Human-Algorithm Collaboration in Gig Work: The Role of Experience, Skill Level, and Task Complexity

Adopting machine learning (ML) and algorithm-enabled decision-support tools in real-world settings is often challenging, particularly in gig work environments where the usage cannot be easily enforced and outcomes depend heavily on the human capital of workers. In this paper, we study how an ML-based solution (i.e., “wayfinding” technology) that provides real-time support for locating and picking products affects the performance and work patterns of gig workers. Using data from a large-scale randomized field experiment on the Instacart platform, we examine how worker experience, skill level, task complexity, and workload moderate the effectiveness of human-algorithm collaboration. Our findings reveal that ML and algorithm-enabled decision support provides significant performance enhancements by improving work efficiency and quality by 3.29% and 3.83%, respectively, particularly among more experienced workers. In fact, the real-time decision-support tool does not diminish the value of experience, but rather, complements it in our context. Conversely, we also find that the ML and algorithm-enabled technology can substitute for workers’ skill levels, especially in helping lower-skilled workers bridge the gap with their more skilled colleagues. However, lower-skilled workers need experience to fully benefit from the ML- and algorithm-based decision-support solution. Moreover, the extent to which the decision-support tool complements experience is influenced by task complexity, with the tool playing a more significant role in complex tasks and when workers experience a high workload. Our comprehensive analysis also indicates that human-algorithm collaboration leads to a 3.16% increase in the volume of work on the platform. In addition, it improves the flexibility of workers, allowing them to handle tasks across multiple stores. To this end, ML and algorithm-enabled technology leads to a relative increase of 32.5% in the likelihood that shoppers will work in multiple stores. Notably, the study underscores that human experience remains vital in shaping the work patterns of the gig workforce when influenced by the ML and algorithm-enabled technology. These insights emphasize the need for companies to balance technological and human factors, such as skills and experience, to optimize the benefits of human-algorithm collaboration for blue-collar workers, ensuring that new tools complement, rather than replace, valuable human experience.

History: Bin Gu, Senior Editor; Gordon Burtch, Associate Editor.

Funding: The authors thank the Fishman-Davidson Center for Service and Operations Management and the Baker Retailing Center, both at the Wharton School, for providing funding for this research.

Supplemental Material: The online appendix is available at https://doi.org/10.1287/isre.2024.1664.