Does the Research Question Assume the Population Can Be Identified?

Who, Exactly, Belongs to the Population in Your Question?

“University students,” “AI users,” “early-career researchers,” “high-performing teachers,” and “students at risk of dropping out” all look like populations. Yet some become surprisingly difficult to identify once recruitment or data extraction begins.

Who counts as an AI user: anyone who has tried a tool once, someone who uses it for academic work, or someone who uses it regularly? When does a researcher stop being “early career”? What observable criteria make a student “at risk” before the outcome actually occurs?

A research question may therefore assume more than the existence of a population. It may assume that researchers can determine who belongs to that population consistently enough to recruit, sample, classify, and eventually state whom the findings concern.

A Population Must Be More Than a Familiar Label

Move From a Population Label to People or Cases You Can Actually Identify

Defining the target population is a fundamental part of study design. The target population is broadly the group about which the researcher wants to make inferences, while the accessible population is the subset available for recruitment under the geographical, temporal, institutional, or other practical conditions of the study.

Eligibility criteria then translate the population concept into rules governing who can enter the study. The important point is that these layers should remain connected. If the eligibility rules identify a materially different group from the population implied by the research question, the study may be answering a narrower or different question than intended.

Start by Asking What Defines Membership

Take the population named in your question and complete this sentence: “A person or case belongs to this population if...”

For “undergraduate students enrolled in fully online degree programs,” membership might depend on enrollment status, program level, delivery mode, institution, and study period. Those characteristics can potentially be translated into observable eligibility rules.

Now consider “students who are overly dependent on AI.” Membership is less obvious. What does “overly dependent” mean? Frequency of use? Inability to complete work without AI? Use for particular tasks? A score above a threshold on a validated measure? A judgment by instructors?

If reasonable investigators could classify the same person differently because the population label has no sufficiently clear operational meaning, population identification requires further conceptual work.

Distinguish the Target Population From the Accessible Population

You may want to understand “university students in the Philippines” but have access only to students enrolled at one institution. Those are not automatically the same population.

A study can legitimately use an accessible population narrower than its broader population of interest. The methodological question is how far the evidence permits inference beyond those actually represented in the study.

Specify Person, Place, and Time When They Matter

Population definitions often become clearer when researchers specify who is being studied, where those people or cases exist, and during what period they qualify.

“Teachers” may be unnecessarily broad. “Full-time public secondary-school teachers employed in Region X during the 2026–2027 academic year” identifies a much more concrete population. Whether that degree of specificity belongs directly in the research question depends on the study, but researchers should know these boundaries when designing it.

Time can be particularly easy to overlook. Enrollment, employment, diagnosis, technology use, program participation, and institutional membership can all change. A person who qualifies today may not have qualified a year earlier.

Eligibility Criteria Should Follow From the Question

Inclusion criteria describe characteristics required for participation. Exclusion criteria generally identify circumstances among otherwise eligible participants that justify exclusion because of safety, validity, feasibility, or other protocol considerations.

These criteria should not be arbitrary administrative conveniences. They operationalize who the study actually concerns.

For example, if a research question concerns first-year undergraduate students' experiences with generative AI, restricting participation to students who have used generative AI for coursework may be justified if experience with such use is necessary to answer the question. Excluding evening students merely because recruitment is easier during daytime classes would have a different rationale and could affect whom the resulting evidence represents.

Do Not Define the Population Using the Outcome You Want to Explain

Some population labels contain the phenomenon the researcher intends to investigate. “Students harmed by generative AI,” for example, already assumes that harm occurred and that AI caused it. “Employees whose productivity declined because of remote work” similarly embeds an outcome and causal attribution into eligibility.

This can create a circular study: participants qualify because the conclusion has effectively been assumed in advance.

When population membership depends on an uncertain empirical proposition, check whether the question treats something as established that should instead be investigated.

Some Populations Exist Conceptually but Lack a Practical Sampling Frame

A target population can be clearly defined even when no complete list of its members exists. This is common in social, educational, community, and online research.

Suppose the population is “Filipino doctoral students who regularly use generative AI for dissertation work.” Researchers may be able to define the population conceptually, yet no national registry may identify every person satisfying those conditions.

This does not automatically make the research impossible. It does affect sampling. Researchers may need institution-based recruitment, respondent-driven approaches, professional networks, screening procedures, or other sampling strategies appropriate to the study. What matters is recognizing that a clear population definition and a complete sampling frame are different things.

Access Is Different From Identification

You may know exactly who qualifies and still be unable to reach them. A hospital database may contain the relevant patient population but be inaccessible because permission cannot be obtained. A school system may have the necessary student records but prohibit researcher access. A professional population may be identifiable but extremely difficult to recruit.

Identification asks, “Can I determine who belongs?” Access asks, “Can I actually obtain the required evidence from them?” Both matter for feasibility, but they should not be confused.

If the required population cannot realistically be accessed, the question may ultimately demand evidence that the proposed study cannot obtain convincingly.

The Population Should Match the Unit of Analysis

Not every research population consists of individual people. Questions may concern schools, universities, research articles, journals, countries, online discussions, classrooms, organizations, courses, or other units.

Researchers should determine what constitutes one eligible case. If the question concerns universities but data are collected from individual faculty members, the unit from which observations are obtained and the unit about which conclusions are drawn require careful alignment.

This becomes particularly important with nested data. Students may be nested within classes, classes within schools, and schools within systems. The population named in the question should correspond to the level at which the substantive claim is intended.

Be Careful With Populations Defined by Self-Identification

For some research questions, self-identification is itself the appropriate criterion. This may be especially relevant for identities, experiences, communities, or roles for which externally imposed classification would distort the phenomenon being studied.

In other cases, self-report may need additional eligibility information. A participant saying “I use AI frequently” does not necessarily satisfy a study definition requiring use on at least four academic tasks per week. The correct approach depends on what population the question actually intends to study.

Your Final Claim Cannot Quietly Expand Beyond the Population You Identified

Suppose a study recruits undergraduate engineering students from one private university. The findings may be valuable, but the population actually represented does not automatically become “all university students.”

Generalizability depends on more than sample size. It involves the relationship among the target population, accessible population, sampling and recruitment processes, eligibility criteria, study context, and characteristics relevant to the phenomenon being investigated.

A narrow population is not inherently a weakness. An inaccurately described one is.

When “AI-Dependent Students” Cannot Yet Be Recruited

The methodological gain is not merely easier recruitment. The revised population gives the eventual findings a clearer referent: readers can understand who was studied and what evidence was used to determine membership.

Common Mistakes When Defining a Research Population

Trace the Population From the Question to the Final Claim

Write the population from your research question at the top of a page. Then identify the characteristics that determine membership, the source from which eligible cases can be located, the rules that will govern inclusion and exclusion, and the population to which you ultimately hope to apply the findings.

Those should form a coherent chain. If the operational eligibility criteria produce a substantially different population from the one named in the question, revise either the criteria or the question.

Population identification is therefore one of the checks worth performing when you stress-test a research question before designing the study.

Check Whether the Population Can Actually Be Identified

Questions About Identifying a Research Population

You Should Be Able to Say Who Belongs Before You Decide Whom to Sample

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.