| Fueled by technological advancement in artificial intelligence,conversational software agents(CAs)in the form of chatbots continues to radically and rapidly replace human service agents as a preferred option in organizational front-line(Adam et al.,2020),changing the nature of service and customer experiences.By means of turn-by-turn text-based natural language processing,live chat interfaces(e.g.,IKEA's Anna,Jing Dong's JIMI)provide a newly cost-and time-saving solution for service providers,resulting in widespread application in various industries(Gnewuch et al.,2017;Pavlikova et al.,2003).From simple tasks such as sending reservation information to complex assignments including giving financial advises(Araujo,2018),conversational software agents have gradually played a significant role on those social media and messaging platforms.Despite the overwhelming implementation of AI-based chatbots,resistance and skepticism from demand sides never diminish and even raised in recent years as emerging technology may trigger a sense of strangeness as well as insecurity,potentially resulting in users' declination to cooperate(e.g.,Elsner,2017).With great business potential and customer's pushback,the critical challenges in chatbot's application highlights the significance of understanding how to promote the design of these AI-based agents in particular the anthropomorphic level and the social presence attributed to the chatbot(Araujo,2018).Therefore,scholars and IT developers have long been working on studying the anthropomorphic aspects of the artificial service agents by deploying human-like design cues to further enhancing human-robot relationship and offering better customer experiences in interacting procedures.Prior literature regarding artificial intelligence anthropomorphism has been primarily focused on embodied conversational software agents with physical body or virtual faces,whereas verbal cues that predominately used by text-based chatbots remain under-explored.Meanwhile,extensive research on the relationship between anthropomorphic design cues and customer attitudes and perceptions are inconclusive.Humanoid robots with the application of anthropomorphic design cues(i.e.appearance,language style and etc.)are anticipated to inspire trust,customer's engagement and compliance,resulting in more pleasant service outcome(Rietz et al.,2019;Cassell et al.,2001;Go and Sundar,2019).An alternative theoretical lens,however,recommends that artificial intelligence with high level of anthropomorphism may trigger averse reactions in users by elicit feeling of eeriness and discomfort proposed by Uncanny Valley theory(Moosa and Ud-Dean,2010).In addition,empirical investigations on chatbot have only recently sparked researchers' interest in some business areas such as retail marketing and hotel reservation,much more service contexts have not been reached including service failure management.The disruption during the service delivery process(Huang et al.,2020)affected customers not only at that time points but also the following service recovery process with strong emotional fluctuation,resulting in negative customers' feelings,evaluations and behavioral intentions.Against the above backdrop,the goal of this research is to study the impact of chatbots with verbal anthropomorphic design cues(versus chatbots without the cues)on customer behavioral intention during service failure recovery process.Herein,we mainly focused on the verbal aspects of text-based chatbot to avoid the disturbance of any nonverbal cues such as embodiment or facial expression.Concretely,we firstly explored whether or not CAs having verbal anthropomorphic design cues(VADCs)compared with CAs without the cues can generate more positive effect on customer's cooperation intention after the breakdown and before service recovery completed(H1).Secondly,we went deep into the underlying mechanism in behind by examining how social cognition dimensions in the form of perceived warmth and competence may be influenced by VADCs,which in turn lead to higher level f customer cooperation intention(H2).Afterwards,our research added situation factors and individual characteristics to study the specific conditions that maximize the usage of VADCs.Extensive researches demonstrated that customer's viewed outcome failure and process failure differently with distinctive reactions based on how they were involved during the service failure process(Albrecht et al,2019;Smith et al.,1999).Since outcome failure usually results in the loss of core service value whereas process failure posits threat to self-esteem caused by the deficient or inferior service delivery,customers experience a series of cognitive process with different needs have been hurt before interacting with the chatbots.Therefore,we hypothesized a moderating effects of service failure types existed on the relationship between CAs with VADCs(versus CAs without VADCs)and customer perceived warmth and competence,inter alia,the positive effect of CAs with VADCs(vs.none)on perceived warmth will be stronger under process failure than