Interações entre humanos e agentes virtuais : efeitos da combinação de comunicações explícitas e implícitas

• Ano de defesa: 2022
• Autor(a) principal: Ana Christina Almada Campos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais
• Palavras-chave em Português: Engenharia elétrica; Interação homem-máquina; Robôs; Sistemas de comunicação
• Link de acesso: https://hdl.handle.net/1843/46156

The field of human-robot interaction has become more important with the increasing application of robots and virtual agents interacting with humans. Communication between humans and robots (or virtual agents) is essential for interaction and is often inspired by human communication, which uses gestures, facial expressions, gaze direction, and other explicit and implicit means. This works aims to compare the use of only explicit communication with the combination of explicit and implicit communication, and observe performance measures and people’s perceptions, contributing to the human-robot communication literature. A multimodal communication infrastructure with systems for human gestures and gaze direction recognition and a virtual agent playing the role of a robot was developed. An interaction experiment similar to a game was proposed and the observed variables were time and number of errors in the task execution, acceptance of the virtual agent, the virtual agent’s sociability and transparency, and perception of efficiency of the interaction. The results were evaluated with frequentist hypothesis tests and Bayesian parameter estimation. Measures more related to the task (time, number of errors and perception of efficiency of the interaction) appear not to have been influenced by the type of communication, which can be attributed to the specific characteristics of the proposed interaction. However, it was observed that the combination of explicit and implicit communications can improve people’s acceptance of the virtual agent and its sociability and transparency. Improvements in the experimental protocol could help in the investigation of variables more related to the task, and information obtained with the Bayesian estimation of parameters associated with the interaction can be incorporated in future studies.