03 · What You Need to Know
Start With the Purpose Behind the Research Question
Exploratory research asks, “What is going on here?”
Exploratory research is particularly useful when a phenomenon, problem, population, process, or context is insufficiently understood for the researcher to begin with tightly specified expectations.
The researcher may be trying to identify important concepts, discover patterns, understand how participants frame an issue, clarify the dimensions of a problem, generate hypotheses, assess feasibility, or determine which questions deserve more systematic investigation.
Exploration is therefore often associated with emerging or underresearched topics. It can also be useful when a familiar phenomenon appears in an unfamiliar context or when existing concepts do not adequately capture what researchers are observing.
Suppose universities begin adopting a new form of autonomous AI agent for student learning, but little is known about how students actually incorporate these agents into their study practices. An exploratory study might investigate what forms of use emerge, what students perceive as useful or problematic, and which issues warrant subsequent research.
The goal is not necessarily to produce a final estimate of how common each behavior is or to establish what causes it. The initial task is to understand the territory well enough to ask better questions about it.
Exploratory does not mean unplanned
The word exploratory can mistakenly suggest that researchers simply collect information and see what happens.
Exploratory inquiry still requires a clear research problem, defensible case or participant selection, appropriate methods, systematic analysis, ethical safeguards, and transparency about how conclusions were developed. What is relatively open is the researcher's prior specification of exactly what patterns or explanations will emerge.
Exploratory research can use qualitative or quantitative evidence. Qualitative methods are common because they can capture unexpected meanings, experiences, categories, and processes. Quantitative exploratory analyses may also be appropriate when researchers are examining unfamiliar patterns, developing measurements, assessing distributions, or generating hypotheses for subsequent testing.
Exploration is a purpose, not a synonym for qualitative research.
Descriptive research asks, “What is happening, and what does it look like?”
Descriptive research aims to characterize a phenomenon systematically.
It may estimate prevalence, frequency, distribution, characteristics, behaviors, attitudes, conditions, patterns, processes, or other features of a population or phenomenon. Descriptions can be numerical, qualitative, or both.
A national survey estimating how many university students use generative AI for different academic activities is descriptive. A detailed qualitative account of how collaborative learning unfolds in a particular classroom may also be descriptive, although its form of description and intended scope are very different.
Good descriptive research can answer questions such as:
- How common is the phenomenon?
- Who experiences it?
- What characteristics does it have?
- How is it distributed across groups, places, or periods?
- What patterns or processes can be observed?
Description should not be treated as methodologically inferior because it does not always answer why something occurs. Reliable description is often necessary before meaningful explanation is possible. Poorly understood prevalence, distributions, or patterns can lead researchers to construct explanations for phenomena they have not adequately characterized.
Descriptive research can involve relationships without becoming explanatory
The boundary between description and explanation deserves care.
A descriptive study may report that AI use differs by year level or that two measured variables are associated. Reporting a relationship does not automatically explain why the relationship exists.
For example, finding that students who use an AI tutoring platform more frequently obtain higher grades describes an association. Explaining that association requires additional theoretical and empirical work. Perhaps the platform improves learning. Perhaps stronger students use the platform more. Perhaps motivation influences both. Perhaps some other factor accounts for the relationship.
The statistical presence of an association should not be mistaken for an explanation of its mechanism or cause.
Explanatory research asks, “How or why does this happen?”
Explanatory research moves beyond identifying or describing a phenomenon toward accounting for it.
Depending on the discipline and theoretical tradition, explanation may concern causal effects, mechanisms, relationships, processes, contextual conditions, or theoretical accounts of why observed patterns occur.
Some explanatory questions are explicitly causal: Does an intervention cause an improvement in learning outcomes? Other explanatory questions may investigate mechanisms: Through what processes does feedback affect students' self-regulated learning? Still others may examine why an organizational practice produces different outcomes under different conditions.
Because explanatory claims can be ambitious, the design must match the particular explanation being sought. If the intended claim is causal, merely observing that two variables are associated is generally insufficient. Researchers need a defensible strategy for temporal ordering, comparison, confounding, alternative explanations, and other threats to causal inference.
Explanatory inquiry is not synonymous with experimentation, however. Experiments are powerful for certain causal questions, but explanatory research may also use quasi-experimental, longitudinal, qualitative, case-based, mixed methods, comparative, or other approaches depending on what kind of explanation is sought.
Evaluative inquiry asks, “How well is this working, for whom, under what conditions, and what should be done?”
