Justice
Optimising for the people already in the data leaves out the people who never got in.
By the end of this lesson
- Explain why a dataset understates the people a service misses
- Find a group your work would not currently reach
- Use analysis to notice absence, not only to find patterns
Watch
Episode V03
The video for this lesson has not been produced yet. The narration below is the same script, and it plays now.
Transcript
Sometimes the biggest failure in a system is not what it does wrong. It is who the system never sees.
Justice begins by looking beyond the average. Beyond the dashboard. Beyond the people already being served. And asking: Who is being overlooked?
It may be a child without transportation. An older adult struggling with technology. A caregiver drowning in paperwork. A veteran who does not know which door to open. A small business disconnected from resources. Or a neighbor nobody thought to ask.
AI can help us map those gaps. Compare neighborhoods. Analyze services. Identify patterns. And reveal where resources are plentiful—and where people are falling through the cracks.
Choose one community system: education, transportation, healthcare, housing, public services, or technology. Ask: Who benefits easily? Who struggles? And whose experience is not represented in the data?
Then do not stop with the analysis. Find a person behind the pattern. Listen. Because justice is not only about better data. It is about recognizing human dignity inside the data.
The next breakthrough may not be serving everyone faster. It may be finally noticing the person nobody designed the system to serve.
When intelligence helps us notice one another, technology becomes a tool for justice.
The lesson
Every record of a service is a record of the people who managed to use it. The ones who could not get through — wrong hours, wrong language, no transport, no trust, gave up at the third form — do not appear as a row marked failed. They appear as nothing at all, and nothing is invisible to analysis.
This is the most consequential blind spot in ordinary data work, because the analysis is not wrong. It is accurate about the population it can see, and confidently silent about the rest. Optimising against it makes the service better for people it already serves and no better at all for the people it is missing.
The practical move is to change what you ask for. Instead of asking which patterns show up in the records, ask who would be absent from these records by construction — who would never have got as far as being counted. That question can be answered from what you know about the barriers, and it points somewhere the data never will.
This virtue is the reason the framework has a step about access barriers and the reason the app tracks a return for access separately from a return for time saved. Reaching someone who was previously excluded is a different achievement from serving an existing user faster, and totalling them would hide it.
Straight from the framework
Virtue 3
Who is being overlooked?
- What it contributes
- Finds the people a system never counted.
- Where it goes wrong
- Optimizing for the people already visible in the data.
- The practice
- Ask who AI could help you notice, not only what it could automate.
Do this in the app
Put the lesson down somewhere real
Write down one group that would be missing from the records of the service you are studying, and why.
Done when — You have named a group and the specific barrier that keeps them out of the data.
/dashboard
Go further
Check yourself