Skip to main content
Research project

Anthropologic Algebra

Project overview

WSI Pilot Project
This project aims to support experimental computational development for greater cultural meaning in AI. The work involves exploration of structuralist anthropological literature, notably the algebraic interests of Claude Lévi-Strauss, alongside practical contemporary AI and big data methods, particularly non-negative matrix factorisation (NMF).

Key to the project is a re-reading of the influential anthropologist Claude Lévi-Strauss, whose work was highly influential in the mid-twentieth century, offering sophisticated approaches to classification models, systems of thinking and cultural meaning. His canonical formula is a serious attempt to organize a series of variants into a permutation group, from which to formulate a law of that group. There are various reasons why this work was not developed further.

In the last few decades, the key shift in AI has been the ‘statistical turn’ (Li 2017; Crawford, 2021), shifting from expert, rule-based systems, and instead, drawing on the availability of massive datasets for innovations in new probabilistic modelling. As a starting point, the proposed project considers these recent techniques to renew connections with Lévi-Strauss’ structuralist methods. While contemporary probabilistic models of AI are highly effective in constructing meaningful syntax (and image models) there remains a problem with the recognition of deeper ‘cultural’ patterns of meaning. In analysing and sorting myths from around the world, a key metaphor for Lévi-Strauss’ was that of the musical score, whereby he would seek to draw out deeper level ‘arrangements’ of meaning (or ‘bundles of relations’). The project will explore and test these such arrangements. Thus, via a series of scoping workshops as well as additional and preparatory research, the project will:

(1) Examine the work of Lévi-Strauss’ canonical formula, and draw upon recent interpretations;
(2) Review contemporary AI methods and specifically non-negative matrix factorisation;
(3) Draw together the understanding of both (1) and (2) to develop speculative and prototype mathematical models in extension of the structuralist methodologies;
(4) Map findings across contemporary contexts of AI, including, but not restricted to, recent developments of GPT-3 image-processing transformer switching.

Staff

Lead researcher

Professor Sunil Manghani

Professor of Theory, Practice & Critique

Research interests

  • Visual Art & Culture
  • Image Studies
  • Critical Theory
Connect with Sunil

Collaborating research institutes, centres and groups

Research outputs