Human Complex Systems

This paper describes methods used at the interface between anthropology and machine learning research. Charles Kemp, a graduate student at MIT in 2004, discovered my numerically coded Alyawarra kinship term applications data (Denham 1973; Denham, McDaniel and Atkins 1979; Denham and White 2005) and received my permission to use the data in his machine learning research. Since then, his co-authored papers (Kemp et al. 2004, 2006, 2010), and other works that cite his papers and mine, have played significant roles in the development of unsupervised pattern detection and machine learning technology as subsets of Artificial Intelligence research. Part 1 of the paper outlines how I produced the Alyawarra (Alyawara) kinship term applications dataset and introduces the structure and content of the dataset and supporting files. Part 2 briefly describes some simple ways to analyze the dataset either manually or with machine learning technology. Minimally these examples demonstrate some ways in which the ethnographic dataset is useful to the machine learning community now. More speculatively, the machine learning technology introduced here may enhance ethnographic research in the future. Part 3 provides links to a sample of 24 papers by Kemp et al. and other AI colleagues, all of which utilize the Alyawarra Kinship dataset. Part 4 contains links to some of my Alyawarra kinship data and documentation files that are available online. Part 5 briefly acknowledges support that I have received for this project over the last half-century. 

In this article, I critique Edward O. Wilson’s (2019) Genesis: The Deep Origin of Societies from a perspective provided by David Christian’s (2016) Big History. Genesis is a slender, narrowly focused recapitulation and summation of Wilson’s lifelong research on altruism, eusociality, the biological bases of kinship, and related aspects of sociality among insects and humans. Wilson considers it to be among the most important of his 35+ published books, one of which created the controversial discipline of sociobiology and two of which won Pulitzer Prizes. Big History is Christian’s recent attempt to graphically depict the history of the universe in a massive, sprawling, well-documented volume that opens with the Big Bang and terminates now, about 13.8 billion years later.

I take four disparate approaches to enhance the strengths of Wilson’s and Christian’s important books. Part 1. Expanding the past examines 1. contextual data for numerous transitions in sociality in the distant past, and 2. ethnographic data pertaining to kinship and warfare in Australian Aboriginal hunter-gatherer societies in the recent past. Part 2. Imagining the future speculates about 1. predictive applications of sociality research as we approach another mass extinction in the near future, and 2. social research concerning globular star clusters in the remote future. Small scale case studies feature, among other things, two species of colonial microorganisms, the Alyawarra speaking people of Central Australia, and social insects as a background for all else. Although Wilson’s extensive quantitative research deals mainly with kinship and related topics among ants, bees, wasps and termites, it is not limited by time, space or species.

This is a quantitative analysis of the replication of Dreamtime traditions among the Alyawarra of Central Australia in 1971-72. A narrative summary presents observational data recorded during the enactment of a 108-hours long tone poem that embodied oral traditions, songs, dances and visual arts as aides-mémoire that facilitated the synthesis and persistence of a reliable society comprised of unreliable people. The tone poem, presented by 69 men and women, marked the beginning of one young man’s lifelong education in the all-encompassing Aboriginal Dreamtime. A tabular summary follows the narrative summary and describes demographic, genealogical, kinship and other quantified relations that were embedded in the narrative and that young men were required to learn before they could marry and sire children. The paper ends with a discussion of the two summaries that together shaped the education of young Aboriginal men. Instantaneous scan sampling and unsupervised pattern detection formed a reductionist research strategy for finding points of entry into an otherwise impenetrably complex alien civilization in which male circumcision was a major feature. The paper does not pretend to be an exercise in explanation, but it has major implications for identifying what needs to be explained. When observations and descriptions are problematic, formulating testable theories of human behavior is doomed from the outset. 

Read’s work on the generative logic of kinship terminologies constitutes one of the most distinctive and stimulating series of publications in the contemporary anthropology of kinship.  His intention to produce a universally valid explanatory (i.e. causal) theory of kinship terminology is highly ambitious – but also appropriate and intellectually refreshing. An important feature of his theoretical framework is that it allows for an interaction between universal cognitive processes and local cultural ideas. Another distinctive feature is that Read usually models whole terminologies – and specific features, such as crossness and generational skewing, are understood in the light of the terminological system as a whole. Read has been continually testing and refining his conceptual apparatus, and in this paper he brings it to bear for the first time on Crow-Omaha systems – offering us an exploratory case study that is intended both to show the insight that the generative logic approach can bring, and to investigate the specific logical features that may give rise to the phenomenon of skewing. In this comment I will look at Read’s approach in quite a general way, embedding my specific comments  on his analysis of Thonga kinship within this more general review.