Evaluation concerns the systematic assessment of an evaluand: something being evaluated, such as a program, policy, intervention, service, initiative, organization, practice, or implementation strategy.
Evaluation questions can address whether implementation occurred as intended, whether intended outcomes were achieved, what unintended consequences emerged, how efficiently resources were used, which groups benefited, under what conditions the program performed well or poorly, and how it might be improved.
The U.S. Centers for Disease Control and Prevention describes program evaluation as systematic data collection and analysis used to assess programs, policies, and organizations and support improvement and decision-making.
Evaluation therefore has a particularly strong relationship to use. The evidence is produced not merely to characterize or explain a phenomenon but to support learning, judgment, accountability, improvement, or decisions about the evaluand.
Evaluation can contain descriptive and explanatory questions
Evaluative research is not separated from the other purposes by an impermeable wall.
Consider an evaluation of a university mentoring program. The evaluation might ask:
- How many eligible students participated?
- Was the program implemented as planned?
- How did participants experience the mentoring process?
- Did participants achieve better outcomes than an appropriate comparison group?
- Why did the program appear to work better for some students than others?
- What changes should be made before the next implementation?
Some questions are descriptive. Some are explanatory. Together, they serve an evaluative purpose because they contribute to understanding and judging the program for practical use.
Current CDC guidance similarly distinguishes process and outcome evaluation questions and emphasizes that evaluation design should follow the evaluation purpose, questions, context, intended uses, resources, and relevant standards.
The easiest distinction is purpose, not method
The same data-collection method can support several research purposes.
| Purpose |
Central concern |
Illustrative question |
Evidence might include |
| Exploratory |
Developing initial or deeper understanding |
How are students beginning to use autonomous AI agents for studying? |
Interviews, observations, usage records, open-ended surveys, exploratory analyses |
| Descriptive |
Characterizing what exists or occurs |
How frequently do students use AI agents for different academic tasks? |
Surveys, records, observations, qualitative descriptions, population estimates |
| Explanatory |
Accounting for how or why something occurs |
Why does sustained AI-agent use improve learning for some students but not others? |
Experimental, observational, longitudinal, qualitative, comparative, or integrated evidence |
| Evaluative |
Assessing an evaluand to support judgment, learning, improvement, or decisions |
Is the university's AI tutoring initiative achieving its intended outcomes, for whom, and under what conditions? |
Implementation evidence, outcomes, comparisons, costs, stakeholder perspectives, contextual information |
Notice that interviews appear potentially useful across several purposes. So do surveys and administrative records. The purpose does not mechanically dictate a method.
This is consistent with the distinction between research design and research methods: what researchers want to accomplish, how the study is structured, and which procedures are used are related decisions, but they are not identical.
The categories are not stages every study must pass through
It is tempting to imagine a universal sequence: first explore, then describe, then explain, and finally evaluate.
Research programs sometimes develop roughly in that direction. An emerging phenomenon may initially require exploration, followed by more systematic description and later explanatory investigation.
But this is not a mandatory progression.
A well-developed field may encounter a new descriptive question. An evaluation may be necessary before researchers have a complete explanatory theory. Exploratory work can occur after decades of research when technology, social conditions, populations, or theoretical perspectives change.
Research purposes respond to what needs to be known, not to a universal maturity ladder.
One study can serve more than one purpose
The categories can overlap within a single investigation.
A study might first describe patterns of student dropout and then model factors associated with those patterns. A qualitative investigation might explore an emerging phenomenon while also producing detailed descriptions of participants' experiences. An evaluation might describe implementation, estimate outcomes, and investigate mechanisms explaining variation in effectiveness.
This does not mean every study should accumulate as many labels as possible. Instead, identify the primary purpose and any secondary purposes that materially affect the research questions and design.
The possibility that a study can be both descriptive and explanatory is especially important because real research questions do not always respect the tidy categories used in introductory methods tables.
Purpose should influence design without being confused with design
Knowing that a study is explanatory does not yet tell you whether it is experimental, quasi-experimental, longitudinal, case-based, qualitative, or something else.
Similarly, saying that research is descriptive does not tell you whether data are cross-sectional or longitudinal, whether the study uses a census or sample, whether the evidence is quantitative or qualitative, or how variables and cases are selected.
Research purpose helps establish what the study needs to accomplish. The research design should then be chosen to produce evidence capable of accomplishing that purpose.
Keeping these levels distinct prevents a common methodological shortcut: treating a statement such as “the study used a descriptive research design” as though it fully explained how the investigation was structured.
