How is the data processing chain able to generate valid information on the objects of study? With the multiple steps involved from “raw” data (which are already constructed from such objects) to final representations, it is surprising that analysts’ discourse on objects of study can still be related to the objects themselves. An important theory to solve this epistemological problem was coined in (Latour 1995) with the “chains of circulating reference”. By observing how scientists transform the soil of Boa Vista forest into scientific facts, Bruno Latour has remarked that scientific studies follow a series of transformations, each one going from matter to forms by creating a gap: forms lose material properties, but gain semiotic properties related to that matter. In this perspective, reference is a property of transformation chains which depend on the quality of transformations. Such chains can conduct truth only if they remain reversible, i.e. changes can be traced back and forth so that valid reference circulates along chains without interruption.
The circulating reference was originally illustrated by Latour on the Boa Vista study. We revamp his schema on the figure below, in an attempt to apply it on the processing chain of graph data. We will see in the next blog post how augmented data (which is part of this chain) generated by data mining algorithms may hasten visual analysis.