On the Different Application Domains of Virtual Agents


Term Paper, 2020

6 Pages, Grade: 1,0


Excerpt

Introduction

Virtual agents are an emerging research topic. On the one side, technology gets more advanced which opens up new possibilities to use virtual agents. On the other side, public interest in the topic is growing which makes the research area more attractive.

In the proceedings of the 16th International Conference on Intelligent Virtual Agents virtual agents are defined the following way: Virtual agents “are intelligent dig­ital interactive characters that can communicate with humans and other agents using natural human modal­ities such as facial expressions, speech, gestures, and movement. They are capable of real-time perception, cognition, emotion, and action that allow them to par­ticipate in dynamic social environments. “ (Traum et al. 2016)

In the following, different usage scenarios of virtual agents are presented. Firstly, the integration of vir­tual agents in pedagogical environments is examined in more detail and secondly a range of other usages of vir­tual agents in daily life is presented.

Virtual Agents in Pedagogic Environments

The use of virtual agents as tutors and teachers has been subject to numerous studies in which different re­searchers proposed designs and use cases for tutoring and learning systems but also evaluated the effects of the usage of such technologies.

Virtual Agents as Tutors

Taoum et al. (2016) proposed a design for a so-called intelligent tutoring system using virtual agents. They extended an existing framework called MASCARET (Multi-Agent System for Collaborative, Adaptive & Realistic Environments for Training) for embodied con­versational agents. The goal was to extend the agent’s behaviour to give it the ability to handle the learner’s feedback based on knowledge about the learning envi­ronment, signals from the learner and the user’s current learning state.

One thing the researchers focused on was the reaction on so-called backchannel signals. These are non-verbal feedback signals humans emit when communicating (e.g. shrugging or face expressions). The virtual agent should be able to react on a set of backchannel signals and emit such signals to make the communication more satisfying for the learner. Furthermore, they presented the structure of their intelligent tutoring system which consists of the following components:

- Domain Model: The domain model is the repre­sentation of the knowledge that should be trans­mitted including all objects the learning process involves and all interactions.
- Pedagogic Model: The pedagogic model defines all pedagogic objectives, activities, prerequisites and environments.
- Learner Model: The learner model contains all information regarding the learner. This includes the learner’s curriculum the history of actions that were already realized and the learner’s profile. The used learning scheme is based on learning by repeating and consists of three learning phases: the cognitive state in which the learner under­stands what to do, the associative stage in which the learner organizes the new information and fi- nally the autonomous stage in which the declara­tive knowledge evolves to procedural knowledge.

Abbildung in dieser Leseprobe nicht enthalten

Figure 1. The virtual agent in the pilot application (Taoum et al. (2016) ).

Abbildung in dieser Leseprobe nicht enthalten

Figure 2. The virtual agent in interaction with a child using the MITp system (Abdullah et al. (2017)).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3. A presentation assisted by the virtual copresenter (Bickmore et al. (2016)).

The virtual tutor is then able to detect in which learning phase the learner currently is and give feedback regarding the performed actions.

There is a pilot application using the system described above in which the user can learn about blood analysis (Figure 1), but this application has not been evaluated yet.

Abdullah et al. (2017) developed a virtual agent that supports discovery-based learning and investigated what it needs to make the virtual agent effective in the role of a tutor. The virtual agent leads learners through a set of activities with the goal that the learner gets new insights regarding a specific topic.

The researchers extended a maths learning envi­ronment called Mathematical Imagery Trainer for Proportionality (MITp) which provides knowledge about proportions to children (Figure 2). The learner needs to touch a screen with both hands in a pre-defined ratio and receives colour-based feedback. Based on the behaviour of human tutors the virtual agent was equipped with the following activity types: giving instructions, valorizing success, wait for the learner to explore a task, provide help if the learner is stuck and advancing the learner through the learning task. A dynamic decision network based on Bayesian networks decides which of the pre-defined actions the virtual agent should perform.

The system was evaluated in an experiment with 10 children aged between 9 and 12 years. All participants developed movement schemes with the guidance of the virtual agent that fulfilled the pre-defined ratios to touch the screen. However, the participants did not get all insights that were expected as the virtual agent did not instruct them in a way that was supportive to gain those insights. The researchers suggest that further investigations are required to solve this issue.

Virtual Agents as Teaching Assistants

Virtual agents are not only used in pedagogic environ­ments when it comes to direct communication between a learner and a tutor but also in typical classroom or lecture situations in which a lecturer is talking in front of people.

Bickmore et al. (2016) designed a virtual agent that is integrated in PowerPoint to assist in oral presentations (Figure 3). The virtual agent is intended to assist students and researchers to give more engaging presentations. The virtual co-presenter is an embodied virtual agent that is able to use verbal and non-verbal signals to deliver content, emphasize important points and actively listen to its human co-presenter. Is is possible for the user of the PowerPoint extension to define and control the virtual agent’s verbal and non­verbal behaviour.

