Working for an Algorithm: Power Asymmetries and Agency in Online Work Settings

The opacity of machine learning algorithms creates a black box that challenges professional groups' knowledge claims and prevents them from questioning AI judgements and outputs.

“In particular, the opacity of machine learning algorithms is seen as creating a 'black box' which represents a challenge to the knowledge claims of professional groups and prevents them from questioning the judgements and outputs produced by AI (Curchod et al., 2020).”

The impact of using algorithmic management to set goals on the reduction of workers' autonomy has been outlined by many scholars, namely in gig work where autonomy in task selection is often viewed by externals as a major advantage of platform-based work.

“The impact of the use of AM to set goals on the reduction of workers' autonomy has been outlined by many scholars, namely in gig-work where autonomy in task selection is often viewed by externals as a major advantage of platform-based work (Curchod, Patriotta, Cohen, & Neysen, 2019;Gerber & Krzywdzinski, 2019;Griesbach et al., 2019;Rosenblat, 2018;Shapiro, 2018).”

Algorithmic goal setting is negatively associated with job autonomy.

“Goal setting Autonomy (-)”

Platforms rely on machine-learning algorithms to filter and control uploaded content.

“platforms rely on machine-learning algorithms to filter and control uploaded content”

Other studies have found that content creators are 'leveraging algorithms' and 'working around the algorithm' by constantly uploading, A/B testing, and diversifying their incomes.

“Other studies have found that content creators are ‘leveraging algorithms’ (Kellogg et al. 2020, p. 391) and ‘working around the algorithm’ (Curchod et al. 2020, p. 17) by constantly uploading (Arriagada and Ibáñez 2020), A/B testing (Cotter 2019), and diversifying their incomes (Cutolo and Kenney 2021).”

Algorithmic systems facilitate high contingency of performance scores on customer satisfaction ratings in gig work and traditional work settings.

“Moreover, algorithmic systems facilitate high contingency of performance scores on customer satisfaction ratings, both in gig work (Curchod et al., 2019;Gerber & Krzywdzinski, 2019;Griesbach et al., 2019) and in traditional work settings such as restaurants (O'Donovan, 2018), retail (Evans & Kitchin, 2018;Levy & Barocas, 2018), or telecommunications (Bakewell et al., 2018).”

The strong impact of customer satisfaction ratings has been associated with workers' perception of job insecurity.

“The strong impact of these ratings, largely questioned about their volatility, the reliability and validity (Gerber & Krzywdzinski, 2019;Greenwood, Adjerid, & Angst, 2019;Luca & Zervas, 2016;Orlikowski & Scott, 2013;Rosenblat, Levy, Barocas, & Hwang, 2017), has been associated with workers' perception of job insecurity (Curchod et al., 2019;Veen et al., 2019).”