Sound research design bridges theory and evidence by specifying research questions, hypotheses, populations, and methods. Strong designs anticipate validity threats and clarify how data collection and analysis answer the research question. Design choices made early constrain and enable later analytic options.
From your study of the philosophy of social science and epistemology, you understand that there are deep disagreements about what counts as knowledge, how causal claims can be justified, and what the appropriate relationship is between theory and evidence. Research design is where those epistemological commitments become methodological choices. A positivist believes that social phenomena can be measured objectively and that causal relationships can be estimated from data; they tend toward surveys, experiments, and quantitative models. An interpretive researcher believes that social meaning is constructed and that understanding requires grasping actors' perspectives from the inside; they tend toward interviews, ethnography, and discourse analysis. Your design must be coherent with your epistemological stance — a mismatch produces studies that answer a different question than intended.
The starting point is a well-formed research question. A useful research question has three properties: it is specific enough to be answerable (not "why is inequality bad?" but "does income inequality increase political polarization?"), it is empirically tractable (there exists evidence that could bear on it), and it is non-obvious (if the answer is already known, you're doing description, not research). From the research question, you derive hypotheses — specific, falsifiable predictions about what you expect to find and why. Hypotheses do two things: they focus data collection, and they commit you in advance to what would count as disconfirming evidence. A hypothesis stated after seeing the data is not a hypothesis — it is a post-hoc story.
The next step is specifying the population and unit of analysis. Who or what are you studying? If you're studying whether police presence reduces crime, is your unit a city, a neighborhood, a precinct, a time-period? The answer shapes everything: what data you need, what comparisons make sense, and to whom your findings can be generalized. This is the question of external validity — whether your findings travel beyond your specific sample and setting. It is easy to get a clean, technically rigorous result that answers your question for a narrow population with no obvious relevance elsewhere. External validity requires deliberate design, not accidental luck.
Equally important is internal validity — whether your design supports the causal claim you want to make. If you observe that cities with more police have less crime, does that mean police reduce crime, or that rich cities (which have both) drive the relationship? Threats to internal validity include confounding (unmeasured third variables that cause both X and Y), reverse causality (Y causing X), and selection bias (non-random sorting into conditions). Strong designs anticipate these threats and build in defenses: random assignment eliminates many confounders; difference-in-differences designs control for stable unmeasured confounders; instrumental variables address reverse causality. No design eliminates all threats, but strong designs name the threats explicitly and argue why residual concerns are manageable.
The key insight — that design choices made early constrain analytic options later — means that the time to think about analysis is before data collection, not after. If you want to estimate a causal effect, you need to design the study so that causal estimation is possible. You cannot randomize after the fact. You cannot add a comparison group after collecting only treatment data. The discipline of research design is the discipline of working backward from your inferential goal to the data structure required to support it, and then asking whether you can actually collect that data.
Topics in reflective domains aren't scored by quiz answers. Read, reflect, and mark when you've thought it through.