Scientific and Engineering Argumentation: From Hypothesis to Solution

A proof answers “what follows from the model.” An experiment is “how the model relates to the world.” A solution is “what to do under remaining uncertainty.”

Scene: the plant wilts

You notice that the plant is wilting and assume: it lacks water. This is not a guaranteed conclusion, but abduction — choosing a plausible explanation.

Next, one can reason:

1. If the cause is a lack of water, the soil moisture will be low.
2. Check the soil.
3. If it is dry, water moderately and observe.
4. If the plant does not recover, consider light, temperature, pests, and root condition.

We turned a guess into testable consequences. This is how the hypothetico-deductive cycle works.

Three types of inference

• Deduction: if the premises are true, the conclusion is necessary.

  All tickets for this session are sold; this ticket belongs to the session; there is no free space for it.

• Induction: from observed cases we estimate the general pattern.

  In ten trips the bus was late; likely, the schedule is unreliable.

• Abduction: choose the best explanation among available ones.

  The lights went out in the entire house; the likely cause is a power outage.

Induction or abduction cannot guarantee deduction. Their strength depends on the sample, alternatives, and checks.

A Good Hypothesis Takes a Chance to Be Wrong

«Everything happens due to a complex combination of factors» is compatible with almost any result and therefore is of little help. A useful hypothesis predicts in advance what we will see if it is true and what will argue against it.

Repair: record:

• observation;
• possible cause;
• prediction;
• check distinguishing alternatives;
• revision condition.

Mill’s Methods in a Everyday Example

Several guests felt unwell after dinner. Without medical advice and premature accusations one can compare:

• what all affected people ate — method of similarity;
• how their diet differed from others — method of difference;
• whether the intensity changed with the amount of the factor — concomitant changes;
• what remained unexplained after excluding known causes — residues.

These methods suggest hypotheses but do not replace qualitative data, specialized expertise, and control of confounding factors.

From research to solution

Even a good research rarely says what to do. A decision adds value, cost, reversibility, and risk.

The data shows a 40% chance of rain.

Whether to take an umbrella depends on the duration of the walk, the cost of discomfort, and how unpleasant it is to get wet. The same fact supports different decisions under different losses.

Short practice

Observation: after rearranging the desk you finish evening work faster.

1. Offer three explanations.
2. For each, output a distinguishing prediction.
3. Devise a reversible check.
4. Record the condition after which you will abandon your favorite hypothesis.

Chapter Rule: A good investigation does not defend a favored explanation. It creates a check where explanations give different expectations.