Copy the page URI to the clipboard
Huang, Xinyu
(2024).
DOI: https://doi.org/10.21954/ou.ro.000176d2
Abstract
Virtual humans and embodied conversational agents play diverse roles in real life, including game characters, chatbots, and teachers. In Augmented Reality (AR), such agents are capable of interacting with the real world. To distinguish between both types of virtual agents, AR agents were conceptually redefined as "holographic Artificial Intelligences (AIs)". Holographic AIs are embodied virtual agents interacting with real objects in Augmented Reality (AR), and can respond to events both in virtual and real environments. This thesis provides a comprehensive investigation into holographic AIs, spanning from their design to their user experience.
The purpose of this thesis is to investigate the creation and use of holographic AIs, by creating specific holographic AIs, and then examining how users perceive such entities in order to contribute to the improvement of the user experience. As a result, this thesis explores the design space for and methods for creating holographic AIs, proposing the novel PICS model which include the dimensions of persona, intelligence, conviviality, and senses.
Following the PICS model, a set of holographic AIs are designed by using a method of semi-automatic reconstruction. An AI that resembles a human being in appearance and behaviour is endowed with multimodal interactions capable of creating the illusion of physicality. The initial proposed model is then refined based on the experience of creation.
Basic body language gestures, such as nodding and opening the arms, are insufficient to engage users, particularly when it comes to intelligent tutoring systems. Therefore, this thesis specifically focuses on an open problem, the generation of re-usable standard instructional gestures. In an experiment, key instructional movements that can be employed by holographic AIs were identified and extracted as animations. The hitherto known range of representational gestures is, epistemologically, further expanded by transformational and imitation gestures, which show how humans manipulate spatio-motor information and characterise posture using hand motion. Therefore, the model can be extended to describe the holographic AI’s behaviour.
Moreover, in order to assess the empirical validity of holographic AIs, this research explores learners' trustworthiness towards this novel technology - as a key criterion for efficacy of this AI approach. Trust and trustworthiness, in terms of holographic AIs, refers to a mindset that aids users in achieving objectives based on good intentions. Young learners’ perception of trust is largely influenced by affective aspects of trust, determined by how emotionally responsive a holographic AI is.
These findings contribute to the design of personal holographic AIs that can perform a series of meaningful gestures that engage the learner’s attention for learning, which in turn fosters a reliable and trustworthy relationship. Both experiments are able to extend elements by adding gestures and holistic perception to this model.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)
