Towards Human-Centered AI Alignment via Interaction: Human-in-the-loop AI Explanation, Evaluation, and Feedback

PRESIDENT’S POSTDOC FELLOWSHIP PROGRAM CANDIDATE

Zoom link

Abstract:

AI has advanced significantly, showcasing previously unheard-of capabilities in solving a wide spectrum of tasks in our daily lives. Despite the incredible progress, the reliance on AI models to consistently align with human intentions remains uncertain. This misalignment poses significant challenges, impeding the performance of models and potentially leading to unforeseen outcomes such as social biases and even fatal accidents. In this talk, I will discuss human-centered AI alignment via interaction, beyond the widely discussed Reinforcement Learning from Human Feedback (RLHF). Specifically, I explore the integration of human-AI alignment throughout the AI lifecycle, encompassing stages like data collection, model construction, evaluation, and deployment. The approaches involve enhancing human understanding and interaction with AI systems through developing human-centered AI explanations and evaluation. Additionally, I will introduce innovative methods to gather and integrate human-centered feedback and metric into data collection and model training.

Speaker Bio:

Hua Shen is a postdoc research fellow at University of Michigan, advised by Dr. David Jurgens, working on facilitating human-centered AI explanation and interaction in the fields of NLP and HCI. She obtained her PhD at Penn State during 2019-2023, supervised by Kenneth Huang, and closely worked with Sherry Wu from CMU. Her research focused on helping humans explain and communicate with AI models (e.g., LLM) via interactions or conversations and, moreover, enhancing AI models by aligning their behavior with human feedback. Besides, she conducted research on Speech Processing in her Google AI and Amazon AI internships. Her papers have been recognized as the Best Demo at CSCW 2023, Best Paper Honorable Mention Award at IUI 2023, and Google Research Science Conference Scholarships. She was selected as one of the 2023 Rising Stars of Data Science, and invited to talk about her research at CMU, Princeton, Google, Amazon, etc. https://hua-shen.org/