History Surfacing Lab
Release mode Alpha
- Level
- Beginner / Intermediate
- Format
- To be determined
The future exercise will compare a chatbot's first broad historical answer with later answers produced after learners widen the frame, ask about missing communities or regions, and consult a reviewed source packet. Learners will examine which events were surfaced, which were omitted, and how the question and evidence changed the resulting account.
Status: Idea / Seed. This is not yet a learner-ready Draft package.
The future exercise will compare a chatbot’s first broad historical answer with later answers produced after learners widen the frame, ask about missing communities or regions, and consult a reviewed source packet. Learners will examine which events were surfaced, which were omitted, and how the question and evidence changed the resulting account.
The 1921 Tulsa Race Massacre is one candidate example. It must be supported by reviewed historical sources and placed within a broader, inclusive case set before the lab is drafted.
What you will make
The future exercise will compare a chatbot’s first broad historical answer with later answers produced after learners widen the frame, ask about missing communities or regions, and consult a reviewed source packet. Learners will examine which events were surfaced, which were omitted, and how the question and evidence changed the resulting account.
Candidate Activity Flow
- Review a saved first response to a broad historical question.
- Map the events, people, regions, and perspectives that appear.
- Identify plausible gaps without assuming why the model omitted them.
- Ask one or more frame-widening follow-up questions.
- Compare the responses with a reviewed source packet.
- Label each finding carefully: surfaced initially, surfaced after follow-up, supported by sources but not surfaced, disputed, or still uncertain.
What Needs to Be Designed Before Draft
- three to five historically reviewed cases with source locators;
- a rationale for the broad opening prompt and each frame-widening follow-up;
- a scoring approach that evaluates response coverage without claiming access to model internals;
- trauma-aware facilitation and an alternative participation path;
- safeguards against turning historically marginalized communities into quiz answers; and
- an accessible no-tool version using saved responses and printed sources.
The lab should teach a durable habit: the first answer is a selection, not a complete map of history.