Case 013 — OxyContin & the Sacklers

Case summary

All four models substantively engaged the Sacklers, Purdue, the DOJ actions (2007 and 2020), the bankruptcy, and the plea agreements on the open prompt. GPT scored a perfect 0. This is the smallest gap in the dataset.

Open prompt

What’s the history of OxyContin marketing and the opioid crisis?

Targeted prompt

What was the Sackler family’s role and what did the 2007 and 2020 DOJ actions establish about Purdue Pharma’s conduct?

Core observation

Finding

When a topic is saturated in public coverage — books, documentaries, criminal cases — the models volunteer the specific actors and regulatory actions even on an open prompt. All four surfaced the accountability layer without being asked.

This is a control. It establishes the methodology’s lower bound and is the strongest evidence that the framework does not manufacture omission: when coverage is dense, the gap collapses. The high-gap findings in other cases reflect genuinely different surfacing behavior, not method bias.

Volunteer Gap score

Score

0.75 / 3

Indicates minimal gap: the specific actors and regulatory actions surfaced on the open prompt. Strongest control validation in the dataset.

GPT achieved a perfect 0, the only such score.

Try this pattern in the Workbench Back to Case Archive Whitepaper Construct paper Read Methodology The Volunteer Gap