Research can be construed in many different ways, but one way to look at it is the exploration of hypotheses: we believe the world works in a certain way, but we don't know for sure, and we want to test that hypothesis.
So we might, to choose a culinary analogy that may not hold up, for example, have a hypothesis that tofu and pomegranate would go well together in a raw salad.
We may have reasons to believe this--we may, for example, have read a review of restaurant that served such a dish, or we may have read a cookbook that suggested that the flavors would work well together, or we may have read some chemistry/biology textbook that leads us to believe that some chemicals in the two would combine well. We have reasons that we believe the hypothesis. To the extent that such reasons are supported in published literature, and that our idea came from reading the literature, we can add such elements to our literature review. But presumably there is not such a preponderance of evidence to suggest that the exact thing that we're studying is certain (e.g., there are no reports of a tofu-pomegranate salad craze in major metropolitan restaurants, or other indication that our hypothesis has been extensively tested).
In order for us to be doing research, we have to be testing something that is at least in question. As far as empirical science is concerned (whether social science or hard science), a hypothesis, no matter its logical antecedents, is worthy of empirical testing if the given empirical test (or one substantially similar to it) has not been executed. The fact that theory suggests that something will happen is no guarantee that it will.
Things are complex. It's simple to say something about preparing tofu and pomegranate, but that hides a great deal of potential complexity, and potential difficulties. To make up an example, it might be the case that tofu and pomegranate go well only with a third ingredient, but that ingredient is rare, or expensive, or hard to work with in some way, making practical execution of a dish infeasible, even if the theory suggests that it should work.
The scientist looks for this complexity within the simpler statement.
"Tofu and pomegranate will go well together" is a simple hypothesis but it suggests more detailed hypotheses and issues: will they go well together when fried? when raw? when boiled? when mixed with vinegar? Will any problems crop up? What will be done to find out? The scientist looks at the simple hypothesis and asks detailed questions about how that can be true. And the starting place for that exploration is intellectual: what do you know about the situation? what are the important ideas that define the situation? what kinds of theories shape your understanding of the situation? what kinds of questions can you ask about the situation? What details are pertinent? Where do theory and practice diverge? And what is the impact of that divergence? If you're attempting to import a theory or practice from one type of endeavor to another, what differences are there going to be? For example, maybe one heard that pomegranate and chicken was really good, but was vegetarian, so thought of pomegranate and tofu instead. What reasons do you have to believe that the translation will work? What reasons do you have to believe that the translation won't work?
Looking at hypothesis with a critical eye looking for detail, many different questions and ideas and possibilities arise. Practically speaking, each needs to be tested individually. So if you start with a general question, you're looking to find a specific aspect of that question that you can test. If you think pomegranate and tofu will work well together, but that they'll need to have some sort of seasoning, then you'll try preparing some with one set of spices/flavors, and you'll see if that works, and then you'll test with a different set of spices/flavors, and see if that works. You won't throw all the spices in at once, because then all you get is confusion. So with a question that can be fragmented and broken down, you want to seek the different questions that could contribute to answering the main question, and look to answer one of those more detailed questions.
But whatever you're going to do, it starts with your looking at the world and putting forth a hypothesis: "I believe the world works this way," you say to yourself. For example "I believe pomegranate and tofu would taste good together," or "I believe that method X, used in field A, will also be useful in field B, despite some differences between those fields."
You start with having an understanding of how the world works, and an idea that it will work in a certain way. Then you look to see what evidence you have to support that view. If you think the evidence is overwhelming that the view is true, then it's not an interesting research question--but if there is doubt--perhaps there are people who believe that it isn't true, perhaps you doubt yourself; it doesn't matter where the doubt arises--then there is a viable research question: you believe the world will work in a certain way, and then you want to test that idea, and you try to find a way to test that idea. It all starts with how you understand the world and your exploration of the places where you are uncertain and curious.