Read’s formal analysis of kinship terminologies is well known and widely respected, as is his leadership in promoting the formal analysis of kinship through the formation of panels and conferences, and his role in this journal. As in all his work the paper is strongly reasoned and draws upon a knowledge of the literature that is long and deep.  All of these are reasons we welcome the piece before us.  On the other hand, the spirit of this work is somewhat different from that of our book, engendering in us some reservations.  Taking the strengths for granted, we will confine our comment to a couple of things in Read’s article with which we take issue.

This is a valuable contribution to the ongoing debate about Crow-Omaha and the overall approach to analysis of kinship systems.

That Omaha patterns can arise for different reasons in different societies is certainly a possibility, and it is good that Read presents here a comparison which claims to show that. I do agree with Read that a formal analysis is a necessary first step before continuing with an explanation of the causes of the specific form of kinship structure in a specific group. The key dichotomy proposed is that between Crow-Omaha as a direct consequence of an inherent generative logic and as resulting from a transformation of the terminology.The idea that there is a major difference between the causes of the two types is attractive but this conceptualisation of it is problematic. It implies that the Thonga-Ronga system is not due to a transformation, so perhaps has remained the same from a very ancient time. But no evidence bearing on this is offered, nor is it likely that no transformations have been at work.

This is the latest of a series of papers on the generative deep structures of kinship terminologies by Dwight Read, which considerably widens the spectrum of methods and concepts employed hitherto. I will therefore discuss it in the context of Read’s more general project to develop a theory and typology of kinship terminologies based on the process of their generation, by concentrating on three main arguments: (1) the newly introduced difference between symmetric and asymmetric deep structures; (2) the use of cross-sex kin terms as gender-switch operators; and (3) the interpretation of generational skewing as an effect of generative asymmetry.

It is always a pleasure to read any of the ongoing elaborations of Dwight Read’s framework for the formal analysis of kinship terminologies. This exploratory case study of the Thonga-Ronga terminological system raises a number of important issues concerning the specificities of skewing as well as the encompassing methodology of Read’s generative logic program for kinship analysis. I first comment on Read’s framework for the analysis of a kinship terminology’s generative logic. I then argue (as other respondents have) for embedding the analysis of Thonga-Ronga skewing within the context of a more locally constrained field of comparison. I conclude with specific suggestions for comparing kinship models and their underlying generative logics as variants and transitions situated within an abstract morphospace. 

Read´s research program for describing the “generative logic” of distinct kinship terminologies in a homogeneous framework has proved its fruitfulness in different ethnographic domains, ranging from North American kinship to Dravidian terminologies, and more. Applied now to the so-called Omaha systems, the framework suggests a new taxonomy of kinship terminologies, in which Thonga kinship terminology – until now a type specimen for the Omaha terminology, based on Junod´s ethnography – is separated from Fox kinship terminology, another type specimen of the Omaha,  as described by Dorsey, and Morgan before him. Read´s thesis, therefore, subverts Lounsbury´s subdivision of “Omaha” taxon in four varieties, among which “Type I” was instanced by the Fox terminology, while Type III had Thonga data as a standard representative. It is not my intention to refute Read´s representation of the logic underlying Thong kinship terminology, expressed in diagrammatic form, but, rather, to suggest that there is more than one way to represent it.

The seven commentators, Thomas Trautmann, Peter Whiteley, Patrick McConvell, Patrick Heady, Franklin Tjon Sie Fat, Klaus Hamberger, and Mauro Barbosa de Almeida, have provided wide-ranging and important observations that go beyond the specifics of my text and bring to the discussion important issues that relate to our understanding of the Crow-Omaha terminologies. Their comments alone provide a major contribution to the discourse on the Crow-Omaha terminologies. Accordingly, my response to their comments focuses on ways that the structural analysis I presented of the Thongan kinship terminology relates to this broader discussion.

I have divided my reply into seven parts: (1) Relationship of Abstract Algebras to Kinship Terminologies, (2) Other Methodologies: Thick Description, Equivalence Rules, Description and Extension, (3) Ethnographic Issues Relating to The Algebraic Representation, (4) Comments by Patrick McConvell, Patrick Heady, and Franklin Tjon Sie Fat, (5) The Formalism Issues Raised by Klaus Hamberger, (6) The Formalism Issues Raised by Mauro Barbosa de Almeida, and (7) Conclusion -- Why Does Ñwana(‘Son’) oMalume(‘Mother’s Brother’) = Malume?