Sketchbook Tags

A Panel of Specialist Readers

ExperimentAsk AI to review a draft article separately as each of the specific readers it will actually face — subfield expert, adjacent specialist, volume editor, fellow contributor.

• four distinct reviews of the same draft, each from a named vantage point
• insights from every perspective that I had not considered
• substantive revisions to the article

AI Integration Ladder

Key questionHow can a course distinguish between different levels of acceptable AI use and make those levels usable on assignments?

• AI use is not binary, but depends on the task
• different levels of AI use require different forms of accountability
• assignment labels are clearest when they say what AI may do, what it may not replace, and what students should make visible

AI Reading Investigation

Key questionHow can AI make a difficult reading more investigable without doing the reading for you?

ActivityStudents work through a dense scholarly article with a structured prompt sequence, then report what AI helped them see, what they verified, and what they had to correct.

• that a generic summary is the least useful thing AI can do with a reading
• how to use AI to generate questions and interpretations, then test them against the text
• where AI flattens the texture of a scholarly argument

AI Use Notes

Key questionHow can students disclose AI use in a way that supports learning?

• AI use can be documented as part of process
• disclosure should distinguish assistance from substitution
• students are accountable for accepting, rejecting, and revising AI output

AI as Second Opinion

Key questionWhat does AI see in a primary source, and what does it smooth over?

ActivityStudents form their own first impression of old primary sources, ask AI for its take, then go back to the originals to find where AI's version was right, flattened, or wrong.

• that tone, voice, audience, and style carry much of a primary source's meaning
• that AI is good at the gist of a text and weak on its texture
• that a first impression of their own is worth having before asking anyone else
• how to sample a long historical text without reading all of it

Argument Audit

Key questionHow can AI help sharpen writing skills instead of replace them?

ActivityStudents use AI-generated objections to test whether a thesis is vague, vulnerable, or persuasive.

• the difference between tone and analytical precision
• what makes an objection substantive vs. generic
• how vague writing produces vague critique

Citation Test

Key questionHow can AI output help students learn scholarly integrity?

ActivityStudents verify an AI-generated reading list and discover how convincingly LLMs invent sources that sound real but do not exist.

• why polished prose is not evidence of accuracy
• how hallucination happens and why it's convincing
• verification is a scholarly habit that connects classroom work with library expertise

Complicate the Obvious

Key questionWhat does a confident, ordinary answer take for granted?

ActivityStudents ask AI a question with a boringly familiar answer — what is a healthy diet? — then use the course to explain why that answer is neither timeless nor neutral.

• that practical advice carries a history and a set of assumptions
• how authority gets constructed in a voice that sounds neutral
• what a well-designed follow-up prompt can surface that a first answer hides

Design the AI Assignment

Key questionWhat does an assignment look like that uses AI to learn, not just to get an answer?

ActivityStudents design an AI-assisted learning exercise on a concept they once struggled with, then trade assignments and find out whether someone else actually learns from it.

• the difference between using AI to finish work and using AI to understand something
• that evidence of learning has to be designed in, not assumed
• how much a learning process depends on clear goals and iteration
• how to write instructions clear enough for someone else to follow

Differentiate Yourself From AI

Key questionWhat if the focus is on the product, not the process?

• AI use does not remove responsibility for the final work
• generic AI-like work may not provide enough evidence of learning
• students are expected to go beyond what AI can produce for free

Disruptive Expertise

Key questionWhat can earlier panics about new information technologies tell us about the debate over AI?

ActivityEach student researches one moment when a new technology upended how people made, shared, or trusted information — with heavy AI help and verified sources — and the class publishes the case studies together as a public website.

• how to orient quickly to an unfamiliar period, technology, and society
• that AI is a useful research assistant and an unreliable authority
• that arguments about new technologies repeat, with differences that matter
• how to write public-facing history for a general reader

Effect of Model Tiers on LLM Responses

ExperimentThis sketch shows the difference in response quality baseed on the LLM model tier.

• hallucinations occur far less frequently with higher tier models

Generate 3D Prints from 2D Drawings

ExperimentAI can transform a historical line drawing into a 3D-printable file, adding a tactile dimension to research that images alone can't provide.

