Whole Earth Curriculum WEC

Lesson 04 — Bias: How It Creeps In and How to Reduce It

Learners meet bias as a steady lean in one direction — different from random error — and see how selection bias and observer bias creep into results. They play a sampling game, label the bias in real scenarios, and practice the fixes: random sampling, blind measurement, and recording before judging.

D06 P3: Intellectual & Cognitive Awareness D06.S1 50 minutes Draft

Objectives

Essential question

How does bias differ from error, and what can we do to keep it out of our results?

biaserrorsamplerandomselection biasobserver biasblindfair test

Materials

Standard materials

Low-tech / no-cost

Enriched / lab & device

Works in different contexts

A diagram contrasting biased and fair sampling. On the left, a garden bed is shown with many plants; a dashed circle encloses only the plants nearest the path, and a caption reads 'selection bias: only the sunniest, easiest plants were measured.' On the right, the same garden shows several small circles scattered at random spots across the whole bed, labeled 'random sample: spots chosen without a pattern, so no corner is favored.' Below, a panel shows a person reading a ruler while peeking at a label, labeled 'observer bias: the reader leans toward the answer they expect,' with a crossed-out version labeled 'blind measurement: the reader does not know which sample is which.' Labels and shapes, not color alone, carry the meaning so it prints clearly in grayscale.
A diagram contrasting biased and fair sampling. On the left, a garden bed is shown with many plants; a dashed circle encloses only the plants nearest the path, and a caption reads 'selection bias: only the sunniest, easiest plants were measured.' On the right, the same garden shows several small circles scattered at random spots across the whole bed, labeled 'random sample: spots chosen without a pattern, so no corner is favored.' Below, a panel shows a person reading a ruler while peeking at a label, labeled 'observer bias: the reader leans toward the answer they expect,' with a crossed-out version labeled 'blind measurement: the reader does not know which sample is which.' Labels and shapes, not color alone, carry the meaning so it prints clearly in grayscale.

Lesson 4 — Bias: How It Creeps In and How to Reduce It

Summary

Learners distinguish bias — a steady lean in one direction — from the random error they met yesterday. Through a sampling game and three real scenarios, they see how selection bias (choosing the easy or familiar sample) and observer bias (leaning toward the answer we expect) creep into results, and they practice the fixes: random sampling, blind measurement, and recording before judging.

Objectives

  • Identify sources of bias in a simple investigation and suggest how to reduce them. (D06.S1.06.02)

Connection

When someone asks “which foods do people here like best?” and asks only their own friends, or measures only the plants closest to the path, the answer comes out leaning one way — not by accident, but because of who or what was chosen. That steady lean is bias. It matters beyond the classroom: whose health gets studied, whose needs get counted, whose voices shape a decision. If the sample is not fair, the answer is not fair. Today you learn to spot the lean and straighten it.

Materials

  • Bias scenario cards
  • Bias-hunt page
  • A bag of mixed counters or beans of two kinds

Preparation

  • Print the scenario cards and bias-hunt page.
  • Fill each bag with mostly one kind of counter/bean and fewer of the other, mixed well.
  • Recall Lesson 3: error is a random spread; today’s topic is a steady lean.

Facilitator note

Written to the learner (“you”). The core distinction: error is random (yesterday), bias is a steady lean in one direction. Two kinds to land: selection bias (you sample only the easy, near, or familiar cases) and observer bias (you unconsciously read toward what you expect). Three fixes: random sampling (choose without a pattern), blind measurement (the reader does not know which sample is which), and record before judging (write all data down first, decide later). The sampling game makes it concrete: a handful that is mostly one color “leans” the same way a chosen sample does. This lesson carries the egalitarian lens directly — whose data counts and whose is left out is a fairness question, not just a technique — so name it as a value, while the mechanics of bias are the evidence-based part. Keep scenarios age-appropriate and connect every bias to a fix (agency, not despair). See docs/facilitation.md.

Procedure

  1. Gather (5 min). Recall yesterday: error made your five measurements a little different, all around the truth. Bias is different — it pushes every result the same way. Think: when has a choice of where or whom to check made an answer come out one-sided?
  2. Play the sampling game (10 min). Your bag has two colors of counters, mixed. First, grab a big handful from the top only — note how much of each color you got. Now shake the bag, close your eyes, and draw randomly, one by one. Compare: which draw better matches what is actually in the bag? The chosen handful leaned; the random draw did not.
  3. Meet two biases (10 min). Look at the diagram. Selection bias is choosing a sample that favors some cases over others — measuring only the sunny edge of the garden. Observer bias is leaning toward the answer you expect — reading the ruler a little toward “taller” because you hope your plants are tallest.
  4. Hunt the bias (15 min). Read the three scenario cards. For each, write on your page: what made the result lean, which bias it is, and one way to reduce it. Use the three fixes — random sampling, blind measurement, and record before judging.
  5. Close (5 min). Bias is a lean we can straighten. When the sample is fair and the reader is blind, the answer can be trusted by everyone — and fairness in science is a fairness to people.

Differentiation

  • Support: Sort scenario cards into “fair” and “not fair” first; use a two-word record (selection | observer) with a fix bank.
  • Extension: Design a small fair survey of the class on a harmless question, describing exactly how you will choose a random sample and record before judging.

Assessment

  • Formative (observation): Can the learner label a scenario’s bias (selection or observer) and suggest a matching fix (random, blind, or record-first)?
  • Self-check: The learner asks, “Would my method let an unexpected answer show up, or does it quietly push toward the answer I expect?”

Home connection

Notice one “count” around you — who is included and who is missed. Ask: is the sample fair, or does it lean? Tell someone what you would change to make it fair.

Resources