Human and Machine Learning — Chiba Institute of Technology, School of Design & Science
A graduate seminar on computational cognitive science: the study of the mind as an information-processing system that performs inference, learning, and decision-making under uncertainty. The course covers Bayesian foundations, hierarchical and causal models, Markov chains and Monte Carlo methods, reinforcement learning and inverse reinforcement learning, Bayesian nonparametrics, and connections to contemporary machine learning (deep neural networks, ethics, adversarial ML). Throughout, we use the probabilistic programming language GenJAX to move fluidly between mathematical models and executable code.
The companion textbook is A Narrative Introduction to Probability, a free, open-source resource written alongside this course.
Students who successfully complete the course will be able to:
| Component | Weight | Notes |
|---|---|---|
| Final project | 50% | Proposal 5% · In-class presentation 7.5% · Final paper 37.5% |
| Programming assignments (4) | 30% | Clusters 7.5% · Generalization 7.5% · Monte Carlo 10.5% · RL 4.5%. All in GenJAX. |
| Weekly written reflections | 12.5% | ~200 words on one assigned reading, pre-class. 5 of 11 required (student's choice; Week 1 not eligible). Pass/fail each. |
| Paper presentation | 7.5% | One 20-minute presentation of an assigned reading, followed by ~15 minutes of discussion you facilitate. See the "Paper presentations" section below. |
| Quizzes | 0% | Self-check only, available via the textbook. |
| Total | 100% |
Assignment weight rationale: the Monte Carlo assignment is weighted highest because it is the most demanding; the reinforcement-learning assignment is lightest because it comes with the most scaffolding code.
With a small seminar class, attendance is self-evident and the paper presentation serves as the visible engagement signal — so participation is folded into those two components rather than graded separately.
The project is the capstone of the course and the largest single grade component. It is an opportunity to explore a question related to human and machine learning that the student is personally interested in. The project does not need to use a specific topic or technique from class.
Three (non-exclusive) categories of project:
Project deliverables (see the project guidelines for full details):
Each week, students write a ~200-word reaction to one of the assigned readings. Reflections are not summaries — they are thoughtful engagements (a point you find compelling, a doubt, a connection to other work). Submit before class. Students choose any 5 of the 11 eligible weeks (Weeks 2–12; Week 1 is the intro session and has no reflection). Pass/fail each.
Each student gives one 20-minute presentation of an assigned reading, followed by 10–15 minutes of discussion that the presenter facilitates. Presentations are slotted across Weeks 4–12; signup happens in Week 2 class.
The focus of the presentation is how the mathematical model connects to cognitive science, and the evidence the authors provide for that connection. Plan to meet with Joe in office hours at least one week before your slot.
Grading (out of 7.5 points — matches the 7.5% course weight):
| Criterion | Points |
|---|---|
| Understanding of the paper | 2.25 |
| Covering key aspects of the paper | 2.25 |
| Presentation clarity | 1.50 |
| Appropriate discussion questions (at least 3) | 0.75 |
| Appropriate use of time | 0.75 |
Useful framing questions (adapted from Tom Griffiths — use these to organize your talk):
See the presentation guidelines for the full rubric and preparation tips.
You are welcome to use AI tools (ChatGPT, Claude, Copilot, etc.) as a resource for technical problems — debugging, looking up syntax, understanding a new concept. The real bar is understanding, not provenance:
All submitted work will be checked. Do not plagiarize. Self-plagiarism (reusing your own work from another class without explicit permission) also counts. Plagiarism cases are handled individually.
I do my best to respond within 36 hours (48 on weekends). Please keep emails clear and concise.
We are still deciding on whether there will be explicit office hours. If you would like to meet, please email Joe to schedule a 1-1 meeting.
course/readings_map.yml and is published to the course website as it is finalized week by week.This syllabus is provisional and may be adjusted as the semester progresses. Any changes will be announced in class and reflected on the course website.