Design thinking for AI
Applying a fairness lens is the first move. The second move is translating what you notice into an actual design. This unit helps you do exactly that: you turn your fairness thinking into a design brief and a simple workflow blueprint that names guardrails, checkpoints, a transparency move, and one monitoring signal.
The goal isn't a perfect process. The goal is a visible, explainable, and improvable one.
Key ideas and models
Let's explore some key ideas and models.
Start with a design brief, not a tool
A design brief names what you are trying to accomplish, who is affected, and what boundaries are required before any tool is selected. Starting with the brief rather than the tool prevents the workflow from being shaped by what AI can do rather than what your learners or colleagues actually need. A brief that names the purpose clearly, identifies at least two stakeholder groups, and sets input and output boundaries gives you something to measure against when the first draft arrives.
The workflow blueprint template
A workflow blueprint translates the design brief into an operational structure. It moves from purpose to stakeholder list, then through input boundaries, output boundaries, guardrails, checkpoints, a transparency line, and a monitoring signal. Each element has a job:
- Input and output boundaries set what goes in and what can come out
- Guardrails prevent the most foreseeable harms
- Checkpoints create required human moments
- The transparency line names accountability
- The monitoring signal creates a feedback loop that makes the workflow improvable over time rather than fixed on launch.
A practical set of guardrails
A guardrail is a non-negotiable boundary applied before a workflow runs. Three guardrails cover most educator use cases:
- Don't input student identifiers or sensitive staff details
- Don't use AI output as a final decision without human review
- Don't publish or send without a transparency step.
Guardrails are most useful when they are written as "don't" statements rather than "try to" statements, because "try to" leaves room for skipping under time pressure. A guardrail should hold even when the week is busy.
Checkpoint and transparency make trust practical
A checkpoint is the required human step before output is used or shared. It names what must be verified, against what source, and what counts as a problem. A transparency line is the one sentence you say or write to explain what was checked and who is responsible. Together, a checkpoint and a transparency line make accountability visible to everyone who relies on the workflow—not just to the person who ran it. When both are present, a workflow is easier to trust and easier to improve because people can see what happened and ask meaningful questions about it.
Why this matters: Design thinking helps educators move from one-off choices to repeatable workflows that protect equity and belonging. When guardrails, checkpoints, transparency, and monitoring are built into the workflow, fairness becomes easier to see and easier to improve over time, because the process doesn't depend on any one person remembering to check. It becomes the way the work gets done.