Public History and History Communication

Knowledge Communication along the Writing and Production Process

Exercise set for target audience-oriented communication of a scholarly article in multiple formats (blog, podcast, social media, poster, interview) – with audit trail, source veto, and methodological transparency.

Author

Affiliation

Moritz Mähr

University of Bern

Overview and Didactic Goal

This exercise trains public history knowledge communication as processual translation work: from a scholarly article (argument, evidence, uncertainties) to audience-oriented formats. Generative AI is used here not as an authority, but as a tool for drafting, variation, editing, and proofreading – with consistent source reference, traceability, and clear limits.

Prerequisites

• Basic understanding of historical research methods
• Basic knowledge of working with generative AI (especially prompting)

NotePrompt Engineering

If you are not yet familiar with prompting, we recommend completing the Prompt Engineering exercise first.

NoteLLM

You can complete this exercise with LLMs from different providers. For this exercise, it is helpful if the LLM has internet access and allows file uploads.

Learning Objectives

• Decompose a communication context into target audiences, medium, purpose, and constraints (communication briefing).
• Derive a core message (message map) and evidence matrix from a scholarly article without inventing new facts.
• Create drafts in multiple public history formats (blog, audio, social, poster) and revise them format-appropriately.
• Systematically verify responses (claim-by-claim, source veto, mark uncertainties) and document changes traceably (audit trail).
• Prepare an interview kit (soundbites, Q&A, no-go list, transparency statement).

The base text is a (preferably open-access) scholarly article from your subject area. E.g.:

• https://doi.org/10.1093/pastj/gtae018

Submission Formats (Minimum)

1. Communication briefing (1 page)
2. Message map + evidence matrix (table or bullet form)
3. Blog post (approx. 700–1,000 words) or podcast script (approx. 6–8 minutes)
4. Two short formats from: social media, poster text blocks, interview kit
5. AI protocol (prompt/response log + reflection note, approx. 300–500 words)

Methodological Framework: AI as Critically Reflected Communication Infrastructure

Terms (for this exercise):

• Audit Trail (AI Protocol): Documentation of prompts, responses, decisions, corrections, and source references, so that your process remains traceable.
• Source Veto: Any narrative sharpening is allowed, provided it does not contradict the statements documented in the article.
• Claim-by-claim Check: Statements are decomposed as verifiable claims, linked to evidence or marked as speculation.

Search engines and AI systems structure visibility in problematic ways.(Noble 2018) AI-generated content for public history applications carries particular risks, as it reaches a broad audience without professional expertise.(Bender et al. 2021)

ImportantBasic Rule: AI May Not “Invent”

Work with a source package: title/DOI, abstract, central results in your own notes, a few key points/evidence locations (pages/sections), possibly figures/tables. Give the AI only material that you can control yourself, and have the AI explicitly work only from that.

TipTransparency in the Product

Plan at the end a brief transparency element (e.g., “How this was created” / “AI assistance”): what AI was used for (structure/language variants/proofreading), what not (facts/interpretation decisions).

Structure of the Exercise

1. Situation Analysis: Context + target audience + purpose + constraints (briefing)
2. Planning: Core message + evidence + narrative route (message map)
3. Drafting: First draft (blog or podcast) with clean evidence logic
4. Revision: Fact/evidence check, comprehensibility, tone, ethics
5. Remix/Adaptation: Short forms (social/poster) + interview preparation
6. Documentation: AI protocol + reflection

1. Target Audience Analysis and Communication Briefing

Goal

Operationalize communication as a context-dependent decision: For whom, where, why, in which medium?

Task

Choose a specific communication context (e.g., museum blog, local newspaper, science podcast, archive’s Instagram, exhibition panel). Create a communication briefing with:

• Target audience(s): prior knowledge, interests, possible triggers/misunderstandings
• Purpose: inform / contextualize / open debate / action knowledge etc.
• Medium/format: length, tone, visuality/audio, interaction
• Constraints: language(s), accessibility, terminology, legal/ethical limits
• Success criteria: What should recipients know / be able to do / see differently afterward?

AI Workflow (Prompt Sketch)

You are an audience research assistant. Generate 2–3 plausible target audience personas for the following context (without inventing facts about real people).
For each persona: prior knowledge, motivations, possible misunderstandings, suitable examples/metaphors, no-gos.
Context: ...
Topic/article in 3 sentences: ...

Critical Comparison

• Persona ≠ reality: supplement/verify assumptions (e.g., with real target audience hints, museum concepts, media guidelines).
• Which simplifications would be epistemically risky? Note red lines.

2. Message Map and Narrative Route

Goal

Build a communicative core architecture that remains stable across formats.

Task

Create:

1. One-liner (1 sentence): What is the core message – and why does it matter for this target audience?
2. 3–4 support points (1–2 sentences each) with evidence reference
3. Evidence matrix: Which support points rely on which data/source locations?
4. Context/framing: What needs to be explained so that nothing is distorted?
5. Uncertainties: What is contested, limited, or only plausible, not proven?

AI Workflow (Prompt Sketch)

Role: You are a structurer. Work only with the source package + briefing.
Generate a message map (one-liner + 3–4 support points + 'so what' for the target audience).
Mark each point with claim IDs and indicate necessary context paragraphs.
Material: ...

