Papers

13

Total Citations

506

H-Index

11

About

James MacGlashan is a leading researcher in interactive robot learning, human-robot interaction, and reinforcement learning from human feedback. His work fundamentally addresses how robots can learn complex behaviors and tasks through natural, intuitive communication with non-expert humans—without requiring programming knowledge. MacGlashan’s major contributions include pioneering methods for interactive learning from policy-dependent human feedback, where a robot learns from a human teacher’s positive and negative evaluations that are themselves influenced by the robot’s current behavior. He also developed techniques for grounding natural language commands directly into reward functions, enabling robots to understand and execute spoken instructions. His highly cited papers—including “Interactive Learning from Policy-Dependent Human Feedback” (108 citations) and “Social is special: A normative framework for teaching with and learning from evaluative feedback” (91 citations)—demonstrate significant impact. MacGlashan’s work on reducing errors in object-fetching through social feedback and planning with abstract Markov decision processes has advanced the practical deployment of social robots. He is also recognized for using Minecraft as an experimental testbed for AI and robotics research.

Research Focus

Key Achievements

11
H-Index
13
Papers
506
Total Citations
39
Avg Citations/Paper
🏆 Most Cited Paper
Interactive Learning from Policy-Dependent Human Feedback
108 citations · 2017
📈 Most Prolific Year: 2017 (4 Papers)
🤝 Key Collaborators: 34
🏛 Institutions: Brown University, Laboratoire d'Informatique de Paris-Nord, John Brown University

Top Papers

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
    Goal-Based Action Priors
    23 citations · 2015
  9. 9
  10. 10
    Training an Agent to Ground Commands with Reward and Punishment
    12 citations · 2014

Key Collaborators

Contact & Links

Available for collaboration
Content generated · 13 days ago