When Predictive Algorithms Break: How an Airline Improvised its COVID-19 Response

This is a lunch seminar; please register your attendance by accepting/declining your emailed invitation by Tuesday, April 2nd, at 10 AM at the latest (for catering). For more information please click here.

Abstract
We investigate how a firm’s reliance on digital data and predictive algorithms shapes improvised response and subsequent adaptations to an unexpected crisis. Our inductive study of an airline facing the COVID-19 pandemic illuminates how seemingly fragile digitalized routines provide building blocks for improvised responses. We show how diverse units quickly improvised non-technological routines by leveraging material “fragments” of existing information systems and shared understandings, followed by progressive revision of routines to regain a predictive orientation. In our case organization, the data-centric organizational culture and a central position in interconnected routines endowed two units with authority and induced them to device technocratic controls to cope with continued uncertainty. These units obtained privileged roles in interpreting and coordinating the pandemic response and used their established understandings and digital infrastructure to improvise new practices. Our study contributes to research on organizational design by articulating how digital infrastructure and related culture shape improvisation. We also nuance the received view of algorithmic breakdown as a liability during unexpected crises, noting how related digital infrastructure facilitates coordination. More broadly, we argue that despite the “entanglement” of digital technologies and organizational routines noted in the literature, experts may often be able to “untangle” their practices materially from information systems. In contrast, shared understandings embedded in those systems appear to be more persistent.