Experimental study on the effect of adopting humanized and non-humanized chatbots on the factors measure the intensity of the user's perceived trust in the Yellow September campaign

Abstract

The article aims to identify if there is a difference between the groups that interacted with the non-humanized and humanized chatbot from the Yellow September campaign and if these factors measure the intensity of the user's perceived trust. The experiment built two chatbots with different characteristics but with the same purpose to inform participants about the campaign. The researchers used Facebook Messenger to conduct the experiment, resulting in a sample of 511 participants. A survey was conducted, and the data obtained was analyzed in the following stages: EFA, experiments (moderations, ANOVA, and GLM), and ANN-based models as a post-hoc analysis. The first one proved that all the analyzed constructs had a high significance level when analyzing the difference between the two groups. We observed that similarity to human, perceived competence, and satisfaction had higher results in the group that interacted with the humanized chatbot. In the moderation analysis stage, expertise proved to be a moderating variable in the chatbot's relationship with perceived competence and satisfaction. Users of the humanized chatbot have greater perceptions of trust than users of the non-humanized chatbot. Notably, humanized chatbots demonstrated significantly greater trust across all factors examined.