• Journal of The Korean Association of Information Education
• /
• v.5 no.1
• /
• pp.69-78
• /
• 2001

Recently, the current trend for computer based learning is moving from CAI environment to WBI environment. Most web documents for WBI learning are collected by aid of search engine. Instructors use those documents as learning materials after they evaluate availability of retrieved web documents. But, this method has the following problems. First, we search repeatedly the web documents selected by instructor. Second, there is a need for another course of instruction design in order to suggest the web documents for learner. Third, it is very difficult to analyze for relevance between the web documents and test results. In this work, we suggest WAILS(Web Agent Instruction Learning System) that retrieves web documents for WBI learning and guides learning course for learners. WAILS collects web documents for WBI learning by aid of web agent. Then, instructors can evaluate them and suggest to learners by using instruction-learning generating machine. Instructors retrieve web documents and the instruction-learning design at the same time. This can facilitate WBI learning.

• Educational Technology International
• /
• v.15 no.1
• /
• pp.1-26
• /
• 2014

The purpose of this study was to develop design principles of both animated pedagogical agents as 'credible' persuasive message source and persuasive fear arousing instructional messages in order to help enhance attitude changes toward a certain issue. Based on the previous pedagogical agent research, this study drew the design principles providing ways to manipulate agent credibility level and fear arousing level of message. Consequently, it specified how to make pedagogical agents perceived less or more credible by learners by manipulating a variety of agent features. For fear arousing message, this study showed how fear arousing messages would be structured into one of three levels: non-threatening, moderately threatening, and strongly threatening. Two different agent conditions and three message conditions were actually developed and experimentally tested with the participants of 40 undergraduate students. The results showed that the agent design principles specified from the previous research worked well enough to make a distinction between the more credible agent and the less credible agent. The overall results of this study may indicate that the design strategies for fear arousing message are retained on the premise of some future refinements.

• 한국HCI학회:학술대회논문집
• /
• 2006.02a
• /
• pp.918-925
• /
• 2006

Recent development of teachable agent provides learners with active roles as knowledge constructors and focuses on the individualization. The aim of this adaptive agent is not only to maximize the learner's cognitive functions but also to enhance the interests and motivation to learn. In order to establish the relationships among user characteristics and response patterns and to extract the algorithm among variables, we measured the individual characteristics and analyzed logs of the teachable agent named KORI (KORea university Intelligent agent) through the student modeling. A correlation analysis was conducted to identify the relationships among individual characteristics, user responses, and learning outcomes. Among hundreds of possible relationships between numerous variables in three dimensions, nine key user responses were extracted, which were highly correlated with either individual characteristics and learning outcomes. The results suggest that certain type of learner responses or the combination of the responses would be useful indices to predict the learners' individual characteristics and ongoing learning outcome. This study proposed a new type of dynamic assessment for individual differences and ongoing cognitive/motivational learning outcomes through the computation of responses without measuring them directly. The construction of individualized student model based on the ongoing response pattern of the user that are highly correlated with the individual differences and learning outcome may be the useful methodology to understand the learner's dynamic change during learning.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.2
• /
• pp.73-82
• /
• 2023

A 2(pedagogical agent: with vs. without) × 2(on-screen text: with vs. without) between-subject design was used in this study to investigate the effects of pedagogical agent and redundant on-screen text on video learning. In the case of the educational video without redundant on-screen text, there was no difference in social presence, satisfaction, and intention to continue learning regardless of the presence of a pedagogical agent. However, when the educational video contained redundant on-screen text, participants who watched educational video with pedagogical agent perceived higher social presence, satisfaction and intention to continue. In terms of academic achievement, no difference was found whether redundant on-screen text was contained or not. It supports some of the previous studies on the reverse-redundancy effects, suggesting that the inclusion of redundant text does not necessarily cause the reduction of learning outcomes. Video learning shows a higher dropout rate than face-to-face learning. Therefore, it is particularly important to understand how to strengthen interactions with learners and motivate them to keep themselves engaged in learning. This study discussed whether pedagogical agent and on-screen text are factors that induce continuous participation of learners in video learning.

• Journal of Korean Association for Spatial Structures
• /
• v.23 no.2
• /
• pp.29-36
• /
• 2023

Reinforcement learning (RL) is widely applied to various engineering fields. Especially, RL has shown successful performance for control problems, such as vehicles, robotics, and active structural control system. However, little research on application of RL to optimal structural design has conducted to date. In this study, the possibility of application of RL to structural design of reinforced concrete (RC) beam was investigated. The example of RC beam structural design problem introduced in previous study was used for comparative study. Deep q-network (DQN) is a famous RL algorithm presenting good performance in the discrete action space and thus it was used in this study. The action of DQN agent is required to represent design variables of RC beam. However, the number of design variables of RC beam is too many to represent by the action of conventional DQN. To solve this problem, multi-agent DQN was used in this study. For more effective reinforcement learning process, DDQN (Double Q-Learning) that is an advanced version of a conventional DQN was employed. The multi-agent of DDQN was trained for optimal structural design of RC beam to satisfy American Concrete Institute (318) without any hand-labeled dataset. Five agents of DDQN provides actions for beam with, beam depth, main rebar size, number of main rebar, and shear stirrup size, respectively. Five agents of DDQN were trained for 10,000 episodes and the performance of the multi-agent of DDQN was evaluated with 100 test design cases. This study shows that the multi-agent DDQN algorithm can provide successfully structural design results of RC beam.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.1
• /
• pp.27-38
• /
• 2023

