Resources
Reference pages
Sociotechnical protocol
The one-page analytic tool the whole course uses. Seventeen labeled questions in three sets: Side A for any technology, Side B for any claimed success, Side M for a deployment a funder or a law requires. Fill it in with the online worksheet and copy the result.
Case bank
Documented harm cases and success cases for the Ethics Case Analysis and class discussion, each with an entry-point source.
Tools
Every tool used in the labs, all free, with honest notes about what the free tiers actually do.
Kinds of technology
The difference between a technology and a product assembled from technologies, five distinctions for placing any part of one (form, position, decider, location, audience), and eight published schemes that sort technology in other ways, from Winner on built-in politics to the EU AI Act’s risk tiers. Ends with a self-check and a Google Trends activity. Used on the Ethics Case and on whatever a partner brings to the Strategy Pitch.
Client-facing technology
Tools aimed at the person receiving a service: appointment reminders, benefits applications, portals, referral lines, apps and devices, and therapeutic programs. What clients withhold from a record, and why.
Practitioner-facing technology
Tools an organization buys for its own staff: documentation time, recovered time, burnout, turnover, referral platforms, and what a system costs to run rather than to build. Includes two widely quoted figures that cannot be verified against their source.
Technology terms
Seventy-eight terms in seven groups, searchable and filterable, each with a plain definition, its source, and a tag for technology, product, or concept. Eighteen have a flag: a published finding about how the word is applied, with a link to it. Ends with a matching self-check.
Kinds of AI systems
Why “AI” covers several different machines. Predictive risk modeling, NLP and computer vision, and generative AI, sorted by the kind of data and the kind of task, plus the two distinctions inside the generative family that decide what an organization can run: model size and openness.
Language models in depth
The mechanism behind Week 4 in more detail: deep learning, scale, retrieval and agents, and the argument over whether any of it resembles human intelligence.
Field directory
Organizations, free courses, volunteering pipelines, job boards, and funding programs in public interest technology.
Interactive activities
Built for this course
Small in-browser tools we made for specific sessions. Each runs entirely in your browser; nothing you do in them is saved or sent anywhere.
Fairness explorer
Pick a real predictive-risk case and move a risk-score cutoff across two groups; a live confusion matrix shows Type I and Type II error and why equal flag rate, false-positive rate, and predictive value cannot all hold at once. Pairs with Week 3.
Ask AI or ask a person
Sort practice tasks by who to ask, then see where the AI-literacy guidance leans and why. Pairs with Week 5.
Adopt, defer, or refuse
Answer structured questions about a technology decision and receive a reasoned adopt, defer, or refuse with its conditions attached; when adoption is mandated by a funder or law, it returns a negotiation agenda instead. Pairs with Week 6.
Tech-org role map
Click through the roles on a technology team to see what each does and which roles a practitioner can occupy. Pairs with Week 8.
Total cost of ownership
Total the one-time and recurring costs over a planning horizon and see the headline subscription as a fraction of the real total. For the strategy pitch.
Media accessibility check
Answer eleven questions about your video and receive the accessibility actions it still needs, ordered WCAG Level A, Level AA, federal caption and player guidance, course rules, then imagery, each one naming its source. The second half turns your own caption error count into a word error rate. For the digital story, and the Week 11 caption sprint.
Technology readiness self-check
Answer seventeen questions quoted from the NCVO Digital Maturity Matrix, the Data Orchard Data Maturity Framework, and CIS Implementation Group 1, and receive each section back on its own instrument’s scale plus the full instrument to run next. For partner intake in the strategy pitch.
Class tools
Week 1, Defining technology
Teachable Machine trains an image classifier from examples you pick, so a lopsided set produces a biased model in minutes. Also Survival of the Best Fit, a short game where a hiring algorithm learns your own bias.
Week 2, History
Mapping Inequality shows the 1930s redlining maps with the original appraisers’ notes. The Hull-House 1895 maps and the Eviction Lab national map are used in the same lab.
Week 4, Language models
Transformer Explainer runs a real model live so you can watch next-token prediction and temperature. Tiktokenizer shows how text is divided into tokens, and Bycroft’s 3D LLM view renders the architecture one layer at a time.
Week 5, AI literacy
Our Ask AI or ask a person sorter is the lead. Which Face Is Real, Spot the Troll, and Bad News each test how well you judge what you see.
Week 6, AI adoption
Our Adopt, defer, or refuse aid is the lead. The AI Incident Database is a searchable record of documented AI failures in real deployments.
Week 3, Bias and fairness
Our Fairness explorer is the lead. Google PAIR’s Measuring Fairness and Hidden Bias explorables cover the same tests, and Parable of the Polygons shows how small biases compound.
Week 7, Data sovereignty
Blacklight scans any site and lists the trackers running on it. Terms of Service Didn’t Read grades a service’s data practices, and Have I Been Pwned reports breach exposure.
Week 8, Tech industry
Our tech-org role map is the lead. The AI Incident Database is read here for which role on a team could have caught each failure before it shipped.
Week 9, Digital maturity
The readiness self-check is the week’s instrument, built on NTEN Tech Accelerate, the CAST and NCVO maturity work, Data Orchard’s data maturity framework, and the CIS Implementation Group 1 safeguards.
Week 10, Digital storytelling
VidStudio is a browser video editor with no signup, for cutting a rough assembly. The heavier production tools you use later live on the Tools page.
Week 11, Accessibility
WebAIM Contrast Checker returns an immediate pass or fail against the Web Content Accessibility Guidelines (WCAG) on any color pair. Also WhoCanUse, which previews a design under color-blindness and low vision, and WAVE, which scans a page for access errors.
Week 12, Design justice and ICT4D
Our World in Data maps mobile-money adoption over time. The Design Justice Network Principles are the seminar handout.
Week 13, Environmental costs
Track Policy maps real data centers, xAI’s Colossus in Memphis among them, with each site’s power and water use. Hannah Ritchie’s comparison tool sets a chatbot query beside a shower, a mile driven, and an hour of streaming.
Note on link rot
Every link on this site was verified in July 2026. Technology writing ages fast, and some of it will break. If you find a dead link, say so on Canvas; there is usually a mirror, and finding it counts as participation.