4.3 Evaluating Claims & Inference vs Observation

4.3 Evaluating Claims & Inference vs Observation

The final part of this skill area is the most grown-up: judging other people's statements. ICAS gives you claims — from advertisements, from characters in a stimulus, from a short news-style report — and asks whether the science behind them holds up. To do that well you need two tools: telling observations from inferences, and matching claims to evidence.

Observation vs Inference — the Exam Version

Earlier chapters taught the basic difference: an observation is what you detect directly with your senses or measure with an instrument; an inference is an explanation you construct from observations. Here the focus is how ICAS tests it. The classic puzzle: the footpath is wet this morning.

• Observation: the footpath is wet.
• Inference: it rained overnight.

But that inference is only one possible explanation — a sprinkler, a burst pipe, or someone hosing the driveway would produce exactly the same observation. The exam habit is this: an observation has one answer (what you actually saw), while an inference can have several rivals. When a question asks which statement is an inference, look for the one that adds an explanation the observer did not directly see. When it asks what you can be sure of, pick the observation.

A quick check is the camera test: could a camera have recorded it? A wet footpath — yes. Rain falling overnight — not unless the camera was running all night. If the camera test passes, it is an observation; if it fails, you are dealing with an inference.

Good scientists, and good ICAS candidates, treat inferences as "best explanation so far" rather than facts. The more rival explanations you can rule out with new observations — say, checking the weather record or looking for a sprinkler — the stronger your inference becomes.

Evaluating Claims in Advertising and Media

A claim is a statement someone says is true and wants you to believe — "9 out of 10 plants grow taller with Brand X fertiliser!" Advertising claims often borrow the costume of science (numbers, lab coats, graphs) without the substance. Ask three questions of any claim:

1. Where are the numbers from? How many plants were tested — 10 or 10,000? Who ran the test — an independent scientist or the company selling Brand X?
2. What was it compared with? Taller than plants with no fertiliser, or taller than plants using other brands? "Grows taller" needs a comparison group to mean anything at all.
3. What is not being said? "9 out of 10" could hide tiny differences, one lucky trial, or one plant in ten dying.

Certain phrases should set off alarm bells straight away:

• "Scientists say..." — which scientists, working where, and paid by whom?
• "Clinically proven" or "miracle" — science rarely talks in miracles, and proof needs published, repeated results.
• "Natural, so it must be safe" — many natural things are harmful; safety comes from testing, not from labels.
• A dramatic before-and-after story about one person or one plant — a single story is an anecdote, not evidence.

None of these phrases proves a claim is false; they are simply signs to demand better evidence before believing it.

What Evidence Would Support — or Refute — a Claim

ICAS loves asking which result would support a claim and which would weaken it. Supporting evidence is a fair test whose results match the claim; refuting evidence is a fair test whose results contradict it. For Brand X, a fair test would grow 40 similar seedlings, half with Brand X and half without, with everything else — light, water, soil, pot size — kept identical. If the Brand X group clearly grows taller across many repeats, the claim is supported; if both groups grow the same, the claim is weakened, no matter how confident the advertisement sounds.

When a Claim Goes Beyond the Data

A claim goes beyond the data when it promises more than the evidence shows. If the test ran for three weeks, "keeps plants healthy for life" goes beyond the data. If only tomato plants were tested, "works on every plant" goes beyond the data. The fix is always the same: shrink the claim to fit the evidence, or gather more evidence to fit the claim.

Does the Experiment Actually Test the Claim?

The sneakiest questions show an experiment and ask whether it tests the claim it cites. Check the match:

The experiment must change the thing the claim is about and measure the thing the claim predicts. If either is wrong, the experiment does not test the claim — however scientific it looks.