Modulation Harms and The Google Home

Abstract

Deleuze’s (1992) modulation is frequently invoked to explain power relations in hyper-connected, sensorised environments. However, attempts to articulate the harmful implications of modulation—a critical step in the process of considering the need for legal intervention—have been modest. In this paper, we theorise four harms arising from the exercise of modulatory power: subsumption, amplification, vibration, and alienation. To do so, we outline the core features of Deleuzean modulatory power (Deleuze 1992), illustrated through contrasts with Foucauldian discipline (Foucault 1995, 1988). Then, drawing on Julie Cohen’s (2013, 2015, 2018, 2019) modulation as a two-way flow of predicted and prescripted modes of governance and knowledge production, we explore and situate our harms in the sensorised and smart home, employing Google’s patented vision as a concrete example (Fadell et al. 2020). We contend that modulation harms arise from the continuous flow and constant agitation of insistent modification (D’Amato 2019) enabled by sensorisation. The core power act that gives rise to modulation harm is the ability to harness, direct, and provide “frequency” to flows of sensor data to achieve continual behavioural modification and shape social norms about the purposes and benefits of such modification. The overarching harm we identify is subsumption, the infrastructural enclosure of all sensorised environments that enables social shaping to take place anywhere, which gives rise to the other modulation harms. Amplification harms regard auto-regulatory norms as an unquestioned facet of an automated human life. Vibration harms arise from the automated ability to prescribe changes in affect. Alienation harms regard subtle denials of access to informational networks. We show that the Google sensorised home both modulates and disciplines occupants concurrently, but more importantly, these concurrent power acts can take place wherever an individual is tethered to the modulation infrastructure and sensor data can be harnessed.