We interact with text-based AI systems (e.g., Google autocomplete predictions and Gmail’s smart compose) dozens of times a day. Behind these systems, however, are models that are primarily trained and evaluated in isolation, without much consideration for human-AI interaction. This gap raises a question at the heart of Mina's research: how can we develop models that can interact with users and eventually augment their capabilities?
Mina Lee is a final-year Ph.D. candidate at Stanford University, who leverages language models to build writing assistants, including an autocomplete system, a contextual thesaurus system, and a general-purpose text editor with suggestions. Mina identifies and measures how these models change the way we write or help us perform a task. For instance, they can help us write faster, by autocompleting sentences, and write better, by suggesting alternative words, phrases, or ideas. Her work has been published in top-tier venues in natural language processing (e.g., ACL and NAACL), machine learning (e.g., NeurIPS), and human-computer interaction (e.g., CHI), and featured in various media outlets including “The Economist.”
Mina envisions language models becoming more personalized and controllable in the future, offering more relevant and targeted assistance. Ultimately, she believes that AI will transform the way we write. She plans to continue her research in this area to ensure that these models are developed in a way that enhances our productivity and creativity rather than merely mimicking human capabilities or automating them.