27 · AI in research: ethics

Tuesday, Dec 1, 2026

Materials for this session are not published yet. They appear here before class.

Objectives

By the end of this session you can:

• State the disclosure norms and apply them to your own work.
• Treat references as unverified by default.
• Say where AI use gets documented in a paper.
• Articulate what, in AI-assisted work, is the researcher’s own contribution.

What we cover

• Discussion-led, and nothing load-bearing. It is the last session before the wrap.
• Twelve weeks of your own disclosure blocks, reviewed: what patterns emerged?
• Hallucinated citations, with a live check: verify references by looking them up in CrossRef — a model can only check something by retrieving it from outside itself, never by generating. Two true stories: the lawyers sanctioned for filing six invented cases (Mata v. Avianca, 2023), and a PhD student expelled over an AI allegation (Minnesota, 2025).
• AEA Data Editor and journal policies; authorship.
• What stays yours, concretely: AI may draft intermediate reports whose numbers you verify — but the prose that enters the manuscript is written by you, because the byline carries the responsibility. That is the reasoning inside every journal policy we read.

Verification habit. References are wrong until verified.