Teaching AI bias in grade 8 — a 45-minute station activity

The hardest part of teaching AI bias to grade 8 students isn’t the concept — they get it fast. It’s giving them concrete examples they can audit themselves, instead of yet another slide-deck definition. Below is the 45-minute station activity we built after the 5th time we said “okay, but show me a real AI output that’s actually biased.”

The activity in 30 seconds

Five stations. Each one has an AI-generated sample (printed, no internet required) with a specific bias problem hidden inside. Students rotate, identify the bias type, and write a one-sentence fix. End-of-class: 10-minute share-out. Done.

Why station rotation works for this topic

Bias-detection is fundamentally a pattern-recognition skill. Students need multiple examples in quick succession to build the pattern. A single example per slide doesn’t do it — they pattern-match on accident, not pattern-recognize on purpose. Stations force five reps in 30 minutes, which is the magic number.

The 5 bias types we cover

Station 1 — Cultural bias

Sample: AI describes “a typical wedding” — the description is entirely Western-Christian. Students identify what’s missing and why the AI defaulted that way.

Station 2 — Source bias

Sample: AI summarizes a historical event citing 4 sources, 3 of which are listed as URLs that don’t exist (hallucinated). Students Google each, mark real vs. fake, write why this is a research integrity issue.

Station 3 — Gender bias

Sample: AI describes “a typical CEO” using exclusively male pronouns and traditionally male names. Students rewrite the description gender-neutrally and discuss what’s lost / gained.

Station 4 — Temporal bias (out-of-date training data)

Sample: AI confidently describes “current” events that happened 18+ months ago and have since changed. Students fact-check against current news.

Station 5 — Recency bias (over-weighting recent events)

Sample: AI summarizes “the history of X” but disproportionately focuses on the past 2 years. Students compare to a textbook timeline and note the gap.

Rubric (the part that makes it gradeable)

The student-facing rubric has 3 columns: Identified the bias (1 pt) · Explained the why (1 pt) · Wrote a one-sentence fix (1 pt). Five stations × 3 points = 15-point activity. Print on cardstock, students self-check answer key after the rotation.

After rotation, project all 5 samples on the board one at a time. Ask the room: “What bias did you spot?” Cold-call works well here because every student has at least one observation. Spend 90 seconds per sample. The point isn’t to land the “right” answer — it’s to surface multiple plausible reads.

Discussion guide (for the next-day debrief)

Three questions worth 5-10 minutes the next class period:

“If AI training data is mostly from the internet, whose voice is over-represented?” (Answer surfaces who creates most internet content vs who uses it.)
“How would you check for bias in an AI answer you use in your own writing?” (Practical takeaway — connects to ISTE 1.3.d source-evaluation.)
“Should the AI tell us when it’s not sure? Why doesn’t it?” (Connects to model-confidence + RLHF training — the rabbit hole some students will love.)

Standards crosswalk

ISTE 1.3.d: Digital Citizen — evaluation of digital sources (primary)
ISTE 1.7.b: Global Collaborator — multiple cultural viewpoints (secondary, via Station 1 + 3)
AI4K12 Big Idea #3: Learning — how training data shapes model outputs
CCSS.ELA-LITERACY.RI.8.6: Determine an author’s point of view or purpose and analyze how the author distinguishes their position from that of others

Get the printable version

The full kit — 5 station cards + answer keys + rubric + discussion guide — is part of the AI Bias Lesson on our TPT store. Print on cardstock, laminate, reuse across class periods.

Or pair it with the free AI Hallucination Fact-Check Worksheet on the Free Resources page — that one focuses specifically on Station 2 (source bias) and is the lightest entry point if you’re new to teaching this.