Research path
Guided course for grad students and researchers. 8 lessons, about 5 h 50 min.
For graduate students and researchers who want AI to speed up their work without weakening it. The course covers how models reason and fail, then literature search, data analysis and grounded answers, with verification at every step.
By the end, you'll be able to
- Choose when a reasoning model helps and when it doesn't.
- Run an AI-assisted literature search and verify every source.
- Analyze data with AI while checking the code it runs.
- Ground AI answers in your own documents and spot when they drift.
- Disclose AI use in research in line with journal and university norms.
Unit 1: How models think, and fail
Research use demands an accurate picture of what models can and can't do. This unit covers how they work, what reasoning models add, and the failure modes to plan for.
- How large language models work: Tokens, transformers, training and why context windows matter.
- Reasoning models: Why some models “think” before answering, and when that's worth the wait.
- What AI gets wrong: Hallucinations, bias, sycophancy, and how to catch them.
Unit 2: Research workflows
The core of the course: literature, data and your own documents. Each lesson pairs a workflow with the checks that make it trustworthy.
- Research and verification: Use AI to find and read faster without losing rigor.
- Analyzing data with AI: Upload a spreadsheet, ask questions, check the code.
- RAG: AI over your own documents: How tools answer from your files instead of from memory.
Unit 3: Writing and integrity
Close with writing: where AI helps a researcher's drafts, and the disclosure and integrity norms that apply to publications.
- Writing with AI, not by AI: Brainstorm, outline, get feedback and edit, and keep your voice.
- Ethics and academic integrity: Privacy, copyright, disclosure and USC policy.
Capstone project: A verified literature map
Use AI tools to map the literature on a question from your field, and keep a verification log that shows every source was checked.