Job candidates' Aggression in chatbot interviews

Abstract

Implementation of a new tool such as a chatbot into the recruitment process profoundly impacts recruiters' jobs, job candidates, and an organization at large. With people spending an increasing amount of time on social media, a chatbot becomes an efficient means of reaching job candidates and conducting job interviews. Chatbots automate processes and decrease the administrative burden. A chatbot proves an effective recruiter in terms of quantity, however, human-computer interactions pose challenges. All nonverbal communication is lost in the process and the interview inherently follows a rather rigid structure. Recruiters need to adopt additional competences; design dialogues and evaluate data of job candidates collected by chatbots. In communication with chatbots, job candidates exhibit a dramatic increase in their use of profane language and aggression compared to a human-to-human interview. An unfavourable impression obtained from social media negatively influences candidates' chances to be recruited. However, little is known about how and to what extent an aggression outburst in a human-chatbot communication influences the employability of the candidates. Our research (i) uses empirical data to measure the levels of job candidates' profanity and aggression in chatbot interviews, and (ii) studies the reasons for this phenomenon from the perspective of chatbot development, psychology and sociology. Identification of these reasons serves as a cornerstone to a new classification for job candidates' aggression towards chatbots; additionally, the classification draws upon a detailed thematic analysis of empirical data. Our primary empirical data are obtained from a leading Czech chatbot developer providing chatbots for an extensive range of companies in the Czech Republic; the data are complemented with structured interviews with recruiters to discuss the effect of chatbot-obtained data in the decisionmaking process in candidate selection. As the data are still being collected, our results are to be considered a work in progress.