Data

A selection of code material will be collected relating to each case-study. This might be source code directly taken from the algorithmic ecology itself. However, due to the proprietary nature of much of this code, most material will be approximations or emulations of the case-study in question. Again using Uber as an example, it is relatively straightforward to execute a generic pathfinding algorithm similar to their inaccessible custom code. Unlike computer science, this analysis is not obsessed with optimization (fastest, most accurate solutions). Instead, the primary concern is the algorithmic ontology, the objects and actions available in the ‘code-world’. These are defined by data types and functions, initiating particular parameters and abilities while suppressing others. This rather restricted software notion of ‘ontology’ can be broadened into its philosophical counterpart, allowing us to analyze this structure not simply as a means to write ‘clean’ code, but as a way of understanding the world. While this code-world ontology is clearly defined, its scope and visibility is necessarily bounded. This is similar to the notion of simplification in actor-network theory (ANT), in which an “infinitely complex world” is reduced to a series of “discrete entities whose characteristics or attributes are well defined” (Callon, 1987, p. 93). This poses an obvious question: what is missed in this reduction? How is the highly complex, highly contingent ‘outer world’ of labour, logistics, bodies and minerals approximated and anticipated by data structures? What slips between these hard-edged ontologies?