Don't waste a good crisis! Finance, Machine Learning, and the Pursuit of Productive Paranoia

Abstract

The last decade has seen a number of flash and hack crashes that exposed how already volatile stock markets are increasingly integrated with other social domains, resulting in entirely unstable media ecologies. For instance, the AP hack crash exposed how entwined technologies that mine twitter data and high frequency trading systems are, and what effects the increasing automation of information processing and trading has already wrought (Karppi and Crawford 2016).

This development has infinitely complicated prediction and aggravated uncertainties that characterized markets ever since. In environments where volatility is the source of productivity and growth as well as of catastrophe and trauma, secret information, hidden patterns, affect and contagious processes now take center stage—and so do practices and technologies conceived to manage and exploit the related, mutating uncertainties.

My paper is of conceptual nature and zooms in on the pursuit of paranoid ideation as a technology in both contemporary machine learning and finance. I draw on sociologies and anthropologies of finance and trading (Arnoldi and Borch 2007; Maurer 2002; Zaloom 2007) to carve out the significance of quasi-paranoid ideation in the market. I argue that paranoid tendencies among traders and in markets more generally are exacerbated in cutting-edge machine learning systems that trade. Mechanisms designed to prevent the becoming pathological of information processing therefore make for an object of study that has implications beyond the narrow confines of computing. I will provide some insights into the design of cutting-edge machine learning systems and work towards a technology-based approach to the libidinal economies of contemporary capitalism.