Ryan Timms

Cognitive scaffolding refers to external structures that change the cognitive demands of tasks. This dissertation begins by reviewing the literature around scaffolding—discussing key thinkers such as Hutchins (1995a), Clark (1997) and Sterelny (2003, 2010). I then develop the above scaffolding characterisation by drawing a distinction between shallow and deep scaffolding. Shallow scaffolding primarily involves cues, whereas deep scaffolding involves the significant offloading of cognitive work. By appealing to the complementarity thesis, I show that deep scaffolds can be explained through a model of cognitive extension (Menary, 2006; Sutton, 2010). Most crucially, despite the abundance of benign/ neutral examples in the literature, I argue that scaffolding can also be ‘hostile’ to agents’ interests—benefitting one agent while undermining another. I draw on Sterelny (2003) to clarify my use of the word ‘hostile’ and also specify what I mean by ‘interests’. I then review authors with similar concerns to my own—i.e., the lack of discussion around external structures that negatively impact agents (Aagaard, 2020; Liao & Huebner, 2020; Slaby, 2016). Despite these authors’ concerns, the idea of ‘hostile scaffolding’ is not quite reached. I then present examples of shallow hostile scaffolding by reviewing how sunglasses (Viola, 2022) and casino interior design and ambience (Friedman, 2000; Schüll, 2012) can be hostile. Most importantly, I show that deep hostile scaffolding is a genuine concern (and not only a theoretical one) by reviewing gambling machines that use player tracking systems and virtual reel mapping (Schüll, 2012), as well as Twitter’s use of gamification (Nguyen, 2021). I close the dissertation by discussing avenues for future hostile scaffolding research. This includes the ethical implications deep hostile scaffolding, scaffolding’s role in forming addictive behavior, other instance of gamification, ‘racist scaffolding’, and hostile scaffolding in developing technologies (such as VR, dark patterns and AI).

Most accounts of cognitive scaffolding focus on ways that external structure can support or augment an agent’s cognitive capacities. We call cases where the interests of the user are served benign scaffolding and argue for the possibility and reality of hostile scaffolding. This is scaffolding which depends on the same capacities of an agent to make cognitive use of external structure as in benign cases, but that undermines or exploits the user while serving the interests of another agent. We develop criteria for scaffolding being hostile and show by reference to examples including the design features of electronic gambling machines and casino management systems that hostile scaffolding exists and can be highly effective. In cases where the scaffolding is deep and permits the offloading of significant cognitive work, hostile scaffolding exploitatively manipulates cognitive processing itself. Given the extent of human reliance on scaffolding this is an important and neglected vulnerability. [This is an updated version of the paper, added in February 2023.]