← Back to series HSCI 488

Directed Studies on Artificial Intelligence in Health Sciences

Office Hours, the companion podcast for HSCI 488

Each episode is a conversation between Sarah and Kiffer about one lesson of the course. Every episode page has the audio with a visual companion and captions, a short summary, a list of chapters and a downloadable transcript.

How AI works

Lessons 01–02

AI Literacy as a Public Health Skill: Machine Learning and Neural NetworksSarah and Kiffer work through accuracy traps, data leakage and a gradient descent step, and debate whether a health team should let a prediction model guide outreach.

How Large Language Models and AI Agents WorkSarah and Kiffer work through what next-token probabilities mean, why instructions cannot stop fabrication, how retrieval misreports real sources, and where people belong in an agent's work.

Working with AI

Lessons 03–04

Working with AI: Prompting, Context and VerificationSarah and Kiffer probe known-answer tests, repair a leading prompt, catch claims whose numbers match the source while the meaning shifts, and debate checking during study.

AI for Learning and StudySarah and Kiffer ask whether assisted practice means learning, size up the Harvard AI tutoring trial, catch a wrong quiz key and debate when to ask for hints.

AI in academic work

Lessons 05–08

AI-Assisted Literature Searching and ReadingSarah and Kiffer work through recall, testing AI-drafted searches, checking AI extraction and auditing citations, with extra worked examples and a debate about deep research reports.

AI-Assisted Academic and Professional WritingSarah and Kiffer test what the writing experiments show, why an assistant's self-check falls short, what a detector flag means and how readability scores mislead.

AI as a Thinking Partner: Generating and Testing IdeasSarah and Kiffer test AI-generated ideas and critiques, work a harder regression-to-the-mean problem, and debate using simulated perspectives to prepare for engagement with interest holders.

AI for Data Analysis and Coding in RSarah and Kiffer work through silent errors in AI-written R code, independent checks, forking paths and data governance, with new examples built on the Canadian Social Connection Survey.

AI in public health practice

Lessons 09–10

AI in Public Health PracticeSarah and Kiffer estimate a news filter's recall, follow a risk score into a new population, weigh evidence on health chatbots, and check an assistant's survey coding with kappa.

Responsible Adoption and Continued LearningThe final episode works through small counts and community data, approvals for new AI features, energy estimates for images, personal test sets and the risk of deskilling.