Positivism v. realism
From InterSciWiki
Contents |
[edit] Introduction
This is an A289 seminar essay -- so free to chime in and collaborative-edit with your own views, observations and questions -- by Doug White with my riff on on where most of network analysis has gone wrong from Clyde Mitchell to the present, with a quick stop at cog anthro. It also references an Introduction I wrote to the network analysis book with Ulla Johansen and a new piece on human social dynamics.
[edit] Positivism v. realism in Anthropology
The use of positivist here is not intended to characterize individuals, but approaches to uses of methods. One article of the same author might fit the rubric described below by Murray Leaf, another not. Certain generic approaches might be more positivist than not, but some works within it might be more realist. In our readings for week 4 the first part of Christenson and Read is more positivist (according to the critiques) and the second part is more realist. The contrast is worth explaining. Earlier in his article, Leaf gives the history of positivism, from Compte and Spencer forward, and its 20th Century ups and down, including logical positivism. The "politics" of positivism is not of interest here, rather, it in the contrast between method imposed and method fitted to understanding of the problem, including direct elicitation of idea systems that constitute what Leaf includes empirical formalism. One not agree with Leaf's argument for a purely Kantian approach to see the contrast: for me as a theoretical realist the approach agrees in many ways with Leaf's "empirical formalism" but is considerably broader in its span. Bourdieu's description of his work is another extension, quite different, from a base in Leaf's extension of Kantian approaches to the social sciences (something missing totally, oddly, in Kant). But when you hear phrase such as "science is scientific method" as if the method was known a priori you can be fairly sure you are in the territory of positivism. Russ Bernard's (2006:18) use of positivism in one of our textbooks -- also published as an e-book is somewhere in between, cleaned up from his earlier usage. His statement on logical positivism is relatively unobjectionable, as is his explanation of how the Hayek's "instrumental positivism" of the 1950s and attack on "scientism" provoked a justified reaction (p.19).
Murray Leaf (2007 Empirical Formalism Structure and Dynamics 2(1):804-824) p. 5:
- "all varieties of positivism involve some version of Comte’s and Mill’s conception of form as inherently arbitrary and non-substantive and all consider theory to be preeminently formal in this sense, not all varieties are prominently associated with a program for “formal analysis.” In philosophy, formal analysis is primarily attached to the varieties of logical positivism and its close allies. In anthropology, it has been primarily associated with ethnoscience, componential analysis, and now cognitive anthropology — and arguably also with Levi-Strauss’s “structural” analysis and structuralism."
- In the ethnoscience of the 1960s and 70s, the positivistic, referential, theory of meaning was clear and explicit. The focus was on taxonomies, and taxonomies were treated as sets of terms (identified morphologically) and the sets of objects that the terms referred to. Each “object,” in turn, was a set of “features” —the counterpart of Mill’s sensations or sense impressions. The possibility that ideas—as such—might have an essential role to play in this process was minimized or ignored. As applied in componential analysis of kinship terminologies, the method was to arrange the terms of the taxonomy and their supposed referents in such a way that “no term overlaps or includes another; every component is discriminated by at least one term; and all terms can be displayed on the same paradigm” (Wallace and Atkins, 1960: 62). The “formal” order in the words, then, was taken as coming from the “formal properties” of what they referred to, on the assumption that each term could have only one referent in this sense — an assumption that is not in fact true."
- "Proponents of contemporary cognitive anthropology reject the earlier ethnoscientists’ positivistic insistence that science can only study “action” or “behavior” and now say they are focusing on “idea systems” (D’Andrade 1995: 12), but the disagreement is superficial. The underlying methodological assumption is still that meaning can only be referential. In consequence, the concept of “idea systems” is still actually the same as when ideas were nominally ignored. They are still not what people actually think, per se, but rather the words (or, in some cases, drawings or images) they use and what in the analyst’s view those words or images are “really about” in terms of the scheme of supposedly “formal” categories that the analyst imposes. D’Andrade describes the aim of the method as to produce an “emic” analysis, and: To carry out an emic analysis one began with a set of categories brought in by the scientific observer and then tried to find out which of those categories really made a difference with respect to the way the natives understood and responded to things. (D’Andrade 1995: 18-19)"
- "What is such an analysis an analysis of? What is it describing? Is it the observers’ categories? Is it the observer’s categories that make some difference in the behavior of the natives? Is it the native’s behavior that the observer’s categories pertain to, or some of each of the above? The critical question is why we should think that any categories “brought in by a scientific observer” should make a difference to the way natives understand and respond to things. And more simply, why should we start with such categories at all? Why not start with the natives’ own categories, find out what they are, and then see what difference they make?"
