I am not sure what you mean by getting the LLM to "align" itself with and idea or theory. Can you say more about the kind of output that you want the LLM to produce?
You need to be an expert in that area. Then think about designing a ‘theoretical framework’ to guide your LLM through a series of steps. Be specific about what you want AI to do in each of these steps. In my paper, I used one framework called ACTOR to prompt ChatGPT in thematic analysis. https://osf.io/preprints/osf/vefwc