Transparency is key to displaying the competence of machine learning models to experts in healthcare and criminal justice. I study how algorithmic transparency affects human decision-making.

Umang Bhatt

The Alan Turing Institute is the United Kingdom's national institute for data science and artificial intelligence (AI). The Turing Enrichment scheme offers students currently enrolled on a doctoral programme at a UK university the opportunity to join the Alan Turing Institute for up to twelve months where they can continue their PhD research while forging new inter-institution collaborations. 

Umang is a second-year PhD student in the Machine Learning Group within the Computational and Biological Learning Lab in the Department of Engineering, where he is supervised by Adrian Weller. His research interests broadly lie in statistical machine learning, explainable AI, and human-machine collaboration.  He is currently studying how transparency affects the formation of a successful and productive human-machine partnership. Specifically, he is concerned with the role of explanation and uncertainty in facilitating interaction between machine learning models and human decision makers. During his enrichment scheme, Umang looks forward to collaborating with the Safe and Ethical AI programme on problems related to the fairness, transparency, and robustness of AI systems. He plans to leverage the Turing Institute’s expertise in public policy to augment his machine learning transparency research and to forge connections with domain experts and regulators.

Umang says "I am primarily interested in trustworthy machine learning. Encompassing, explanation, documentation, and disclosure, transparency is key to displaying the competence of machine learning models to experts in healthcare and criminal justice. I study how algorithmic transparency affects human decision-making."

In October 2020, Umang was appointed a Fellow at the Mozilla Foundation. Umang is one of two PhD students in a highly selective cohort of 29 individuals selected from 14 countries to work on building a more trustworthy Internet. Mozilla Fellows are web activists, open-source researchers and scientists, engineers, and technology policy experts who work on the front lines of ensuring technology remains a force for good. As a Mozilla Fellow, Umang has been understanding the transparency needs of stakeholders affected by AI systems and developing methods to generate diverse example-based explanations from AI systems. By coupling advocacy at a civil society like Mozilla with the technical nature of a PhD, Umang can translate how algorithms for creating safe AI systems implicate practical considerations for policy makers and practitioners. 

Before joining Mozilla, Umang spent a year and a half as a Research Fellow at the Partnership on AI, a nonprofit committed to building responsible artificial intelligence by convening government, civil society, academia, and industry. Umang led their transparency initiative, which resulted in a timely study of how explainable AI tools are used by practitioners. Before coming to Cambridge, he completed a joint bachelors-masters in Electrical and Computer Engineering at Carnegie Mellon University. His PhD research is funded by the Leverhulme Centre for the Future of Intelligence (Trust and Transparency Initiative) with generous donations from DeepMind and Leverhulme Trust.