Text Mining as Creative Différance

• In Event: Critical Data Studies: Ethics and Human Contexts of Data Science I

Abstract

Text mining (TM), with its wide array of contributing disciplines, methods, interests, and histories, remains a field pulled in different directions. As a result, the philosophical foundations of TM remain undetermined. Nonetheless, TM has regularly been defined as an endeavor of discovery, specifically, the computation of new ideas from extant texts. The imaginaries of TM and its closest relatives (information retrieval, natural language processing, data mining, and artificial intelligence) are entwined with the imaginaries of digital capital, the Internet, and computation. This family of imaginaries, fueled by discursive regimes of exploration, colonization, warfare, espionage, criminal science, and profit, have made TM possible but have also limited its possibilities. TM’s characteristic doublespeak of discovery and the double consciousness needed to sustain it gives rise to a similarly limited AI, risking the development and proliferation of a narrow if not altogether dangerous model of human intelligence that at best constructs innovation entirely out of translation, neglecting invention in the process. Through an examination of TM (Claude Shannon, Vannevar Bush, John Tukey, and Marti Hearst), the author argues that TM can be re-framed in creativity rather than discovery. By acknowledging and appreciating TM’s status as a “trading zone,” and elevating the primacy of making, the author reimagines TM through continuous chains of meaning-making, partial knowledge, iteration, creativity, augmentation, and collaboration. Such reimagining leads to more robust tools and methods for a wide array of disciplines and endeavors. Examples of such tools and methods are presented.

Author

• Patrick Herron, Duke University