research overview
- Artificial Intelligence (AI)-assisted software solutions have made substantial inroads into critical aspects of modern life, where they routinely make safety-critical, socio-critical, and legal-critical decisions with certainty and speed. Examples of such AI-assisted decisions include self-driving cars deciding to stop, implantable pacemakers determining when to pace, and the COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) software assessing whether individuals are likely to reoffend. These AI-assisted systems are data-driven: they adapt their behavior based on experiences in the form of data—whether expertly curated datasets in supervised learning, hidden patterns in raw data uncovered through unsupervised learning, or self-generated data guided by expertly designed reward signals in reinforcement learning. Dr. Ashutosh Trivedi's research focuses on enabling rigorous system engineering, specifically through formal methods, to enhance the safety, security, fairness, and accountability of data-driven systems.
