Building Customer Understanding: The Deal of Chatbots and Designers

• In Event: Automation, Skill and Identity in an Age of AI - II

Abstract

In this paper, we propose to focus on chatbots developed in customer relationship ; these « chatter » robots, which have no body avatar - unlike physical robots - are indeed programs of anticipated dialogues between a simulated agent and a human. These devices spreads on brand websites or mobile instant messaging application ; they often are developed by brands and made increasingly available to consumers, including in their personal lives. Based on an ethnography of two projects of chatbot development, this paper demonstrates how these devices, which closely hybridize human activity and algorithmic activity, ultimately redistribute machine skills with human collaboration. Half objects, half subjects, these automatons intermediaries based on commercial conversation (or rather dialogue in current language), evolve with technologies of natural language processing. They have the particularity of providing multiple choices (dialog trees) according to the decision of the semantic recognition system ; these dialog trees are then linked to several « intentions » (wich are semantic groups) according to an algorithm… But the automatic system ends there. Actually, these automatons are not as automatic as they seem : equipped with humans supervisors for their constant evolution, these talkative machines commonly referred to as "intelligent" machines, actually require a constant, meticulous and manual work of readjustments : they learn as they interact with human Internet users. Thus, these new hybrid intermediaries are shaking up models of "scalability" and questioning the ideal of AI applied to customer service as a continuation of industrialization in the "relational" domains, hitherto attributed to humans.

Author

• Charlotte Esteban, Université Toulouse