Culturally Affected Behavior (CAB)
The CAB research effort seeks to develop an approach for authoring modular cultural representations and applying those cultural representations to affect how AI characters behave and relate to other characters, objects and the world around them. For this project, a character's cultural identity will influence physical appearance, internal knowledge and reasoning and external behavior. Cultural identity will be used to select an appropriate three-dimensional model for the character as well as a graphic texture to put on that model. This will allow culture to influence the virtual character's style of dress, hair color and style, facial hair style, skin color, and apparent age. The character's primary cultural identity will also influence aspects of internal reasoning, including attitudes, assumptions and biases towards other people and things, what goals are selected, and even emotional coping strategies. External behavior will be influenced by culture both indirectly through the internal reasoning and directly through the selection of culturally-specific animations (for example a greeting animation), body language, facial expressions, verbal language and accent. The CAB effort has focused on defining the problem; conducting an initial literature survey of a wide variety of fields including artificial intelligence, modeling & simulation, sociology, anthropology, psychology and international business; establishing collaborations with partners with USC and within the Army; and creating a technical approach and prototype.
The CAB prototype, based on ICT's virtual humans, is integrated with the ELECT BiLAT application. In the prototype the human player can meet with Farid, the Iraqi policeman, or Fritz, the German policeman. The cultural differences between these two characters are represented as different states and tasks and different weights across a set of socio-cultural norms. Future efforts will continue the work on individual cultural behaviors, including evaluation and experimentation, and additionally explore population-level culturally-affected behavior.
This project differs from others in the following ways:
The CAB effort distinguishes itself from other approaches to modeling culturally-affected behavior in that it is:
- based on theories from social psychology and sociology
- modular; allowing cultures to be "swapped" in and out of a character
- deep; modeling both surface elements of culture (appearance, gestures, accent) and the deeper, cognitive elements of culture (cultural norms, taboos, attitudes, biases, stereotypes).
Tags: attitudes, behavior, cultural, identity, modeling, representation
Goals
- CAB seeks to develop a computational approach for representing, encoding, and using cultural knowledge at the individual and aggregate level.
- Integrate these cultural representations into AI behavior generation algorithms and existing virtual environment simulations.
- Explore how these cultural modules can be efficiently authored.