Having believable interactions between synthetic characters and with people is crucial to the creation of immersive interactive experiences. The believability of such experiences is strongly connected to having the characters express emotions with the correct timing throughout their interaction while retaining the human participants’ agency. Unfortunately, most video games adopt static animations, which usually only convey a single emotion throughout, making their expression look artificial, or allowing for greater emotion expressiveness but sacrificing the player’s agency. With this work, we propose the creation of a model based on the principles of animation and anticipation that allows for emotion expression during gameplay while retaining the participants’ agency. We evaluate our approach using three models. In the first model, characters express emotions in reaction to what happens (this is similar to how games approach emotion expressions), in the second in anticipation of what could happen, and in the third in reaction to what happens given what was anticipated. We tested these models in Adfectus, a 3D arena fighting game, and found improvements in some dimensions relating to the believability of interactions in particular player demographics.

To better understand how emotions are recognized, we propose the Triad Affect Interpretation (TAI) method that creates a model of how users perceive animations of emotions of specific synthetic characters, with the additional intent of improving the way emotions are communicated so that they are more easily recognized. The method is divided into two main steps and focuses on a set of animations of emotions taken from the synthetic characters under development. The first step aims to elicit meaningful constructs through content analysis. The second step asks participants to rate the same animations against the selected constructs. Using principal component analysis and cluster analysis, we then create a model of the relevant factors to the perception of emotions that can inform future improvements of the animations related to the expression of emotions. The results of a user study suggest that our characters' expressions of Anger and Disgust are confused with each other due to a similar contraction of the face; and Fear and Surprise are confused due to a similar expansion of the face, which provides specific directions for improvements.

When designing interactive scenarios that depend on emotion expression, it is imperative to consider the levels of recognition associated with said expressions, to ascertain whether or not an acceptable degree of emotional communication has been achieved. In this work, two experiments were conducted with that aim, one asking participants to compare two different versions of an application side-by-side when conveying a specific emotion, and another asking the participants to recognize the emotion being expressed in each version. We found that, for some emotions, the approach rated higher in terms of emotion expression during the side-by-side comparison would not translate to the approach with a higher emotion recognition in the second experiment. Although this discrepancy is generally consistent with what happens with emotion recognition in humans, it is noteworthy that some higher-rated choices ended up not being as effective in the expression of emotion. We discuss how these discrepancies might have originated from forced-choice and feature dominance, and why context should be taken into account when designing experiments.

Long distance learning has always faced several challenges in regards to motivation and availability of opportunity for students. While the latter has seen efforts in terms of collaborative and non-traditional learning avenues, such as online courses and virtually hosted classrooms, the former still suffers due to external influences and responsibilities, which in turn lead to an increase in student's disinterest levels. To combat these hindrances, we propose the creation of an intelligent virtual agent based on the Social Regulatory Cycle, to fill the role of a coach in a readily available mobile application. This virtual coach is capable of supporting and adapting to the needs of human students, helping them in their tasks based on their intentions, motivations and feelings. By gathering objective and subjective user data, the virtual coach is able to interpret learning situations of individual users and, with the use of a synthetic character with analogous mannerisms to those of an actual human, help them by modulating their emotional state in the context of the learning process. Preliminary evaluations performed within an online learning environment class showed users continued interest in engaging with the virtual coaches on a regular basis.