← Back to course
Applied AI / Week 05 / Friday
5/6
Week 05 · Fair for Everyone

Friday

Week 5 review
// Who is missing, and who is misjudged?
⏱ about 15 min

Friday: Week 5 Review

At the end of the week, Maya, Ben and Rosa compare what they found.

Maya explained the photo app to her son. "The tool is missing people, not the other way round."

Ben decided he will listen to every student himself before trusting any reading score. Rosa rewrote her spring email for all her customers.

This week in five lines

  • Biased training data can lead to some people being less well served (AI4K12).
  • Generative AI can amplify historical and social biases and work worse for some groups or languages (NIST).
  • This can come from training data that does not represent everyone (NIST).
  • Examples: voice tools and regional dialects, and exam monitoring that may unfairly flag some students (Department of Education).
  • Trustworthy AI is fair, with harmful bias managed, and that takes people who look for it (NIST).
REVIEW
  • Read the question.
  • Tap your answer.
Which two questions guide a bias check in this course?
NIST says performance gaps may come from what?
A tool flags a student during a test. What is the fair next step?
A draft says "older customers do not read email." What is it?
StatementTrue or false?
NIST lists fairness among the traits of trustworthy AI.?
If a tool works for you, it works for everyone.?
Bias in writing can look like a stereotype stated as fact.?
Automation bias can make a biased output more harmful.?
WHY THIS EXERCISEFairness is one of the traits you will build into your AI plan in week 8.
A BIAS CHECK, STEP BY STEP
  • ?Ask who the tool or output is for.
  • ?Ask who is missing from it.
  • ?Ask who it describes unfairly or gets wrong.
  • ?Remove or fix the biased parts.
  • ?Have a person decide before anyone is affected.
WHY THIS EXERCISEA short routine makes fairness a habit, not an afterthought.
In your own words, what is a performance gap?
WHY THIS EXERCISENaming the gap is how you start a fair conversation about a tool.

Sources: AI4K12 Five Big Ideas · NIST AI 100-1 (2023) · NIST AI 600-1 (2024) · US Department of Education (2023)

Tomorrow is a Try It day: a bias hunt in a printed example.

← Thursday