Human-Centered Artificial Intelligence (AI) Research
Explores how to evaluate and align artificial intelligence (AI) systems to serve human and community needs. Integrating human-centered design, AI evaluation, and alignment research, students examine how human goals translate into system behavior, measure capabilities, and assess context-specific performance. Key topics include benchmarks, human evaluation, red teaming, robustness, fairness, transparency, oversight, and AI agent design. Through research paper readings, evaluation exercises, and a hands-on project, students analyze existing AI systems, gather empirical evidence, test for failures and harms, and communicate findings and limitations. Building or training models is not required or covered in course.
Topics
- Define a human-centered AI research question that accounts for the goals and perspectives of relevant stakeholders. Explain major approaches to AI alignment and distinguish intended behavior from observed behavior in a specific use context. Design evaluations using appropriate tasks, measures, comparison conditions, and human judgment; explain what each measure misses. Investigate AI capabilities, failure modes, and potential harms through systematic testing and analysis. Use HCI methods, where appropriate, to understand how people use, interpret, and are affected by AI systems. Critically assess the validity and limitations of published AI evaluation and alignment research. Conduct and communicate a small-scale, ethically planned evaluation of an AI system, including findings, limitations, and recommendations.
Official catalogue course objectives and student learning outcomes
Sources
- Course catalogue ↗2026-2027