competition awards with her team, including WMT and NIST open machine translation. Her team, with members from five  nationalities, was ranked as the second place in the latest AlexaPrize social bot challenge.