Quoting from my MA thesis:
A dispositif, for Foucault, is a heterogeneous system that connects its constituent elements through relations of power and knowledge (Power/Knowledge 194-6). Foucault notes that his method follows the counter-intuitive claim that the phenomenon he’s examining doesn’t exist. And instead of trying to establish the facticity, truth, or cause of its emergence, he asks how events and practices can be organized around something that never existed (Birth of Biopolitics 3, 33). Drawing from the range of notes Foucault made about his methodology, we can surmise that the connections between its elements are not be based on linear or expressive causality, but are be based on partial, multiple, and indirect immanent relationships. A dispositif has no essential constitutive elements, it is only the product of numerous contingent forces in an encounter. As the elements and their force are intensified, reproduced, diminished, or replaced, the characteristics of the dispositif also change. The concept was developed in order to describe formations, at any given historical moment, while maintaining a commitment to the radical contingency, indirect causal relationship, and weak ontology of its constituent elements. While similar to his work in Archaeology of Knowledge, dispositif is put to use in more capacious ways. Unfortunately, Foucault never developed his account of dispositif in any detail, making it difficult to find specifically articulated ways to distinguish it from the forms developed according to his archaeological method. The concept of strata, which Deleuze uses when translating archaeology into his own conceptual system, works as a bridge between archaeology and dispositif. The use of strata is an attempt to make explicit the super-lingusitic semiology that is present in Foucault’s work that often gets ignored when systematized. The two critical elements in strata are its forms of content and expression that establish a complex set of relations between discursive and non-discursive elements (Foucault 48-51). Later, Deleuze argues that the archaeological task is to open up the discursive and non-discursive content of strata, making visible what and who is being spoken (52-60). As Reid’s analysis shows, Foucault’s later work retained but also moved beyond the problematics of archaeology’s focus on what is being spoken. Therefore, it may be useful to buttress Foucault’s dispositif with Deleuze and Guattari’s notion of an assemblage. Similar to the dispositif, an assemblage is a collection of disparate elements, what in physics is referred to as a multiplicity. Assemblages never make up a totality, there is always too much or too little, yet there are usually some elements that have cohered enough to create a contingently stable form.
The alternative I propose is tracing lines of force, following a topological mapping of strategic points of intervention within a dispositif or assemblage. This mode of inquiry shifts from establishing the truth about phenomenon to a Nietzschean exercise of power. As Deleuze reminds us, we shouldn’t ask ‘What is power and where does it come from’, but instead ‘How is it practiced?’; how to develop a practice used “to incite, provoke and produce,…constitute active affects, while to be incited or provoked, to be induced to produce, to have a ‘useful’ effect, constitute reactive affects” (Foucault 71). Fitting with Foucault’s analytics of power, analyses would proceed to develop a strategy. This is not to abandon knowledge, but to understand the force of power and truth that is immanent to its productive capacity.
 Is this also what Foucault in Security, Territory, Population calls a milieu (where he footnotes canguilhem on basically a proto-auto-poesis-esque paper on biology?) [post-MA notes: this is where recent work on Simondon and the ‘transductive’ method of studying ‘onto-genesis’ is making inroads in the social sciences, particularly the geography of space (following Lefebvre’s concept of the ‘production of space’)]