Facilitator guide: Algorithmic Bias & Systemic Justice
Materials
- Projector.
- Devices for the fairness slider once it exists, or two decks of 20 cards per pair (see below).
- Printed case cards for the Step 2 sort.
- Four corner signs.
Lesson steps
- 1. Same tool, different mistakes (Explainer, 5 min)
- 2. Three ways inequity gets in (Explainer, 7 min)
- 3. Two kinds of fair (Explainer, 7 min)
- 4. The fairness slider (Activity, 8 min)
- 5. You are the auditor (Scenario, 6 min)
- 6. Quick check (Quick check, 4 min)
- 7. Where it touches you (Reflect, 5 min)
- 8. Recap (Recap, 3 min)
Note for whoever leads this: this module touches on race, gender, disability, and class. Keep the focus on how systems work, never on individual students, and let anyone pass.
Card decks for Step 4
Use the Step 3 numbers scaled to 20. Group A deck: 20 cards, 10 marked “will reoffend.” Start with 15 in the high-risk pile (9 marked) and 5 in the low-risk pile (1 marked). Group B deck: 20 cards, 6 marked. Start with 5 high (3 marked) and 15 low (3 marked). Both high piles start at 60% marked, and both low piles at 20%. Wrongful flags start at 6 of 10 in Group A and 2 of 14 in Group B. Learners move cards to try to equalize both.
Timing for a 75-minute class
For 60 minutes, do the Step 2 sort as a whole class. For 90 minutes, let groups present their Step 4 choices and defend them.
| Minutes | What happens | Format |
|---|---|---|
| 0-3 | Pre-check | Individual |
| 3-10 | Step 1: guess the sources of the gap before reading on | Whole class |
| 10-22 | Step 2 and the sort | Pairs |
| 22-34 | Step 3: work the numbers on the board, then the Amsterdam case | Whole class |
| 34-48 | Step 4 Try it with the slider or cards | Pairs |
| 48-58 | Step 5 audit scenario | Groups of four |
| 58-62 | Step 6 quick check | Individual |
| 62-71 | Step 7: ARK’s view, reflect, then the fix-it corners | Individual, then whole class |
| 71-75 | Step 8 recap and post-check | Whole class |
Before you start
- Keep discussion on systems and evidence. Never ask students to share their own experiences of discrimination, and let anyone pass.
- Some students may have family members affected by criminal justice, health, or housing systems. Present cases calmly and avoid graphic detail.
- In Step 3, slow down. Draw both groups on the board and count together. The point lands when students see that the math, not anyone’s bad intent, forces the trade-off. Amsterdam then shows it happening to a city that tried to do everything right.
- ARK’s view is an opinion, labeled as one. Invite students to argue with it. No grade depends on agreeing.
Discussion prompts
Use the deck’s “Ban it or fix it?” debate and “Where’s the real fix?” corners.
Answer notes
Quick check: 1 B, 2 C, 3 A, 4 C, 5 B. Pre/post: 1 B, 2 A, 3 C. If learners choose A on quick check question 5, ask who the average hides.
Words to know
- disparate impact
- When a rule that looks neutral harms one group more than others.
- proxy variable
- A data point that stands in for a protected trait, such as a zip code standing in for race.
- base rate
- How common an outcome is in a group before any prediction is made.
- intersectionality
- How overlapping identities, such as race and gender together, shape what a person experiences.
- algorithmic audit
- An independent test of a system for unequal outcomes.
Workshop quick checks
Learners can answer these on their own phones with no account. Answers count toward ARK's anonymous totals.
- Before the lesson: aireadiness4kids.org/check/algorithmic-bias-and-justice/pre
- After the lesson: aireadiness4kids.org/check/algorithmic-bias-and-justice/post