Week 6

Week 6: 2006/10/2 - 2006/10/6

Synthetic Characters with Emotional States

Avradinis, Nikos, Themis Panayiotopoulos, and Spyros Vosinakis. "Synthetic Characters with Emotional States." Lecture Notes in Computer Science: Methods and Applications of Artificial Intelligence. Berlin: Springer, 2004. 505-514.

In this article Avradinis, Panayiotopoulos, and Vosinakis discuss why synthetic characters need to have emotions, and how to create a emotion system in a synthetic character. Emotions are an important part of creating synthetic characters because they need to be able to respond appropriately , and to accomplish this they need an understanding of emotion. The authors give the example of 3D models in a virtual environment. Not only will users expect them to look realistic, but they will expect them to also behave consistently with their own internal attributes as well as outside inputs. To generate these emotions they propose a system based on their work with SimHuman and Carroll Izard's theories about emotion activation. They propose a three layer architecture of cognitive, non-cognitive, and physical layers. Actions are generated in either the cognitive or non-cognitive layers and can be projected (acted) out through the physical layer. A few things they note that their design doesn't implement, and neither do other designs, is the idea of dynamically creating rational processes in the characters as well as spontaneous behavior.

---

A Model for Personality and Emotion Simulation

Egges, Arjan, Sumedha Kshirsagar, and Nadia Magnenat-Thalmann. "A Model for Personality and Emotion Simulation." Lecture Notes in Computer Science: Knowledge-Based Intelligent Information and Engineering Systems. Berlin: Springer, 2003. 505-514.

In this article the authors express that what is missing from synthetic characters is individuality, what drives them. They present mathematical functions to represent personality, emotional state and emotional state history, and mood, as well as functions to update these categories. Beyond that, there is very little of value in this article.

---

Why We Play Games: Four Keys to More Emotion Without Story

Lazzaro, Nicole. "Why We Play Games: Four Keys to More Emotion Without Story." XEODesign Inc. 2004.

Lazzaro, in this article, discusses why we play games, what makes playing video games fun, and the emotions these kinds of fun elicit in us. Through her research with XEODesign she has broken fun down into four different categories and the emotions these forms of fun elicit. Hard Fun is the satisfaction that comes from being sufficiently challenged and overcoming that challenge and evokes emotions such as frustration and triumph. Easy Fun is about players discovering the world through immersion, exploration, and adventure. Easy Fun provokes wonder, awe and mystery for these players. Altered States moves the player from one mental state to another, making the player feel different. Finally the People Factor is about the social aspects of video games. Whether it's cooperative team work, or competing against other players, these players derive pleasure and pride from such activities.

Social characters have the potential for playing a role in the enjoyment people get out of playing video games. They could add an additional layers of challenge, exploration, depth and competition to video games.

---

Changing personalities: towards realistic virtual characters

Poznanski, Mike, and Paul Thagard. “Changing personalities: towards realistic virtual characters.” Journal of Experimental & Theoretical Artificial Intelligence, Volum 17, Issue 3, Sept. 2005. 221-241

Poznanski and Thagard discuss their personality model, SPOT (simulating personality over time), that satisfies the needs of both psychology and computer science, namely the game development industry. The criteria they used to design this are: psychological plausibility and realism, simplicity and efficiency, and interesting and varied model behavior. Using the Java Neural Network Simulator (JavaNNS) they’ve implemented a three layer forward feeding neural network. These layers are an input layer, a personality/emotion layer, and an output layer. The layers contain nodes and are connected through links between the nodes. Each node and link has a value, when a situation occurs the input is evaluated through these nodes and corresponding links to determine the characters behavior. The interesting thing about this model is that they’ve also implemented a mechanism that allows the character’s personality to change over time based on the situations the encounter but also take into account their “genetic dispositions”, the personality it started off with. So for example, a person who is disposed to being disagreeable but encounters many positive situations will only be able to obtain a certain level of agreeableness. These personalities, the rate at which they change, and the node and link values are all customizable, allowing each character to be different.