Computational Behavioral Sciences - Foundations of Data Science & Applied AI/ML for Behavioral Research

Welcome to “Computational Behavioral Sciences”, a 4-credit, research-based project that I lead at Harvard University. This project is offered as a course focusing on learning to conduct research that requires quantifying behavior in humans and other animals using automation, data science, and AI/ML methods using Python programming.

The course at a glance¶

From the precise movements of animals, or an athlete to the intricate decision-making in animals, humans, and organizations — each represents a form of behavior at a different scale. Understanding and predicting behavior is fundamental to advancing research and innovation across academic disciplines, industry, and businesses. For conducting any objective research, defining and quantifying the target behavior is essential. For such quantification, behavior is defined as a temporal sequence of discrete actions — whether movements, vocalizations, or other measurable signals — occurring in a given context. These quantitative measurements underpin advances in neuroscience, medical sciences, biomechanics, biophysics, biomimetics, ethology and ecology, as well as transformative applications in business, sports, performing arts, and digital media.

Highlights of the course

6 modules · ~27 notebooks · 7 reading guides ·from zero Python to data science and AI/ML workflow

Please click the “hamburger” menu button on the top left of this site to explore all modules and other contents. You can also search any topic through the search box.

What you’ll learn¶

The courser aims to enable learners to solve real-life problems and answer questions with a research framework. By the end you will be able to:

Formulate behavior-related questions as testable scientific hypotheses.
Collect and process video, audio, and other digital data.
Use and fine-tune open-source AI/ML models (for example, models used for pose estimation, object tracking, audio embeddings) to extract behavioral metrics.
Apply statistical analyses to validate results.
Train classical ML models (supervised and unsupervised) to predict on new data.
Communicate findings through visualizations, talks, and written reports.
Use large language models responsibly in research workflows.

No prior programming experience required — Module 00B is meant for learners with zero experience with Python programming.

Who this is for¶

Students enrolled in LS100, and anyone using computation to make sense of behavior.
Learners seeking a data-backed understanding of human or animal movement, vocalization, or behavioral data from sources like health portals and web services.
Instructors seeking non-commercial teaching resources in computational ethology and behavioral data science.
Expected time commitment: approximately 6–10 hours per week.

You can