https://orcid.org/0000-0002-5770-1647
![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
ABSTRACT
This study presents one of the earliest empirical investigations on how to harness the power of chatbots for improving key public relations outcomes. Specifically, this study integrates the construct of social presence that has been widely studied in the computer-mediated communication literature with the concept of conversational human voice in public relations to conceptualize chatbots’ social conversation. We evaluate chatbots’ social conversation as an important antecedent driving user perception, not only of chatbots’ listening capability, but also of the organizations’ listening efforts, which, in turn, enhance the essential perceptual outcomes of organizational transparency and organization-public relationships. Our theoretical model was tested through an online survey of 778 adult Facebook users in the US, who were directed to have a 5-minute conversation with a real chatbot. The study results advance the organizational listening literature and contribute to the growing body of knowledge on artificial intelligence in public relations.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1. Note that the accountability dimension of perceived organizational transparency as defined by Rawlins (Citation2008) was not examined in the current study given that accountable transparency pertains to organizations being accountable for their actions, words, and decisions, which requires “persons in transparent organizations contemplate their decisions and behaviors” (Rawlins, Citation2008, p. 75). Such a notion may not be easily discerned by external stakeholders without accumulated communications, interactions or experience with the organization, and judgment of the organizations’ decisions and actions.
2. These five chatbots were selected based on comprehensive secondary research regarding best-performing social chatbots conducted from October 2019 till April 2020 followed by the research team’s independent screening and testing. Specifically, through examining Fortune 100 companies’ and Top 100 Unicorn companies’ (The Global Unicorn Club, 2019) which had Facebook chatbots at the time of our study and reviewing top chatbots reported in the media, we compiled a list of 30 brands’ active Facebook chatbots. These 30 chatbots were then tested by the researchers and rated on a scale of 1 (worst performing) to 3 (best performing). We picked five best-performing chatbots to represent a variety of brands from various industries to include in our study.
3. Specifically, the numbers of respondents randomly assigned to each of the chatbots (i.e., Domino’s Pizza, Jobbot, Toni, Eddy Travels, and Swelly) were 138, 180, 173, 129, and 158 respectively.
4. Prior to the formal data collection, a pretest with 100 respondents using the same procedure was conducted to ensure the reliability and validity of the measures, which yielded no changes to the survey questions. Therefore, the pretest data were included in the final data set.
5. This is comparable to the average usage of Facebook among Americans which is 58 minutes a day in 2020 (Suciu, Citation2021).
6. Byrne (Citation2010, p. 111) argues that “‘forcing large error terms to be uncorrelated is rarely appropriate with real data.’” Allowing error covariance within the same construct can also explain content redundancy. c (SP1, 2) =.22, c (S 5, 6) =.18, c (PCL2, 3) =.23, c (PCL4, 5) =.23, c (PCL5, 7) =.29, c (C1, 2) =. 20, c (C3, 4) =. 51.
7. Furthermore, a separate discriminant validity analysis was conducted using the CFA (merger) Chi-Square difference test (Zait & Bertea, Citation2011) on the two dimensions of chatbot social conversation, social presence and conversational human voice, which showed significant result, χ2 (1) = 21.89, p < .001. The constrained model fits significantly worse, thus supporting the discriminant validity of social presence and conversational human voice.
8. Prior attitudes toward the chatbot showed small significant positive effects on perceived chatbot listening (β = .05, p<.05) and perceived organizational listening (β = .06, p<.01). Prior attitudes toward the organization showed small positive effects on perceived chatbot listening (β = .06, p<.01), perceived organizational listening (β = .07, p<.001) and organization-public relationships (β = .13, p<.001). The organization’s name showed a significant effect on perceived organizational transparency (β = −.03, p<.01).