Drawing conclusions from evidence (age 9+)

Activities to try

1. 1

   Data-and-trust bench: watch confidence respond to the evidence

   Why it works: How much we can trust a conclusion is earned by how the evidence was gathered — a larger sample and repeat readings that turn out to cluster closely justify higher confidence, while small samples, widely scattered repeats and unexplained anomalies justify caution. A real causal claim requires ruling out other explanations, not just two things occurring together.

   1. Study a provided dataset (e.g. plant growth at different fertiliser doses) and write a conclusion as a causal claim using the frame 'X caused Y because ___ (evidence/mechanism)'.
   2. Explore how trust changes: add repeat readings and observe whether the readings cluster closely or scatter; shrink the sample to only two plants; reveal a hidden outlier/anomaly. Remember that adding repeats reveals whether readings are close — close, repeated readings support higher trust, scattered ones support lower trust; adding repeats does not by itself force the readings together.
   3. After each change, state whether your confidence should rise or fall and why, before consulting the optional 'trust meter' scaffold (a proposed, unvalidated aid) to compare your reasoning.
   4. When an outlier appears, decide: repeat the measurement, investigate the odd reading, or keep it — and justify the choice.
   5. Feed a causal claim into the optional causal-claim checker (e.g. type 'the fertiliser caused more growth') and answer its prompt about whether anything else differed between the groups.
2. 2

   Defend it to a skeptical scientist (oral report and confidence)

   Why it works: Presenting findings and discussing degree of trust is a dialogic skill: a well-reasoned 'we can't be fully sure because…' is stronger science than false certainty, and a causal claim must survive the question 'what else could explain it?'

   1. Present your findings aloud, stating the causal conclusion in appropriate scientific vocabulary.
   2. Respond to a skeptical questioner who presses: 'You say the fertiliser caused it — could something else explain that?' and 'How confident are you, given only three plants?'
   3. Qualify your confidence out loud on a spectrum: 'We can be fairly confident because we repeated it three times and the readings were close, but the sample was small.'
   4. Answer follow-ups that probe whether other variables were controlled and whether any measurements might be off.
3. 3

   Rank and sort: which evidence deserves more trust?

   Why it works: Reliability can be compared and reasoned about: evidence quality (sample size, repeats, spread, control, anomalies) determines how much weight a conclusion deserves, and a causal claim says more than a correlational observation, so it needs more support (a controlled comparison and a plausible mechanism) before it can be trusted.

   1. Compare two or more results and choose which should be trusted more, giving the reason (e.g. more repeats, larger sample, readings that cluster closely).
   2. Order several datasets from most to least reliable and justify the ordering.
   3. Sort statements into 'causal' vs 'correlational' categories (e.g. 'the plants given more light grew taller because they could photosynthesise more' vs 'the taller plants also happened to be near the window').
   4. Examine a results table with a flagged anomaly and state how it affects the trust in the conclusion.