Lev Manovich, "Data," in Critical Terms in Futures Studies, ed. Paul Heike, Palgrave, 2019.
From the article:
"How is a data representation of some phenomenon or process different from other kinds of cultural representation humans used until now, be they representational paintings, literary narratives, historical accounts, or hand-drawn maps? First, a data representation is modular, i.e., it consists of separate elements: objects and their features. Secondly, the features are encoded in such a way that we calculate on them. This means that the features can take a number of forms—integers, floating-point numbers, categories represented as integers or text labels, spatial coordinates, etc.—but not just any form. And only one format can be used for each feature.
In other words, today “data” is not just any arbitrary collection of items existing in some medium such as paper. In a computational environment, “data” is something a computer can
read, transform, and analyze. This imposes fundamental constraints on how we represent anything.
What is chosen as objects, what features are added, and how these features are encoded—these three decisions are equally important for representing phenomena as data and, consequently, making them computable, manageable, knowable, and shareable through data science techniques.