In this paper, we propose a conversational AI agent based on continual learning that can continuously learn and grow with new data over time. A continual learning-based conversational AI agent consists of three main components: Task manager, User attribute extraction, and Auto-growing knowledge graph. When a task manager finds new data during a conversation with a user, it creates a new task with previously learned knowledge. The user attribute extraction model extracts the user's characteristics from the new task, and the auto-growing knowledge graph continuously learns the new external knowledge. Unlike the existing conversational AI agents that learned based on a limited dataset, our proposed method enables conversations based on continuous user attribute learning and knowledge learning. A conversational AI agent with continual learning technology can respond personally as conversations with users accumulate. And it can respond to new knowledge continuously. This paper validate the possibility of our proposed method through experiments on performance changes in dialogue generation models over time.

• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.672-680
• /
• 2003

A shopbot is a software agent whose goal is to maximize buyer´s satisfaction through automatically gathering the price and quality information of goods as well as the services from on-line sellers. In the response to shopbots´ activities, sellers on the Internet need the agents called pricebots that can help them maximize their own profits. In this paper we adopts Q-learning, one of the model-free reinforcement learning methods as a price-setting algorithm of pricebots. A Q-learned agent increases profitability and eliminates the cyclic price wars when compared with the agents using the myoptimal (myopically optimal) pricing strategy Q-teaming needs to select a sequence of state-action fairs for the convergence of Q-teaming. When the uniform random method in selecting state-action pairs is used, the number of accesses to the Q-tables to obtain the optimal Q-values is quite large. Therefore, it is not appropriate for universal on-line learning in a real world environment. This phenomenon occurs because the uniform random selection reflects the uncertainty of exploitation for the optimal policy. In this paper, we propose a Mixed Nonstationary Policy (MNP), which consists of both the auxiliary Markov process and the original Markov process. MNP tries to keep balance of exploration and exploitation in reinforcement learning. Our experiment results show that the Q-learning agent using MNP converges to the optimal Q-values about 2.6 time faster than the uniform random selection on the average.

• The Journal of Society for e-Business Studies
• /
• v.6 no.3
• /
• pp.63-80
• /
• 2001

The construction of the students knowledge in e-Learning systems, namely the student modeling, is a core component used to develop e-Learning systems. However, existing e-Learning systems have many problems to share the knowledge in a heterogeneous student model and a distributed knowledge base. Because the methods of the knowledge representation are different in each e-Learning systems, the accumulated knowledge cannot be used or shared without a great deal of difficulty. In order to share this knowledge, existing systems must reconstruct the knowledge bases. Consequently, we propose a new a Case Markup Language based on XML in order to overcome these problems. A distributed e-Learning systems fan have the advantage of easily sharing and managing the heterogeneous knowledge base proposed by CaseML. Moreover students can generate and share a case knowledge to use the communication protocol of agents. In this paper, we have designed and developed a CaseML by using a knowledge markup language. Furthermore, in order to construct an intelligent e-Learning systems, we have done our research based on the design and development of the intelligent agent system by using CaseML.

• International Journal of Contents
• /
• v.5 no.1
• /
• pp.51-55
• /
• 2009

In this paper, we present a structure of personalized e-Learning system to study for a test formalized by uniform multiple-choice using on/off line mixed estimations as is the case of Driver :s License Test in Korea. Using the system a candidate can study toward the license through the Internet (and/or mobile instruments) within the personalized concept based on IRT(item response theory). The system accurately estimates user's ability parameter and dynamically offers optimal evaluation problems and learning contents according to the estimated ability so that the user can take possession of the license in shorter time. In order to establish the personalized e-Learning concepts, we build up 3 databases and 2 agents in this system. Content DB maintains learning contents for studying toward the license as the shape of objects separated by concept-unit. Item-bank DB manages items with their parameters such as difficulties, discriminations, and guessing factors, which are firmly related to the learning contents in Content DB through the concept of object parameters. User profile DB maintains users' status information, item responses, and ability parameters. With these DB formations, Interface agent processes user ID, password, status information, and various queries generated by learners. In addition, it hooks up user's item response with Selection & Feedback agent. On the other hand, Selection & Feedback agent offers problems and content objects according to the corresponding user's ability parameter, and re-estimates the ability parameter to activate dynamic personalized learning situation and so forth.

• Journal of Information Technology Services
• /
• v.15 no.4
• /
• pp.85-95
• /
• 2016

Recently, deep learning and machine learning have attracted considerable attention and many supporting frameworks appeared. In artificial intelligence field, a large body of research is underway to apply the relevant knowledge for complex problem-solving, necessitating the application of various learning algorithms and training methods to artificial intelligence systems. In addition, there is a dearth of performance evaluation of decision making agents. The decision making agent that can find optimal solutions by using reinforcement learning methods designed through this research can collect raw pixel data observed from dynamic environments and make decisions by itself based on the data. The decision making agent uses convolutional neural networks to classify situations it confronts, and the data observed from the environment undergoes preprocessing before being used. This research represents how the convolutional neural networks and the decision making agent are configured, analyzes learning performance through a value-based algorithm and a policy-based algorithm : a Deep Q-Networks and a Policy Gradient, sets forth their differences and demonstrates how the convolutional neural networks affect entire learning performance when using pixel data. This research is expected to contribute to the improvement of artificial intelligence systems which can efficiently find optimal solutions by using features extracted from raw pixel data.