Page 321 - CyberSecurityEssentialsEnglish

A summary condenses one source. A comparison places two or more sources on shared
dimensions. A synthesis builds a wider picture from a body of evidence. An inference states
the conclusion that picture permits. Confusing the levels produces writing that looks

comprehensive but explains nothing; a list of “study A said” followed by “study B found” is
not a synthesis.

A dividing rule: a summary serves one source; a synthesis serves the question across a set of

sources. A summary asks what this study said. A synthesis asks what the evidence together
allows us to say, with what degree of confidence, and under what conditions.

Synthesis begins from an organizing question. If the question concerns the effect of an
intervention, we may organize the findings by outcome type, follow-up duration, or intensity

of the intervention. If it concerns participants’ experiences, we may organize by theme,
stage, or group. If it concerns policy, we may compare effectiveness, cost, equity, and
feasibility. The structure comes from the question, not from the order in which the sources

were found.

Thematic grouping is useful, but it can hide contradiction within a theme. Within each strand
we should write: what is the range of the evidence? Which designs? Is the direction
consistent? What are the exceptions? Which conditions change the result? And what is the

certainty? AI can propose groupings or headings, but they must be tested against the cases
that do not fit.

When results differ, several explanations are possible: differences in population,

measurement, dose, follow-up duration, quality of implementation, analysis, context, or
chance. It is not acceptable to select one explanation without evidence. A conditional
explanation can be written: “the difference may be due in part to...”, indicating what

supports it and what remains uncertain.

In quantitative synthesis, the presence of numbers does not mean a simple average can be
computed. Measures, designs, and groups may differ. A meta-analysis needs comparable
effect sizes, a model, assumptions, and an examination of heterogeneity. AI should not

choose a pooling method merely because it can write the code for it. The decision is
methodological and requires expertise.

In narrative synthesis, tables, figures, and text are used to interpret patterns without
statistical pooling. The narrative must also be systematic: describe the studies, group them,

explore the relationships, and appraise the strength of the synthesis. A narrative is not a
space for selecting supportive examples or concealing inconsistency.

320