Showing versus Doing: Teaching by Demonstration
People often learn from others' demonstrations, and inverse reinforcement learning (IRL) techniques have realized this capacity in machines. In contrast, teaching by demonstration has been less well studied computationally. Here, we develop a Bayesian model for teaching by demonstration.
This paper presents a Bayesian model for teaching by demonstration, distinguishing between intentional teaching (showing) and task performance (doing) in machine learning contexts.
What happened
Researchers created a Bayesian model to analyze how humans modify teaching behavior when intentionally demonstrating tasks versus performing them. Experiments with human participants confirmed the model's predictions, and standard inverse reinforcement learning (IRL) algorithms showed improved performance when trained on demonstration data.
Why it matters
The findings advance understanding of human-AI interaction by clarifying how teaching strategies affect machine learning outcomes, particularly in scenarios requiring human guidance.
- Intentional teaching strategies significantly influence AI learning outcomes
- Bayesian models can effectively capture human teaching behavior
- Demonstration-based training outperforms passive task execution for IRL algorithms
The bigger picture
The research provides a framework for improving AI training through structured human demonstrations, with implications for interactive machine learning systems and human-in-the-loop AI development.
The study establishes a computational framework for analyzing teaching-by-demonstration, demonstrating that intentional instruction methods improve both human-AI collaboration and machine learning performance.
Technical details
- Bayesian model for teaching-by-demonstration
- Inverse reinforcement learning (IRL) algorithms
- Experimental setup with human participants
- Comparison between 'showing' and 'doing' task execution modes
FAQ
What is the key distinction between 'showing' and 'doing' in AI teaching?
Showing refers to intentional teaching behavior where demonstrators structure their actions to convey knowledge, while doing involves performing a task without explicit instructional intent.
How does the Bayesian model contribute to teaching-by-demonstration research?
The model provides a computational framework to predict and analyze how demonstrators adjust their behavior when intentionally teaching versus performing tasks.
What evidence supports the effectiveness of demonstration-based learning?
Experiments showed human participants modified their teaching behavior according to the model's predictions, and standard IRL algorithms achieved better results with demonstration data.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by arXiv cs.LG on Oct 14, 2026.
Comments
Sign in to join the discussion.