The AI Triple Whammy

The gig economy is a rapidly growing global labour paradigm involving casualised labour across sectors as varied as transportation, freelancing, domestic help, programming and vacation rentals. The gig economy has been studied extensively within social science, political, legal and allied circles. It has been widely observed that gig economy, powered by myriad AI algorithms, has led to poorer working conditions and a downward pressure on wages. The focus of many such studies is split across governance and administration of algorithms on the one side, and consequences of algorithmic control on the other. Consequently, it could well appear that the underlying gig AI infrastructure is not necessarily blameworthy and may be amenable to ‘dual use’. This leaves room for optimism that the very same gig AI algorithms may be leveraged for alternative and responsible operation under models such as platform cooperativism. In this article, I consider whether the observed worker harms are substantively traceable to choices made at the domain-specific algorithm design stage, consequently exploring the possibility that the seeds of the harms are baked into the algorithms, undermining their intrinsic potential for alternative or responsible use. Focusing on two key gig sectors of transportation and freelancing, I consider three algorithmic tasks within the gig economy and three key forms of worker harm. Given the proprietary nature of gig AI and their consequent opacity, I adopt an interdisciplinary and critical desk research approach examining extant scholarly and other published literature across AI and social sciences and assemble evidence to derive insights. My analyses indicate strong support in favour of the assertion that worker harms are significantly traceable to algorithm design choices embedded within gig AI; I characterise these connections into separate streams of worker behaviours that are predicated by algorithm design which in turn lead to harms. This suggests that effectively developing responsible and convivial gig models would require not just innovations in governance and administration, but also technological research for the development of new and appropriate algorithms. Finally, I outline some technological possibilities for the development of such alternatives and responsible algorithms.