A parallel critique is made in White and Johansen's (2004) Introduction: Networks, Ethnography and Emergence from Network Analysis and Ethnographic Problems: Process Models of a Turkish Nomad Clan, namely that in spite of all their advances:
- The Manchester school approach led by Mitchell (1969) did not envision the more general possibility of embedding anthropological problems in a network approach in which the network data is suited to the problem. Their views of the contribution of network data and network analyses were highly restrictive, rarely if ever rising to the level of interactions between multiple networks in different domains or at different scales. Even for community studies, the methods of choice were institutional analyses that were wedded to structure-functionalism or functionalist varieties of conflict theory. Mitchell and others in his group failed to see network studies as providing contributions in this context that institutional analyses could not provide. The presumption stemming from structure-functionalism was that shared culture and a stable social structure were intrinsic to social life in traditional rural communities, barring periods of change and adjustments like those studied by Turner (1957). It was only with migration, mixing of populations, multiethnic groups, industrialization, and globalization that they recognized pressures toward change and rapid adjustment as features of social life that they thought required network analysis if only because of the fluidity of these processes. Members of the Manchester school tended to treat networks as special types of structures that required a distinct toolkit rather than as a more general and flexible ontology for situating social theory. The net-work approaches of the 1960s and early 1970s were abandoned by practicing ethnologists well before many of the newer network modeling approaches had developed. For most anthropologists, the uses of network concepts reverted to those of the earlier structure-functional period in anthropology, as metaphors for social relationships.
The problem here again was a form of positivism: they emphasized a plethora of formal measures with which they demonstrated correlations to specific problems but failed to put forward (except for Victor Turner) an approach to how networks fit, for example, larger institutional configurations. Networks, for Mitchell and many of his colleagues, could not supply, for example, an alternative approach to standard institutional or community analysis.
[edit] Network positivism today
We see this again in the network tutorials of Steve Borgatti. His glossary, for example, while incomplete, discusses concepts in the following terms
- Structural embeddedness. A tie between nodes A and B is embedded if there is a third party C that both are connected to. The idea is that the presence of C affects/constrains/enables the relationship between A and B. The more third parties connected to both A and B, the more embedded is the tie between A and B. See also Simmelian tie.
- Embeddedness. See relational embeddedness and structural embeddedness. Normally refers to a property of ties between nodes. Occasionally (and confusingly) also used to refer to a property of nodes (in which case it is a vague allusion to being in the thick of things, such as having high eigenvector centrality).
These definitions never rise above the triad (triadic measures) and attribute confusion and vagueness to the possibility of a property of nodes (like structural cohesion or alternative definitions -- available from 2003 -- of structural embeddedness) having implications for linking individual to the properties that they might inherit by membership is groups (for which centrality is an inappropriate measure).
This might seem like a trivial point, but it is not. When Moody and White submitted their structural cohesion and embeddedness (2003) article, several reviewers repeated their belief that group membership properties had to be tested against other group-level phenomena and it was logically impossible for individuals to possess properties by virtue of membership in groups that could be related to or tested against other individual properties.
Borgatti's Group Cohesion & Shape modules repeat this kind of logical error. The slides under "graphcohesion.pdf (new!)" exhibit the first awareness (2008) of work done in 2003 that won the ASA Mathematical Sociology prize on this subject. Measures discussed are (like the Mitchell measures, but expanded):
- Density & variations
- Average tie strength
- Average degree
•Number of components •Fragmentation / connectedness •Size of largest clique •Cliques per node •Connectivity / independent paths / max flow (the illustration shows no comprehension of the concepts or of Menger's theorem)
- Dyadic connectivity cohesion (which breaks up the phenomena in a way that destroys group properties, as computed in UCInet - this measure from White and Newman 2001). Instead, cohesion is decomposed into components of
- Average distance
- Diameter
- Breadth
This repeats the same problem as componential analysis: the notions of measurement supercede the logic of the actual problem of cohesion. The actual problem of cohesion is taken on at the logical level by Menger and at the substantive level by Moody and White (2003), Powell, White, Koput and Owen-Smith (2005), (and a host of other references)..., and in historical dynamics by Peter Turchin (here again a host of references), as newly summarized in Dynamics of Human Behavior (2008, Douglas R. White, preprint, Encyclopedia of Complexity and Systems Science.
[edit] Ethnographic and Network Realism
Check out:
Network Realism: Process Modeling and Network Analysis
see also Work on Reason
Back to Weekly riff
