Contribution
The paper argues that Lévi-Strauss’s transformational method, long dismissed in Anglo-American anthropology as “pseudo-mathematical mystification”, can be made operational with a discrete-event simulation formalism, and presents the authors’ software for it. Beyond the argument, it adds a geographic visualisation: a myth’s mythemes, given coordinates by hand, are displayed in narrative order on a map. The transformation-generating machinery is summarised from their 2020 paper.
Key insight
The paper’s own: treat a myth as a network of components (mythemes) whose connections can change during a simulation, so that a transformation of a myth is a change of model structure, which DEVS with dynamic variable structure supports. In the software as described, which change to make is supplied by the user.
Assumptions
- That Lévi-Strauss’s myth analysis is an algorithmic, generative system (“myths are something like informal algorithms”).
- That a myth decomposes into mythemes, each a pair of a term and a function, and that a variant is obtained by operations on those pairs.
- That the canonical formula, with a boundary condition external to it, captures how myths transform across the boundaries between peoples.
Key results
- DEVS models defined (summarised from the 2020 paper): MythGen, TransformationADEVS, Collector, Supervisor, MythemADEVS and Observor; a myth as a coupled model of mytheme atomic models; generation of a new myth (M2 Bororo) from the reference myth (M1 Bororo) by the operations held in TransformationADEVS; a graph of generated myths (Fig. 5).
- Visualisation models (new here): ViewerGen reads a text file of mythemes and a second, hand-made file of coordinates; MythVisu adds a PointMyth instance to the MythVariant coupled model for each mytheme; SIGviewer writes a KML placemark (Algorithm 1); Google Earth reloads the KML periodically, and a step-by-step plugin advances the narrative.
- Validation. The 20 mythemes of U Lurcu, from the Nebbiu in northern Corsica, are placed on the map in narrative order (Figs. 11–14). The simulation’s time base is not historical time.
- Programme. A neo-structural model of identity transformations, combining discourse analysis, Bayesian inference and probabilistic DEVS, proposed for EU and US politics.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | DEVS with dynamic structure can represent a myth as mythemes and apply transformation operations to generate variants | moderate (as a software claim) | the model inventory; details deferred to the 2020 paper |
| C2 | The software displays a myth’s mythemes on a map in narrative order | strong (demonstration) | Figs. 11–14, Algorithm 1 |
| C3 | This validates or revitalises Lévi-Strauss’s transformational theory | not supported here | no prediction of the theory is tested; the transformations are user-specified |
| C4 | The method can automatically validate anthropologists’ hypotheses and analyse many myths | not supported here | asserted in the introduction; not shown |
| C5 | The approach can anticipate identity transformations and inform policy | not supported here | future work |
Method
Discrete-event system specification: atomic models (states, transition and time-advance functions) composed hierarchically into coupled models, run by an abstract simulator; dynamic structure by a supervisor atomic model that rewires a coupled model during simulation (after Baati et al. 2007). Implemented in Python 2 in DEVSimPy, with plugins for step-by-step runs and map display through KML.
Concepts
- mytheme: the basic element of a myth, characterised by a term (a person, animal, divinity or thing) and a function (its role).
- transformation (of a myth): a variant obtained from another by operations such as inversion, opposition, homology or symmetrization; all variants form a group with no privileged member.
- boundary condition: in the authors’ reading of the canonical formula, the external condition of crossing a spatial, linguistic or social boundary under which the “double twist” transformation occurs.
- DEVS / DSDEVS: discrete-event system specification; its dynamic structure extension.
Connections
The paper’s base is Santucci, Doja and Capocchi (2020), “A Discrete-Event Simulation of Claude Lévi-Strauss’ Structural Analysis of Myths Based on Symmetry and Double Twist Transformations”, Symmetry 12(10):1706, not held. It reads Lévi-Strauss through Mythologiques and the canonical formula (Lévi-Strauss’s 1955 article “The Structural Study of Myth”, later collected in Structural Anthropology, LIT-775, unread), and through mathematical readings (Petitot, Maranda, Morava). It cites Propp’s Morphology (English 1968; LIT-828) among earlier formalisations of narrative. Descola’s lecture (LIT-822) is not cited.
Bearing on the record
- On computational modelling of Lévi-Strauss as prior art. The paper shows that Lévi-Straussian transformations have been implemented as software operations on mytheme representations. It does not show that the software discovers transformations, decides between analyses, or tests any structural hypothesis against a corpus: the operations are supplied, and the validation is a map display of one folktale. Whoever cites it as a computational model of Lévi-Strauss’s analysis should say it is a framework for encoding analyst-specified transformations; the 2020 Symmetry paper, unread, is where the transformation model itself is described.
- Descola’s point (LIT-822) that in myth the analyst cuts the transformation continuum applies here unaltered: the software inherits the analyst’s cuts.
- No THEORY is filed. No instruction for machine-learning practice; nothing for the anthology.
Limitations
- No corpus beyond one reference myth (named) and one folktale (mapped); no evaluation, no comparison with any other analysis.
- The transformation engine is summarised, not specified, here.
- The geographic coordinates are entered by hand.
- Much of the text is a defence of Lévi-Strauss’s legacy and a programme for future work.
Open questions
- Given a corpus of variants, can such a system recover the transformations an analyst proposes, or propose ones the analyst did not, with some criterion for preferring one over another?
- Does the 2020 paper’s double-twist implementation reproduce a published Lévi-Strauss analysis end to end?