UC Santa Barbara

The ways that people construct their identities at work organize their social actions. Considering that so much of people’s work increasingly involves technology, a significant body of research has focused on what people do with and through technology to construct distinct and desirable identities. But emerging developments in artificial intelligence (AI) make it possible for technologies to be part of identity construction in a way that has not yet been explored: by acting autonomously on behalf of the people who use them. Individuals’ growing reliance on intelligent machines in their work requires reconceptualizing identity construction as a process that people share with technologies themselves, which I call a joint action perspective of identity construction. I argue that two capabilities of AI technologies—their capability to learn from aggregated data and their capability to make decisions autonomously—shape identity construction processes when these technologies mediate people’s interactions. I explored these possibilities by conducting a comparative field study of two AI scheduling technologies created by different companies. I followed these technologies from development into use by working professionals to investigate how the capabilities of AI scheduling technologies shape how they are involved in identity construction and with what consequences. Through a set of three studies, my findings illuminate how people’s outsourcing of work to AI scheduling technologies that facilitate communication and communicate on their behalf shape their identity construction. In the first study, I showed that AI scheduling technologies carry out new work practices on users’ behalf, shaping how they enact identities, which users accepted or resisted depending on how these practices were presented and whether they helped users enact identities to which they had aspired. In the second study, I showed that communication partners can form impressions of AI agents that communicate on users’ behalf, which sometimes were transferred to partners’ impressions of users’ identities when communication partners did not have a strong relationship with those users. Users worked to manage their own and communication partners’ relationship with AI technologies in order to maintain or construct desirable identities for themselves. In the third study, I showed how one AI technology made it difficult for users to enact multiple identities, because the AI technology was optimized only for a narrow range of outcomes and limited users’ flexibility to draw on different identities at different times. I discuss the theoretical and practical implications of these findings for individuals’ identity construction in the age of intelligent machines.