In short, DeLanda argues for a 'realist' conception of society where the social analyst's conceptions meet something intransient and real -- i.e. where the sociologist has a concrete referent object to analyse, not where said sociologist just builds up what society is through hypotheses, ideal types or signifiers. Having said this, DeLanda then immediately acknowledges that in some cases it is not this simple; he uses the case of a refugee as an example of an instance in which the social scientist may affect the referent object of study by altering the system of classification by which 'refugee' qua social subject is constituted. This happens and it is important; however, DeLanda argues, it is the exception rather than the rule. Most object of sociological inquiry are not reflexive in this manner (1-3).
On his blog, Alex Reid picks up this idea from Levi Bryant's introduction to DeLanda's introduction and writes:
[T]here is an interesting reflexive quality to a social realist ontology that it different from an ontology of natural objects. That is, calling a tree a tree doesn't impact what the tree is. Calling a plant a weed, doesn't change the plant (though it may change the way people react to the plant). On the other hand, the way we name things in a social milieu can be cybernetic. For example, students who become labelled as smart or troubled or whatever can tend to take on those roles.In response Levi writes:
This reflexivity is one of the key features of the social. Social entities are capable of relating to the manner in which they are described, such that their description modifies their nature through this relation. If my doctor, for example, diagnoses me as suffering from depression, I might do research on depression and begin emulating some of these descriptions.
All of which reminds me of an essay called 'Social Thought and Social Action' by the philosopher Martin Hollis that appeared in a book called The Social Dimensions of Science in 1992. In particular I am reminded of what Hollis called 'double' and 'triple hermeneutics.'
"Guinea pigs do not read books", begins Hollis, "Biologists do." Humans react to their observation in a way that non-humans do not; in short, they are reflexive beings. This is a "universal social fact" (68). I think we can all agree that guinea pigs, in fact, do not read books. But this highlights something rather important. For Hollis the reflexivity of social beings divides the world in two: on the one hand, there are intelligent, reflexive humans who, when observed, necessitate 'double' or even 'triple hermeneutics' as we are then layering interpretations on interpretations on interpretations (hence leading to 'double' and 'triple hermeneutics'); on the other, there are mute, prostrate non-humans largely oblivious to their manipulation and utterly incapable of affecting the observer in response to the observer's manipulation.
Both Alex and Levi imply (perhaps they would not agree, but this is how it seems from the discussion so far) that reflexivity is a defining characteristic of sociality and that only social beings can be properly reflexive. DeLanda does perhaps imply something like thus but I do not think this is the right way to interpret him at this point.
DeLanda asserts that some social relationships are reflexive and others are not. Social relationships are reflexive when interpreting them affects the original phenomena. Writing policy guidelines for government refugee legislation is, therefore, a decidedly reflexive act as the category of 'refugee' is being reshaped as it is being observed. If I am sat watching a news report about refugees in Australia, for example, this is not reflexive in this way as my experience has no (or at least negligible) impact on the category of 'refugee.' Fair enough, but none of this restricts reflexivity to the social or, indeed, makes societies defined by reflexivity.
Firstly, DeLanda is clear that reflexivity is the exception rather than the rule. Secondly, I see no reason why, for example, an assemblage of gases locked in a relatively stable cycle of disequilibrium (so that the assemblage is constantly changing but cycles back to repeat itself relatively consistently) cannot be said to be 'reflexive'. Certainly it is not of the order of sentient beings attaining self awareness but nor is it totally different. Various beings in an assemblage react to other beings in such a way that all their properties are altered over time.
For Hollis humans are reflexive and non-humans are not, therefore humans and non-humans are completely different (reality is thereby duly bifurcated). For DeLanda some human relationships are reflexive and some are not and we already know that humans and non-humans are not altogether different. On that basis, I would rather put it this way: reflexivity is a possible property of all assemblages, not just social ones. Although social, human assemblages display unusually enhanced properties of reflexivity this does not place them ontologically apart from less reflexive or even un-reflexive assemblages. All can develop reflex mechanisms of one kind or another.
In this reading, human/human, human/non-human and non-human/non-human assemblages may all attain 'reflexivity' or they may not. It is an open question.