Avaliação de interfaces de usuário em chatbots de serviço público e o uncanny valley

Abstract

The work developed is an analysis of the use of chatbots for governments and public services in communication and service to the public. For this, it was used a set of heuristics for conversational agents adapted from Nielsen’s heuristics recommended by Langevin et al., (2021) and the semiotic inspection method, using strategies and classes of signs proposed by Valerio et al.,(2020), to understand the presence of uncanny valley effects in conversational interfaces and communicative ruptures in the interaction between human agents and machines. At first, chatbots used by the public sector in Brazil were surveyed, and after filtering, a classification based on different parameters such as type, architecture and mechanisms for formulating responses was performed. From the data obtained after the categorization, the chatbot of the Government of the State of Ceará, Plantão Coronavírus, was chosen to undergo the interface evaluation methods and investigate its performance in communicating with users during interactions. It was found the effectiveness of the methods in chatbots with usability problems and the potential efficiency of using the results obtained in the evaluation for a conversational design with more realistic expectations and fewer communicational gaps, which would also reduce effects of Uncanny Valley, the phenomenon about a person’s emotional response to a human-like robot changing abruptly from empathy to repulsion as the machine approaches, but fails to achieve, a realistic appearance or response.