Where it goes wrong
Traditional teaching
Compute the mean, median, mode, and range of a list. Draw the bar chart. Statistics becomes arithmetic homework wearing a lab coat — and the actual skill, deciding what data can and cannot tell you, is never practiced.
The reteach
The Alice Method
Statistics is detective work: every dataset is evidence, and the question is what conclusions the evidence supports. Start with real questions the child cares about (do our family's pancakes get eaten faster on Saturdays?). Teach the summary statistics as compressions that lose information — the mean of {50, 50} and {0, 100} is the same, and the story isn't. Always ask: what got left out? Who collected this? Compared to what?
Mental model
Compression with loss
An average is a lossy summary: one number standing in for many, chosen to be representative. Every summary throws information away — the skill is knowing what got discarded and whether it mattered. 'The average hides the spread' is the statistician's first reflex.
Transfer
Where this shows up for the rest of their life
This is the defense kit for modern life: news claims, product reviews, medical studies, school dashboards (including MAP reports — parents use exactly this skill in Layer 3). Correlation vs. causation, sample vs. population, outlier vs. trend — children who ask 'compared to what?' grow into adults who can't easily be fooled.
Watch for these
Common misconceptions
The misconception
“The average is the typical case.”
The repair
One billionaire in a café makes the 'average customer' a millionaire. Skewed data breaks the mean — that's why medians exist, and why 'average' claims deserve suspicion.
The misconception
“Correlation implies causation.”
The repair
Ice cream sales and drownings rise together (both follow summer). Give kids a stock of silly confounders and they'll spot real ones for life.
Seen, not said
The visual explanation
Two datasets with identical means and wildly different spreads, drawn as dot plots side by side. The mean as a balance point you can see — and see failing to tell the whole story.
Try it
Interactive example
Sampling variability, live: small samples mislead, large samples stabilize.
Probability is long-run behavior, not fortune telling
Ten flips are wild. Three hundred flips settle near 50%. Randomness is messy up close and orderly from far away.