TAP: An Effective Personality Representation for Inter-Agent Adaptation in Games

AAAI Conferences

Tactical Agent Personality (TAP) is a modeling concept to capture tactical patterns in game agents, based on a personality concept introduced by Tan and Cheng (2007), to allow behavior adaptation to different play styles. We introduced a weighted sequential topology to the actions set to capture tactical preferences rather than individual action preferences. This produces a personality representation of much higher expressive power that improves the adaptation performance and subsequently enables a larger variety of action genres to be adaptable. A TAPbased learning framework is then constructed and it is shown to perform better than the one based on the previous agent personality. Consequently, we also implement an RPG scenario that demonstrates its ability to generate adaptive plausible behavior in a much larger variety of action genres.


Personality-based Adaptation for Teamwork in Game Agents

AAAI Conferences

This paper presents a novel learning framework to provide computer game agents the ability to adapt to the player as well as other game agents. Our technique generally involves a personality adaptation module encapsulated in a reinforcement learning framework. Unlike previous work in which adaptation normally involves a decision process on every single action the agent takes, we introduce a two-level process whereby adaptation only takes place on an abstracted actions set which we coin as agent personality. With the personality defined, each agent will then take actions according to the restrictions imposed in its personality. In doing so, adaptation takes place in appropriately defined intervals in the game, without disrupting or slowing down the game constantly with intensive decision-making computations, hence improving enjoyment for the player. Moreover, by decoupling adaptation from action selection, we have a modular adaptive system that can be used with existing action planning methods. With an actual typical game scenario that we have created, it is shown that a team of agents using our framework to adapt towards the player are able to perform better than a team with scripted behavior. Consequently, we also show the team performs even better when adapted towards each other.


Computer-based personality judgments are more accurate than those made by humans

#artificialintelligence

Judging others' personalities is an essential skill in successful social living, as personality is a key driver behind people's interactions, behaviors, and emotions. Although accurate personality judgments stem from social-cognitive skills, developments in machine learning show that computer models can also make valid judgments. This study compares the accuracy of human and computer-based personality judgments, using a sample of 86,220 volunteers who completed a 100-item personality questionnaire. We show that (i) computer predictions based on a generic digital footprint (Facebook Likes) are more accurate (r 0.56) than those made by the participants' Facebook friends using a personality questionnaire (r 0.49); (ii) computer models show higher interjudge agreement; and (iii) computer personality judgments have higher external validity when predicting life outcomes such as substance use, political attitudes, and physical health; for some outcomes, they even outperform the self-rated personality scores. Computers outpacing humans in personality judgment presents significant opportunities and challenges in the areas of psychological assessment, marketing, and privacy.


Computer-based personality judgments are more accurate than those made by humans

#artificialintelligence

Judging others' personalities is an essential skill in successful social living, as personality is a key driver behind people's interactions, behaviors, and emotions. Although accurate personality judgments stem from social-cognitive skills, developments in machine learning show that computer models can also make valid judgments. This study compares the accuracy of human and computer-based personality judgments, using a sample of 86,220 volunteers who completed a 100-item personality questionnaire. We show that (i) computer predictions based on a generic digital footprint (Facebook Likes) are more accurate (r 0.56) than those made by the participants' Facebook friends using a personality questionnaire (r 0.49); (ii) computer models show higher interjudge agreement; and (iii) computer personality judgments have higher external validity when predicting life outcomes such as substance use, political attitudes, and physical health; for some outcomes, they even outperform the self-rated personality scores. Computers outpacing humans in personality judgment presents significant opportunities and challenges in the areas of psychological assessment, marketing, and privacy.


An Environment for Transforming Game Character Animations Based on Nationality and Profession Personality Stereotypes

AAAI Conferences

A vast body of literature has dealt with the challenges of creating the impression of human appearance and human-like motion in the animation of game characters. In this paper, we further refine these efforts by creating a flexible environment for animating game characters endowed with personality, which is a core descriptor of stable characteristics of human behavior and which is often expressed in human movement. We base our work on the Big Five personality traits, also known as OCEAN (Openness, Conscientiousness, Extroversion, Agreeableness, Neuroticism). Our environment incorporates a procedural mapping from OCEAN personality traits to movement modifiers that alter existing motions in ways compatible with a desired personality. Using Amazon Mechanical Turk, we collected stereotypical personality profiles for 135 nationalities and 100 professions. We integrated these stereotypical personality expectations into an interactive interface in Unity3D. Users can linearly blend the nationality and profession OCEAN parameters and individually adjust them for specific characters or groups. The results are validated using Amazon Mechanical Turk pairwise judgments on character types based on movements.