Think Lab exercise

Test Your Assumptions

What you will be able to do: Identifying which unstated assumption a conclusion actually depends on

Published
Updated
By
ARAC International
Reading time
8 minutes
Level
Intermediate

Before you start

Every argument rests on assumptions, and listing them all is not useful. What matters is finding the one the conclusion cannot survive without.

Both items below come from ordinary organizational life, where this kind of reasoning does real damage precisely because it looks unremarkable.

Every scenario on this page is fictional and was written for this exercise. None of them describes a real organization, person, or event.

Practice

Choose a response for each item, then read why it is stronger or weaker. There is no score, no account, and nothing is sent anywhere. Your answers stay in this browser tab and disappear when you close it.

1. Find the load-bearing assumption

A charity notes that donations arriving through its website fell sharply the month after it redesigned the donation page, and concludes that the redesign is at fault.

Fictional scenario constructed for this exercise
Which assumption is doing the most work?
Show the full answer key for this item

What remains uncertain: One month is a short window. Monthly donation totals in small organizations are often dominated by a handful of large gifts, which makes single-month comparisons unstable.

That nothing else relevant changed in the same period. Strongest reading
Strongest. The whole inference rests on the redesign being the only relevant difference between the two months. Seasonality, a lapsed campaign, a large recurring gift landing in a different month, or an email schedule change would each produce the same pattern.
That the donation figures are accurate. Partly right
A real assumption and worth checking, but a smaller one. Payment records are usually reliable, and if they were wrong the conclusion would be wrong for a different reason than the one being examined.
That donors noticed the redesign. Weaker reading
Not required by the argument. A redesign can reduce completions without anyone consciously registering that it changed, which is in fact the common case.

2. Find the load-bearing assumption

A survey of a neighbourhood association's members finds 78 percent support for a proposed parking change, and the association reports that the neighbourhood supports the change.

Fictional scenario constructed for this exercise
Which assumption is doing the most work?
Show the full answer key for this item

What remains uncertain: The response rate matters as much as the membership question. If 78 percent of a small fraction of members responded, self-selection compounds the coverage problem.

That association members are representative of the neighbourhood. Strongest reading
Strongest. The claim moves from members to residents. People who join a neighbourhood association differ systematically from those who do not, typically in tenure, homeownership, age, and availability. The percentage may be accurate and the generalization still unsupported.
That respondents understood the proposal. Partly right
A genuine concern, especially for technical proposals, and worth checking in the question wording. But it affects the meaning of the 78 percent rather than the leap from members to the neighbourhood.
That 78 percent is a large majority. Weaker reading
This is not really an assumption in dispute. Seventy-eight percent of respondents is a large share of respondents. The problem is who the respondents are.

What to take away

Look for the 'nothing else changed' assumption.
Before-and-after comparisons almost always depend on it, and it is almost never stated.
Watch the population shift.
A great deal of misleading reporting happens in the step from 'respondents' to 'people', or from 'members' to 'the community'.
Distinguish assumptions that would change the conclusion from those that would only change its precision.
Only the first kind is load bearing.

Nothing you select on this page is transmitted anywhere. There is no account, no score, and no record. See the privacy page for details.

Next step

Read the lesson on confirmation bias

Topics covered: Critical ThinkingInformation Resilience

  • Article

    How Confident Should You Be?

    Confidence language, what it should be based on, and how to say what you believe in a way that can be checked later.

    10 minIntermediate

  • Think Lab

    What Would Change Your Mind?

    Practice naming the specific observation that would reduce your confidence, which is the fastest route out of a stuck disagreement.

    8 minIntermediate

How this page is produced

Published by ARAC International under theTHINK editorial standards. If you find an error, please tell us through the corrections page.