Page 303 - CyberSecurityEssentialsEnglish

An operational definition turns a concept into something observable or measurable.
“Productivity” may mean the number of transactions, the quality of the product, the time
taken, sales, or a manager’s rating. “Intensive use of AI” may mean minutes of use, the

number of tasks, or the proportion of work revised with the tool. If we do not fix the
meaning, the study may answer a different question from the one the reader assumes.

A hypothesis is a testable expectation about a relationship or a difference. A null hypothesis

assumes no difference or relationship under the statistical model, and an alternative
hypothesis proposes one. But not every study needs a directional hypothesis, and a
hypothesis must not be invented after seeing the results and then presented as though it

preceded the data. Here the distinction between exploratory and confirmatory analysis is
essential.

AI can act as an intellectual opponent: ask it to identify the assumptions embedded in your
question, propose alternative explanations, and say whether the question confuses

association with causation or overlooks a mediating or moderating variable. But do not ask
it to “prove my idea is right”; that builds confirmation bias into the question before any
evidence is gathered.

In institutional projects, the question must connect to a decision. Instead of “what do staff

think of the system?”, the question could be “which factors prevent a transaction being
completed within the target time, and which of them can be changed within three months?”

The second question gives the result a use, and stops you collecting attractive data that leads
to no action.

FIGURE 7-2: FROM A BROAD TOPIC TO A RESEARCHABLE QUESTION
What to notice: The path narrows through the problem, the gap, the question, and then the
operational definition. Jumping straight from a topic to a question usually produces wording

that is too broad or that cannot be tested.
Source: Prepared by the author.

302