Digital platforms are complex algorithmic structures connecting dispersed buyers and sellers of services. On platforms such as Guru or Upwork, for instance, customers can search for logo designers or IT developers through algorithms, which provide suggestions about potential matches and allow customers to filter freelancers according to their needs. At the end of the contract, algorithms also ask customers and freelancers to rate their working experience, and, starting from these ratings, compute an aggregate individual score associated to performance. The extensive use of algorithms to manage the workforce is capturing the attention of scholars and policy makers. Whether algorithms are solely constraints to workers’ action, or whether there are ways in which workers can still freely operate, even thrive, when managed by ‘algorithmic logics’, is being increasingly debated in recent literature. Our recent article, published in the Journal of Management Studies, aims to set new light on this debated issue.

What do algorithms on online platforms do?

It is well known that algorithms create power asymmetries (Cameron and Rahman, 2022) and tensions between platform providers, clients, and workers, who are subjected to algorithmic evaluations – i.e., algorithmic computed scores summarizing workers’ past performance and clients’ reviews – whose computing mechanisms are opaque and invisible to the workforce (Rahman, 2021). However, recent evidence highlights that workers develop some knowledge about how algorithms work, for example through peer discussions in forums and online communities (Lehdonvirta, 2016) and they deal with opaque algorithmic evaluations by actively experimenting with new ways to improve scores or by developing practices aimed at complying directly and indirectly with algorithms (Bucher et al., 2021). These findings suggest that, despite algorithmic opaqueness, workers can still exert agency to reduce the asymmetries of an algorithmic-driven environment.

As the online workforce is steadily growing, it is paramount to develop a better theoretical and empirical understanding of the way algorithms work and why online workers behave differently, or how they can thrive under uncertain working conditions.

Exploring reactions to algorithms

In our research, we take an agentic perspective on the way online workers deal with algorithmic scores and investigate, first, how algorithmic scores are interpreted by online workers and, second, the consequences in terms of behavioral and emotional responses. We thus conducted an exploratory study and built a database of 66 interviews with freelancers from a major online labor platform, 66 associated online profiles and 190 articles from the platform’s blog.  

Can algorithms support freelancers’ work? A Technology Affordances perspective

Our results show a key mechanism explaining the heterogeneous freelancers’ interpretations of algorithmic management, i.e., the different technology affordances, or, in other words, what freelancers believe scores can do for them. Specifically, we show that, to some freelancers, algorithms provide a sense of individual visibility and self-extension. Other freelancers, however, perceive them as “rules of the game.” Our grounded model further shows that behavioral and emotional responses vary accordingly. When algorithms are perceived as tools enhancing individual visibility, workers react by trying to manipulate algorithmic computation and preserving their image on the platform. This way, they are also able to keep their emotions in check. When algorithms are perceived as an expansion of their working selves, online workers indirectly deal with algorithmic computations by nurturing their relations with clients to confirm their positive selves on platforms. This proves to be a stressful situation, though. When algorithms are interpreted as a rule of the game, they decide to ignore scores and try to do their best with clients. However, they face a paradoxical situation in which they feel frustrated, as they have not developed specific behaviors to deal with algorithmic scores.

Finally, our findings reveal that workers change their responses over time. In particular, they adjust their behaviors according to the perceived affordances only after an initial period of compliance to algorithmic rules, that is, the time when they are new to the platform and lack positive reviews. During this initial period, algorithmic scores are perceived as barriers, thus, instead of providing opportunities for action, they constrain workers’ actions.

A comprehensive model on the relationship between freelancers and algorithms

While previous research on algorithms has emphasized issues of control and opacity, we adopted an affordance perspective and proposed that the way gig workers respond to algorithmic management depends on the possibilities and constraints they see in algorithmic scores. Furthermore, even if our findings confirm that negative feelings, such as frustration or anxiety, are likely to arise, our affordance perspective allows us to underscore when such negative feelings happen, and when, instead, gig workers’ experience becomes bearable, even exciting.

In terms of practical implications, this study shows that the way algorithmic scores and algorithmic rules are currently designed is very likely to produce deviant behaviors hampering either workers or the score system itself, leading freelancers and clients to distrust it. Although scores on these platforms can be useful tools, they tend to be ambiguous and subjective, despite being designed to be objective and meritocratic. Our practical implications address platform providers, then, and our suggestions are twofold. We first suggest sensitizing clients to how the score system can profoundly impact workers’ experience and encouraging a fair use. For instance, our informants revealed they accepted lower payments in exchange for good reviews, which contributes to making their online experience unfair and hampering the platform’s reputation. Second, we suggest deemphasizing the importance of scores on freelancers’ profiles and allowing workers to personalize the design of their CV or other quality signals.

References

Bucher, E. L., Schou, P. K., and Waldkirch, M. (2021). ‘Pacifying the algorithm – Anticipatory compliance in the face of algorithmic management in the gig economy’. Organization, 28, 44–67.

Cameron, L. D. and Rahman, H. A. (2022). ‘Expanding the Locus of Resistance: Understanding the Co-Constitution of Control and Resistance in the Gig Economy’. Organization Science, 33, 38–58.

Lehdonvirta, V. (2016). ‘Algorithms That Divide and Unite: Delocalization, Identity, and Collective Action in ‘Microwork’’. In J. Flecker (Ed.), Space, place and global digital work, London: Palgrave Macmillan, 53–80.

Rahman, H. A. (2021). ‘The Invisible Cage: Workers’ Reactivity to Opaque Algorithmic Evaluations’. Administrative Science Quarterly, 66, 945–988.

IMAGE: https://unsplash.com/photos/H7yW_lVGJuI