History & AI

Resources and strategies for integrating artificial intelligence into history education, research, and public engagement.

View the Project on GitHub AMSUCF/HistoryAndAIDemo

Prompt Engineering for History

The quality of AI output depends heavily on how you ask. This page covers strategies for writing effective prompts when using large language models for historical research, teaching, and writing.


Core Principles

1. Be Specific About Time, Place, and Scope

Weak: “Tell me about slavery.”

Better: “Describe the daily labor routines of enslaved people on cotton plantations in the Mississippi Delta between 1830 and 1860, based on primary source accounts such as slave narratives and plantation records.”

Specificity reduces the chance of generic, decontextualized responses.

2. Specify the Audience and Register

Weak: “Explain the Reformation.”

Better: “Write a 300-word summary of the causes of the Protestant Reformation suitable for an undergraduate survey course. Focus on theological, political, and economic factors. Avoid jargon.”

Telling the AI who the audience is shapes the level of detail and tone.

3. Request Evidence and Sources

Weak: “What caused World War I?”

Better: “What were the major historiographic interpretations of the causes of World War I? For each interpretation, name a key historian and their major work.”

Note: AI may fabricate citations. Always verify that cited works exist and say what the AI claims.

4. Ask for Multiple Perspectives

Weak: “What happened during the Partition of India?”

Better: “Describe the Partition of India in 1947 from three perspectives: a Hindu family displaced from Lahore, a Muslim family displaced from Delhi, and a British colonial administrator. Note the limitations of each perspective.”

This pushes the AI beyond a single dominant narrative.

5. Use Role and Persona Prompts

“You are a historian specializing in early modern Japan. A graduate student asks you to explain the significance of the Tokugawa shogunate’s sakoku policy. Provide a nuanced answer that acknowledges scholarly debate.”

Role prompts help the AI adopt a more disciplinary voice, though they do not guarantee accuracy.


Prompt Templates for Common Tasks

Summarizing a Primary Source

I am going to provide a primary source document. Please:
1. Summarize the main argument or purpose of the document in 2-3 sentences.
2. Identify the author's perspective and potential biases.
3. List 3 questions a historian might ask about this source.
4. Note any claims that would need verification against other sources.

[Paste or describe the document here]

Generating Discussion Questions

I am teaching an undergraduate course on [topic]. Students have just read
[specific reading]. Generate 5 discussion questions that:
- Move beyond factual recall
- Encourage students to connect the reading to broader course themes
- Include at least one question about methodology or sources
- Include at least one question that invites comparison across time or place

Historiographic Overview

Provide an overview of how historians have interpreted [topic/event] from
the mid-20th century to the present. For each major school of thought:
- Name 1-2 representative historians and key works
- Summarize their central argument
- Note major critiques of that interpretation

Flag any works you are uncertain about so I can verify them.

Translating and Contextualizing a Source

Translate the following passage from [language] to English. Then:
1. Identify the likely date, author, and genre of the text.
2. Explain any culturally specific terms or references.
3. Note where translation choices might affect interpretation.

[Paste source text here]

Lesson Plan Drafting

Draft a 75-minute lesson plan for an undergraduate history course on [topic].
The lesson should include:
- A brief lecture segment (15 min) on [specific focus]
- A primary source analysis activity (25 min)
- A small-group discussion (20 min)
- A wrap-up reflection (15 min)
Assume students have read [assigned reading] before class.

Common Pitfalls

Pitfall Example Fix
Too vague “Tell me about colonialism” Specify region, period, and analytical lens
Assumed accuracy Trusting AI-generated dates and citations Always verify factual claims independently
Single-query reliance Using one prompt for a complex topic Break complex questions into sequential prompts
Ignoring positionality “What really happened at…” Ask for multiple perspectives and name whose account is centered
No output constraints Letting the AI write unlimited text Set word limits, format requirements, and audience level

Iterative Prompting: A Worked Example

Round 1: “What was the significance of the Haitian Revolution?”

AI gives a general overview. Useful but generic.

Round 2: “How did the Haitian Revolution influence abolitionist movements in the British Caribbean between 1800 and 1838? Focus on specific events and figures.”

AI provides more targeted information. Some claims need checking.

Round 3: “You mentioned the influence on the 1831 Jamaica revolt. What primary sources document connections between Haitian revolutionary ideas and the leadership of Samuel Sharpe? Flag any sources you are uncertain about.”

AI narrows further and flags uncertainty. The historian can now pursue specific archival leads.

This iterative approach treats AI as a research interlocutor, not an oracle.


← Back to Home