Labs / Chapter 5
Chapter 5 argues that automated map generation amplifies ethical responsibility. In this lab, you will create a publishability packet for an LLM-assisted map concept: provenance, prompt metadata, attribution, bias and privacy checks, and a short public disclosure.
Chapter 5 treats ethics as craft hygiene: ordinary habits that keep elegant maps from becoming careless or harmful. Your goal is not to make a finished map. Your goal is to decide what evidence, attribution, restraint, and disclosure a map would need before it deserves to be public.
Choose a map idea that could help people but could also mislead, expose, stereotype, or overstate if handled carelessly.
Map idea: Audience: Purpose: Potential benefit: Potential harm if handled badly: One dataset or data type I imagine using: One design style or inspiration I might be tempted to borrow:
Provenance in prompt cartography extends beyond dataset origin. It includes model use, prompt history, design choices, and publication decisions.
Act as a provenance reviewer for an LLM-assisted web map. Using the map idea below, create a chain-of-custody checklist with these columns: - item to document - why it matters - evidence I should preserve - risk if missing Include dataset origin, license, last update, transformations, model name, prompt archive, classification method, style influences, human review, and publication date. Map idea: [PASTE SETUP NOTES]
Prompt metadata makes a map reproducible enough to critique later. It also makes your directorial decisions visible.
Create a prompt metadata template for my map project. Use plain JSON-like fields, but keep it readable for a human reviewer. Include: - project title - human cartographer - date created - model or models used - prompt archive location - data sources - transformations - classification method - style influences - privacy protections - human review notes - publication disclosure text
Fill in as many fields as you honestly can. Mark unknowns as unknown rather than inventing details.
Chapter 5 distinguishes legality from ethics. If your map borrows a visual lineage, name it clearly instead of hiding it inside adjectives.
Review this map concept for attribution and style lineage. Identify: 1. dataset credits I should show 2. licenses or update dates I should verify 3. basemap or tile credits that might be required 4. style influences I should acknowledge 5. places where an LLM might invent or omit attribution Then draft a concise Data Sources and Credits section.
In prompt cartography, bias can hide in adjectives, labels, classification choices, and the implied audience. Ask the model to explain its design logic before you accept it.
Act as a skeptical cartographic ethics reviewer. Audit my map concept for explicit and implicit bias. Pay special attention to: - adjectives or labels that frame people or places unfairly - classification choices that could exaggerate or hide patterns - color palettes that imply blame, danger, safety, or moral judgment - missing perspectives or affected communities - claims the map might imply without evidence Then propose one alternative framing and explain what it changes ethically.
Privacy is not only about whether a dataset is public. It is about whether your map makes people, routines, protected categories, or vulnerable places easier to target.
Screen this map concept for privacy and representation risk. Classify each risk as low, medium, or high. Include risks related to: - individual addresses or precise points - small counts or rare categories - sensitive populations or protected classes - stigmatizing place labels - zoom levels that reveal too much detail - popups, labels, or downloadable data For each medium or high risk, recommend a mitigation such as aggregation, masking, coarser zoom, removing labels, changing tone, or adding uncertainty language.
Disclosure is a design choice. It tells readers that AI assistance did not erase human responsibility.
Draft a short public disclosure for this map. It should say: - that the map was AI-assisted through natural-language prompts - that a human reviewed the output - what datasets or source types were used - where credits and limitations can be found - what privacy protections were applied Keep it under 90 words and avoid defensive legal language.
Then write an archive note naming what you would save: prompts, outputs, model name, data sources, transformations, screenshots, credits, review notes, and final disclosure.
Submit or save the following:
Tiny nudge, large payoff: make the cartographic decision visible.
Do the tiny-but-mighty judgment check. If these answers are fuzzy, revise the map idea before polishing the prompt.
For use with Prompt Cartography: Interactive Web Map Design with LLMs, CRC Press. www.promptcartography.com
Prompt Cartography