Mazda Moayeri – ARCS-MWC Endowment Scholar
1st Year Scholar, PhD Candidate, Computer Science
University of Maryland

Research: 
I research the reliability of cutting-edge artificial intelligence (AI), with an emphasis on making models more interpretable and robust to deployment settings. More specifically, I develop tools for understanding autonomous reasoning, discovering hidden biases, and debugging machine-learning models, centered around improving the trustworthiness and reliability of AI. 

Describe the expected benefit of your research to society: 
As the field of AI continues to rapidly develop, serious reliability risks have emerged, many of which are rooted in the fact that modern deep networks lack interpretability (i.e. the way they make decisions is opaque) and robustness (they often fail dramatically when deployed to settings slightly different than the setting the model was trained in). My research focuses on ameliorating these reliability risks, so that humans can trust AI and AI can realize its incredible potential sweeping across countless domains. I wish to develop interpretability tools that allow humans to learn from how AIs reason, potentially uncovering new underlying insights in any scientific discipline. Further, understanding what AIs can and can’t do opens the door for human-AI collaborations to flourish, bringing forth new technologies and opportunities for all of society. My robust research aims to ensure these rapidly advancing technologies remain ethical and safe. Whether discovering hidden biases in AI or predicting failures before they occur, the development of safeguards and mitigation techniques will be essential to the proliferation of responsible AI for human good. In summary, AI is incredibly powerful and can benefit countless fields, from medicine to art to science to economics, but before that happens, we need to make AI interpretable and reliable enough to where people trust it, and that is precisely what I am setting out to do. 

Indicate how an ARCS award might benefit your research: 
I am most excited to join a network of young STEM researchers across disciplines. As AI has the potential to help nearly all fields, getting the opportunity to learn more about the hopes and professional needs of various scientists would offer inspiration for how I can make my research the most impactful. I also hope to meet more scientists from underrepresented groups. I appreciate how ARCS explicitly promotes diversity, equity, and inclusion, and I would hope to find like-minded allies or scientists from genders, races, and socioeconomic backgrounds that are minorities in STEM, a field that can often be exceedingly homogeneous. Sharing solidarity, learning from each other, and promoting one another would be a key benefit of the ARCS fellowship.