The Role of Reuse in the Successive Implementation of Conversational Agents

Abstract

More and more companies rely on the implementation of conversational agents (CAs) to automate certain of their processes. While CAs' initial development often mirrors an innovative process, their successive implementation can be made more efficient by drawing on other CAs' prior developments. We rely on data from a case study where a series of chatbots (which represent one type of CAs) was implemented. Routine theoretical concepts help us better understand how CAs may dynamically evolve and how their implementation can be accelerated. We found that a) the reuse of emerging or intentionally constructed means may allow accelerating the implementation of multiple successive CAs, b) means can be reused to extend or transfer functionality (through mutation or inheritance); and c) changes in the conversational context determine if means can be reused directly or not (through reproduction or recreation).