Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.704801
Title: Simulating ethical behaviour in virtual characters
Author: Headleand, Christopher J.
ISNI:       0000 0004 6057 150X
Awarding Body: Prifysgol Bangor University
Current Institution: Bangor University
Date of Award: 2016
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
The goal of virtual human simulation is to produce behaviour which is visually indistinguishable from reality. However, while various aspects of human behaviour have been extensively explored, there has been little research into behaviour motivated by moral objectives. Virtual characters are often simulated in charged environments, where rational behaviour is greatly challenged and in reality moral judgement plays a significant role. This thesis explores and presents novel solutions to the problem of simulating ethical behaviour. The research is presented in three stages. In the first, a reactive approach to the simulation of ethics inspired by Braitenberg’s Vehicles is described. This is achieved by iteratively augmenting a Type 3 Vehicle with new sensorimotor connections to produce further emergent results. The approach was capable of producing behaviour which was consistent with various normative specifications. Although successful, the Braitenberg Vehicle approach yields behaviour which is visually robotic. This is explored in the second stage of the research where a novel method for modelling affective behaviour is presented. In the third stage, a new architecture for the simulation and modelling of ethical behaviour called Trilogy is presented. This approach, inspired by classical and contemporary tripatriate theories of thought, serves as a computational substrate to bring together the ethical and affective simulation methods previously developed in stages one and two. Two experiments are conducted to evaluate the architecture where participants observed videos of simulated behaviour. The first experiment tests the hypothesis that the inclusion of affective states make an agent’s ethical behaviour more believable, and this was demonstrated to be the case. The second experiment compares the behaviour of the ethical agents against agents which are not ethically motivated. The results of both experiments demonstrate that the approach is capable of producing visually ethical behaviour beyond chance accuracy.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.704801  DOI: Not available
Share: