HCC: Responsible AI Guidance for Visual Data Analysis

Artificial intelligence (AI) tools are being rapidly integrated into the software that people use to explore and make sense of data. These tools can suggest charts, highlight patterns, and even guide users toward conclusions automatically. While such assistance can be helpful, it is not yet well understood when AI guidance improves decision-making versus when it leads people to over-rely on automated suggestions and potentially miss important insights. This project addresses that gap by studying how different forms of AI guidance affect the way people analyze data and draw conclusions, with the goal of ensuring that future AI-powered visual data tools are designed responsibly. The findings will benefit anyone who uses visual data analysis software in fields ranging from business and healthcare to national security, and will help establish design practices that keep humans meaningfully in control even as AI capabilities continue to advance. This project pursues four interconnected research activities. First, the research team will develop a software testbed with modular AI guidance components focused on chart creation and visual analysis of tabular data. Second, the team will develop computational metrics capable of measuring, in real time and after the fact, how much a person relies on AI guidance while performing analysis tasks, enabling detection of both under-reliance and over-reliance. Third, a series of controlled user studies will be conducted using this testbed to determine which types of visual analytics tasks are most and least suited to AI guidance, and to establish baselines for appropriate guidance utilization. Fourth, the team will design and evaluate user interface components that use these metrics to detect suboptimal guidance usage and prompt users toward more effective engagement. The testbed and associated code will be released as open-source resources for the broader research and development communities. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.