The researchers conducted several studies to eval­uate the virtual co-presenter. They conducted a laboratory-based study with 12 participants in which the participants had to deliver two prepared 7-minute presentations. One presentation was assisted by the virtual agent and one presentation was held by the participant alone. The study showed that the assistance of the virtual agent significantly decreased public speaking anxiety and increased speaker confidence for the non-native English speaking participants. In a follow-up study, 12 judges rated the presentations. The result was that the assisted presentations were rated better on note reliance, quality of speech and presentation quality.

After the laboratory-based study, the co-presenter was also evaluated in real-life conditions. It was included in a computer science research methods course at university and the students were asked to rate the lecture. The study showed that the students found the lectures given by the professor and the virtual agent significantly more novel, exciting and entertaining. It has to be noted that the professor was rated as more competent without the agent. However, the students reported that they would like to see more assisted lectures in the future.

A second lecture in which the virtual agent was included was a public speaking course and the par­ticipating students used the virtual agent in one of their presentations. The students reported that they prepared their presentations more thoroughly using the virtual co-presenter. As before, the participants again reported that the virtual agents increases engagement, variety and energy level of a presentation and decreases anxiety.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4. The animal that is instructed in the learning game (Lindberg et al. (2017)).

Abbildung in dieser Leseprobe nicht enthalten

Figure 5. The robot used for the study (Lindberg et al. (2017)).

Lindberg et al. (2017) also investigated teaching situations. In contrast to the previous approaches, they did not only look at virtual agents but also at robots. They compared the effects that a virtual agent and a robot has on learners when they follow a so-called learning-by-teaching approach. Learning-by-teaching means that the learner gains knowledge about a certain topic by teaching the topic to another person (or virtual agent/robot).

For their investigation, they used the learning game Magical Garden in which children can learn basic maths concepts by teaching an animal (Figure 4). For the robot version, they used a humanoid robot (Figure 5). The participants of the study were eight children aged between 5 and 9 years. Each participant interacted with the robot as well as with the virtual agent. The study showed that the robot attracted more attention but lost the attention in the interaction. Both the virtual agent and the robot were able to convey mental states to the children.

Abbildung in dieser Leseprobe nicht enthalten

Figure 6. Interaction of a psychotherapist with the virtual patient (Johansson et al. (2017)).

Virtual Agents for Educational Simulation

Virtual agents can also support educational simulations and simulate human behaviour for learners to prac­tice certain types of interaction with the virtual agent. Johansson et al. (2017) implemented a psychotherapy training system in which the Furhat social robot sim­ulates virtual patients (Figure 6). The virtual patient’s behaviour is based on the so-called Intensive short-term dynamic psychotherapy which provides knowledge on in-session verbal and non-verbal behaviour.

The training system focuses on the assessment phase at the beginning of a treatment. The user should learn how to deal with barriers that can occur during the in­teraction with the patient. These barriers like anxiety or defences likely not only occur during the assessment phase but also during the actual therapy.

The user learns to detect verbal and non-verbal signs of anxiety and defences and how to deal with these be­haviours. There are multiple virtual patients available for the user to work with different characters and barri­ers.

Virtual Agents in Other Areas of Life

Virtual Agents are not only emerging in pedagogic environments but in multiple other areas of daily life. They can occur at work, in health care and in recreational activities like sports.

Kraemer et al. (2017) argued that numerous people nowadays work in remote working environments and therefore focused on how virtual bosses are perceived. They investigated how people experience negative feedback given from a human boss and from a virtual boss (Figure 7) in a remote working environment. For the study, they split 183 participants aged between 20 and 69 years into four groups: human boss and factual feedback, human boss and emotional feedback, virtual boss and factual feedback, virtual boss and emotional feedback. The participants had to do a proofreading task and received negative feedback, no matter how they performed. The study showed that it did not make a difference if the boss was human or virtual regarding how the feedback was perceived. People of the emotional feedback group rated the boss less warm and more human like regardless the boss condition. Also there was no difference for the boss conditions regarding the perceived psychological safety. Participants of the factual feedback group reported a higher psychological safety than participants from the emotional feedback group. Finally the boss condition did not have an effect on the perceived social presence of the boss or on the performance of the participants.

Abbildung in dieser Leseprobe nicht enthalten

Figure 7. Pictures of the human and computer version of the bosses used for the study (Kraemer et al. (2017)).

[...]

Excerpt out of 6 pages

Details

Title
On the Different Application Domains of Virtual Agents
College
University of Constance
Grade
1,0
Author
Year
2020
Pages
6
Catalog Number
V974344
ISBN (eBook)
9783346322142
Language
English
Keywords
different, application, domains, virtual, agents
Quote paper
Ramona Burger (Author), 2020, On the Different Application Domains of Virtual Agents, Munich, GRIN Verlag, https://www.grin.com/document/974344

Comments

  • No comments yet.
Read the ebook
Title: On the Different Application Domains of Virtual Agents



Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free