Autonomous Assessment of Videogame Difficulty Using Physiological Signals

• Autor(a) principal: Rodrigues, Pedro Mendes
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/153932
• topic: Affective Computing; Emotion Assessment; Physiological Signals; Virtual Reality; Videogames; Domínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e Tecnologias

Given the well-explored relation between challenge and involvement in a task, (e.g., as described in Csikszentmihalyi’s theory of flow), it could be argued that the presence of challenge in videogames is a core element that shapes player experiences and should, therefore, be matched to the player’s skills and attitude towards the game. However, handling videogame difficulty, is a challenging problem in game design, as too easy a task can lead to boredom and too hard can lead to frustration. Thus, by exploring the relationship between difficulty and emotion, the current work intends to propose an artificial intelligence model that autonomously predicts difficulty according to the set of emotions elicited in the player. To test the validity of this approach, we developed a simple puzzle-based Virtual Reality (VR) videogame, based on the Trail Making Test (TMT), and whose objective was to elicit different emotions according to three levels of difficulty. A study was carried out in which physiological responses as well as player self- reports were collected during gameplay. Statistical analysis of the self-reports showed that different levels of experience with either VR or videogames didn’t have a measurable impact on how players performed during the three levels. Additionally, the self-assessed emotional ratings indicated that playing the game at different difficulty levels gave rise to different emotional states. Next, classification using a Support Vector Machine (SVM) was performed to verify if it was possible to detect difficulty considering the physiological responses associated with the elicited emotions. Results report an overall F1-score of 68% in detecting the three levels of difficulty, which verifies the effectiveness of the adopted methodology and encourages further research with a larger dataset.