Critical Comparison

• Where does “presentism” (present-day overlay) or moral simplification threaten?
• Where are terms/definitions needed instead of storytelling?

3. Draft I: Blog Post

Goal

Create a textual base model that serves as a reference for further adaptations.

Task

Write a blog post (approx. 700–1,000 words) with:

• An opening that matches target audience and purpose
• Clear section structure (subheadings)
• Explicit evidence logic (claim IDs or footnotes/references to evidence locations)
• “What we know / what we don’t know” (short section)
• Call-to-action/further reading (e.g., link to article, glossary, sources)

AI Workflow (Prompt Sketch)

Role: You are a blog editor. Write a blog post for the following context.
Constraints: Length 900 words, clear subheadings, no new facts.
Use claim IDs in the text in square brackets, e.g., [C3].
Material: Briefing + source package + message map + claim table.

Critical Comparison

• Check each paragraph: What claims are being made? Are they supported?
• Check tone and register: “Sounds plausible” is not a quality criterion.
• Check suitability: Which terms need to be explained?

4. Revision: Fact-Check, Comprehensibility, Ethics

Goal

Turn a plausible draft into a robust communication product.

Task (Checklist)

• Claim-by-claim: Every statement either supported, marked as framing, or deleted.
• Comprehensibility: Reduce or define unnecessary jargon; check examples.
• Ethics/responsibility: Sensitive groups/persons, violence, stigmatization, data policy, image rights.
• Transparency: Add brief note on method/source/AI assistance.

AI Workflow (Prompt Sketch)

Role: You are a fact-checker.
1) List all claimed facts/statements in the draft.
2) Assign them to claim IDs or mark 'not in source package'.
3) Suggest more precise formulations that correctly express uncertainties.
Input: Blog draft + source package.

Critical Comparison

• AI can identify gaps, but cannot reliably verify: verification remains your work.
• Document changes in the AI protocol (why changed? which evidence location?).

5. Remix: Format Adaptations

The following tasks are deliberately designed as adaptations: You recycle message map + baseline text instead of “inventing” anew in each medium.

5A Social Media (Thread + Single Post)

Goal: Translate core message into an attention medium without overclaiming.

Task

• 1 single post (max. 280–500 characters, depending on platform) + 1 thread (5–7 posts)
• 1 “context post”: terms/framing
• 1 “uncertainties post”: limits, open questions
• 1 link/source reference (format-appropriate)

AI Workflow

Generate 3 variants (conservative / narrative / provocation-averse) for a thread.
Rules: no new facts, no clickbait exaggeration, include an uncertainties post.
Input: Message map + blog text excerpt + context.

5B Podcast Script (6–8 Minutes)

Goal: Communicate argument and context in audio format (voice, rhythm, examples).

Task

• Script with segments (hook, context, 2–3 key points, framing, conclusion)
• Speakable language (short sentences, active verbs)
• Show notes: 5 bullet points + source reference + glossary (3 terms)

AI Workflow

Write a podcast script (7 minutes) for [target audience/context].
Structure: Hook (15s) – Context (60s) – Key points (3x90s) – Uncertainties (45s) – Conclusion (30s).
Use only claims from the source package and mark claim IDs in the margin of the manuscript.

5C Poster Text Blocks (A1/A0, Exhibition or Conference)

Goal: Visual hierarchy + concise texts that work even without prior knowledge.

Task

• Headline (max. 10 words), subheadline (max. 20 words)
• 3 text blocks (max. 60 words each): “What’s it about?”, “What’s new?”, “Why relevant?”
• Image/graphic caption (max. 30 words) + source reference
• QR text (max. 15 words) + link target

AI Workflow

Generate poster text blocks with strict word limits and no new facts.
Pay attention to: clear terms, no implicit value judgments, claim IDs per block.
Input: Message map + evidence matrix.

6. Interview Kit

Goal

Speak about research confidently, precisely, and with minimized risk – including “no-go” control.

Task

Create an interview kit (1–2 pages):

• 3 soundbites (12–20 seconds each, paraphrased)
• 5 core questions + short answers (3–5 sentences each)
• 2 critical questions (e.g., “Isn’t that…?”) + bridging strategies
• No-go list: Statements you won’t make (due to evidence/complexity/ethics)
• Transparency statement on method and data situation (1–2 sentences)

AI Workflow (Prompt Sketch)

Role: You are a media coach. Generate an interview kit for radio/newspaper.
Rules: no new facts, soundbites must correctly convey claims, don't hide uncertainties.
Input: Message map + source package + blog/podcast draft.

Critical Comparison

• Do soundbites match claim IDs or “slip” into exaggeration?
• Are bridging strategies content-honest (no evasion without framing)?

Further Resources

• AI Pedagogy Project: What is AI? – Conceptual introduction for non-technical audiences.
• Critical AI Literacy: Curated Resources – Ethics, power, labor, and politics of AI.

Bibliography

Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜.” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21), 610–23. New York, NY, USA: Association for Computing Machinery. https://doi.org/10.1145/3442188.3445922.

Noble, Safiya Umoja. 2018. Algorithms of Oppression: How Search Engines Reinforce Racism. New York: NYU Press.

WarningAutomated Translation Disclaimer

This exercise was automatically translated from German using AI and may contain errors or inaccuracies. Please refer to the original German version for the authoritative text. If you notice any translation issues, please report them.