All videos
0:00 / 0:00
research

Reward Design and Evaluation in Reinforcement Learning, Calarina Muslimani

Amii Intelligence16 June 2026Watch on YouTube

Description

The AI Seminar is a weekly meeting at the University of Alberta where researchers interested in artificial intelligence (AI) can share their research. Presenters include both local speakers from the University of Alberta and visitors from other institutions. Topics can be related in any way to artificial intelligence, from foundational theoretical work to innovative applications of AI techniques to new fields and problems. In this seminar from the Alberta Machine Intelligence Institute and the Department of Computing Science, Calarina Muslimani, PhD Student at the University of Alberta, discusses the challenges of reward design in RL presents approaches to simplify this task for RL practitioners. Bio: Calarina (Callie) Muslimani is a fourth-year PhD student at the University of Alberta in the Reinforcement Learning and Artificial Intelligence (RLAI) Lab, advised by Matthew E. Taylor. Her research focuses on designing human-aligned reward functions for reinforcement learning, including developing metrics to evaluate reward functions and creating reward learning algorithms.

What you'll learn

  • Reward design is a critical challenge in reinforcement learning, as it directly determines how AI systems behave
  • Human alignment means reward functions must reflect human values and objectives, not just technical optimization goals
  • Evaluation metrics are needed to measure whether reward functions actually produce the desired behavior in RL agents

Frequently asked questions

What is reward design and why is it important in reinforcement learning?
Reward design involves creating reward functions that indicate what behavior an RL agent should learn. It is important because the reward function directly guides how an agent behaves and learns, making it crucial for achieving desired outcomes.
How does human alignment relate to reward design?
Human alignment focuses on ensuring reward functions align with human values and objectives. Muslimani's research addresses simplifying the design of reward functions that genuinely reflect what people want AI systems to do.
What role do evaluation metrics play in reward design?
Evaluation metrics determine whether reward functions actually produce the desired behavior in RL agents. They are essential for verifying that designed reward functions work as intended.
What is the practical relevance of this research for RL professionals?
Muslimani's research focuses on simplifying the reward design process. This can make it easier for RL practitioners to design better, human-aligned reward functions.

Topics

reinforcement learningreward designhuman alignmentrl algorithms