• generated 3D-printable files from 2D historical images
• reconstructed material culture objects for research
• incorporated tactile elements into research presentations

Historians' Café

Key questionCan AI represent a school of thought, or only its vocabulary?

ActivityStudents script an argument among three historians from different schools of thought, then judge whether AI captured real methodological differences or just swapped labels.

• how methodological assumptions produce different readings of the same evidence
• what a caricature of an intellectual position looks like next to the real thing
• that the quality of AI output depends on how well they already understand the material

How Else Could This Look?

Key questionWhat did this article decide to be about, and what did that decision cost?

ActivityStudents read a long encyclopedia entry on fast food with AI, then spend class figuring out what the article decided to be about — and asking AI to draft the versions that were never written.

• that a one-clause mention is how a text claims a subject without thinking about it
• how an article's organization is an argument about what the subject is
• that AI will name silences fluently and generically if you let it do the noticing

Initiate Research with an Annotated Bibliography

ExperimentThis sketch demonstrates how to construct a prompt for a high quality annotated bibliography.

• a sophisticated and reliable annotated bibliography to initiate research

Narrate the Slides

Key questionWhat does a smooth narrative leave out of the images it strings together?

ActivityAI writes a caption linking each image in an unnarrated slide deck to the next; students go back through the slides to find what the course lets them see that the AI story missed.

• that images carry historical detail a connecting narrative tends to skip
• that AI imposes a tidy story on a sequence whether or not the sources support one
• how much course knowledge they bring to looking at a source

Photos to Map Pins

ExperimentCreate an interactive map with pins for hundreds of photos, using GPS metadata already embedded in your phone's images — in under an hour.

• extracted GPS metadata from image files
• built a map visualization with AI-assisted coding
• presented geolocated data in a public-facing format

Pipelines for Medieval Handwriting Recognition

ExperimentTo create an AI agent to work with Gemini and Claude to bulk process 300 images of archival documents and enable full-text search of medieval handwriting.

• built an agentic pipeline for bulk document processing
• combined multiple LLMs to improve transcription accuracy
• enabled full-text search of handwritten archival sources

Remixing Plato

Key questionIf Plato's worries about writing were recast as worries about AI, what would change and what would stay the same?

ActivityStudents remix Plato's worries about writing into a new dialogue about AI — building the characters themselves, then iterating with AI until each position is sharp.

• what AI can and cannot preserve in philosophical argument
• how old anxieties about new media resemble current debates about AI
• how form and genre reshape meaning
• that prompting requires the same clarity as writing

Same Prompt, Different History

Key questionDoes the same prompt give everyone the same history?

ActivityThe same prompt to ChatGPT produces different histories depending on whether you're logged in or not — and that difference is the lesson.

• how context shapes historical interpretation
• what filter bubbles look like in practice
• the difference between pronouncing and puzzling about sources

Start With the AI Answer

Key questionIf you can't see how students work, what if the assignment starts where AI would have?

• that an AI answer is a starting point to interrogate, not a finished product
• which kinds of thinking AI can help with and which have to be their own
• how to check a confident answer against the actual sources

Table of Contents to Zotero

ExperimentWith one prompt get entries in Zotero from a book of essay chapters.

• extracted data from the Table of Contents
• built a file in BibTeX format
• imported to Zotero

Take the AI Draft to the Archive

Key questionWhat does the archive know that the model doesn't — and why can't it?

ActivityStudents have AI write a short history of a narrow topic, critique it, then test it against unpublished material in a physical collection — and write about what the model could never have known.

• that fluent, specific-sounding history can rest on nothing checkable
• that most of the historical record has never been digitized, and so is invisible to AI
• that archives are shaped by decisions about what was worth keeping, just as AI output is
• how to find and work with unpublished material in a physical collection

What Does Cilantro Taste Like?

Key questionHow to introduce students to the basics of AI output differences?

ActivityA hands-on demo to show how model size and settings change what AI says — using one simple, relatable question.

• AI is a spectrum of models, not one fixed thing
• how temperature, token limits, and sampling shape output
• what training data has to do with what a model knows