1.1.1 Methodology. In this paper we use the Viable System Model (Beer, 1979, 1981, 1985) – VSM- and the VIPLAN Methodology (Espejo, 1993; Espejo & Reyes, 2011); they help us to discuss the braiding of organisational learning (Espejo, Schuhmann, Schwaninger, & Bilello, 1996) and technological processes. We support model and methodology by a systemic epistemology, which highlights holism, in particular the relevance of communications, interactions and complexity. More specifically we adapt the Viplan Methodology (Espejo & Reyes, 2011) to the use of BDA.
The emphasis is in the interactions and relationships of agents at multiple levels, from the global to the local. We use a systemic epistemology that highlights structural determinism in organisations and structural coupling between agents and actors (Maturana & Varela, 1992). Structural determinism highlights the autonomy of organisational systems; it is the closure of their structures that determines which environmental data makes sense within the organisation. Structural coupling highlights the history of communications and interactions between agents and actors leading to the structural congruence between them.
Big data is produced by the huge number of transactions natural to all situations. The problem is their management. Crucially to focus on relational aspects we use Ashby's Law of Requisite Variety (Ashby, 1964)and the ideas of variety operators to balance performance at satisfactory levels. Dealing with data requires considering how they are absorbed by the structures affected by them, as well as their responses. It is in absorption that the structural, ethical and technical issues of big data and people come together.
It is not always the case that an enterprise shows the property of closure necessary for a desirable autonomous behaviour in its environmental context. To improve its structure we focus on processes of individual and organisational learning. Individual learning is increasing their capacity to take effective action and organisational learning is increasing effective action in their environments. Among several factors restricting this learning are poor models of these environmental situations. "Every good regulator of a system must be a model of that system" (Conant & Ashby, 1970). But it is not useful to be a good regulator of a poorly structured situation, hence the duality of structure and data models that we explore in this paper. Overcoming structural fragmentation helps making data more meaningful to those affected by the contextual changes.