in outcome failure situation(H3a)and the positive effect of VADCs on perceived competence will be stronger under outcome failure than in process failure(H3b).Furthermore,customer's technology readiness(TR)level as a personal trait-like factor was also investigated with the predication that the TR level,when jointly considered with service failure types strengthened the impact of CAs with VADCs(versus CAs without VADCs)on customer perceived warmth and competence(H4).Thus,the present research examined the impact of verbal anthropomorphic design cues,service failure types and individual technology readiness level in the context of service failure recovery process.To test these predications,a series of online survey were carried out using scenario-based questionnaires.Altogether,three studies with the sample of 218,200 and 402 respectively were conducted in which participants were either recruited via survey platform or invited by acquaintances on social apps.Study 1 provided the initial test for the main effect between CAs with anthropomorphic design and customer cooperation intention,supporting the proposition that application of human-like cues on live interfaces are beneficial to customer's service encounter.The output of Study 2 verified how VADCs on CAs influenced the perception of warmth and competence compared with non-anthropomorphic situation,which in turn resulting in more positive customer's behavioral intention.We examined the predication proposed by van Doorn et al.(2017)and concluded customers' sense of social cognition in terms of perceived warmth and perceived competence related to the verbal anthropomorphic cues on CAs will influence its effect on customer's social cognition aspects,as the underlying mechanism.Finally,Study 3 supported hypotheses 3a,3b and 4,examining the moderating effect of service failure types and technology readiness level,whereby respondents were randomly assigned into 4 different scenarios(outcome-VADCs;process-VADCs;outcome-non VADCs;process-non VADCs)with different background information and the manipulation of VADCs.Customer's technology readiness level was collected through self-evaluation questions developed from previous literature(Parasuraman and Colby,2015).The result of hierarchical regression analysis demonstrated the moderating role of service failure types as well as technology readiness,implying TR level moderates the two-way interaction effects of VADCs and service failure types on customer perceived warmth and competence.Consequently,we found out the positive effect of CAs with VADCs(vs.none)on customer perceived warmth will be the strongest under process failure and when customer technology readiness is high;the positive effects of CAs with VADCs(vs.none)on customer perceived competence will be the weakest under outcome failure and when technology readiness is low.This research provides valuable avenues for future study especially on the design of anthropomorphic cues and the following applying procedures in specific service context.Through empirical study,we identified whether or not verbal anthropomorphic design on CAs should be placed and under what circumstances the application can be most effective,considering service failure types and customer's general attitude towards cutting-edge technology.The finding of the more positive effect generated by VADCs on customer cooperation intention contributes to the on-going debate of artificial intelligence's anthropomorphism,supporting the application of verbal cues are beneficial to improve service encounter.Meanwhile,the underlying mechanism regarding the perception of warmth and competence spotlights chatbot's distinctive ability to engage customer on a social level in service recovery processes,indicating the great potential of artificial intelligence as the replacement of human-employee.The last but also the main theoretical contribution of our paper regarding the moderating effects of technology readiness level,which jointly with service failure types amplify the effect of VADCs on perceived warmth and competence.We not only derive first insights into underlying the effect of technology-based agent's anthropomorphic design in the area of service failure management,but also can be viewed a catalyst to future research by revealing the three-way interaction effects of VADCs,service failure types and individual technology readiness level.Apart from theoretical contributions,we offered valuable practical insights for enterprises and IT developers,particularly in the field of service recovery process,the significant but under-explored area.We suggest the perception of warmth and competence can be raised by chatbots with VADCs(versus chatbot without VADCs)and in turn lead to more positive customer's service recovery cooperation intentions,supporting the incorporation by service organizations of cutting-edge innovations into frontline service.Responding to the notion of 'Technology Dominated Market 2025',this article contributes to the effective usage of verbal anthropomorphic design in real practice. |
PDF Full Text Request