Earticle

Most everyday meetings today involve presenters connecting their computers to a shared TV or projector via HDMI during their turn. This requires the hassle of switching HDMI connections every time a presenter changes, and it also limits the ability to view multiple presenters' screens simultaneously. This paper proposes and implements a Multiplexer System that allows presenters to easily share displays. This system is a form of display sharing that connects the presenters' screens to a general-purpose multiplexer computer that is directly connected to a large display and is connected in real time. We used WebRTC to handle real-time transmission of screen images, and leveraged the hardware video decoder built into the multiplexer computer for fast decoding of encoded streams. To evaluate the performance of the Multiplexer System, we measured the screen latency required for the screen image of the presenter's computer to be displayed on the multiplexer's display. The result was evaluated to be around 221-362 ms when the number of simultaneous presenters is two, which is considered to be a level that does not cause any inconvenience to the presenters. We conclude that the Multiplexer System proposed in this paper is a system that conveniently supports everyday meetings without the use of special hardware or additional costs.

This paper proposes three one-time password (OTP) generation methods suitable for firmware OTPs, one of the most representative methods for addressing the rapidly growing demand for user authentication in online services, based on stream and block ciphers. The first method is a stream cipher-based approach consisting of a 127-bit linear feedback shift register (LFSR) and exclusive-OR (XOR) operators. The OTP output bits are determined using a bit-position selection derived from the digits of  . The second method is a block cipherbased approach employing triple data encryption standard (TDES). Part of the output bits are used as the OTP output and the remaining bits are fed back to the input through an output feedback (OFB) mode. The third method adopts advanced encryption standard (AES) as the block cipher, using a portion of the output bits as the OTP output and feeding back a subset of the remaining bits to the input. All methods generate initial values through key-based random number generation applying message authentication code (MAC). The proposed methods are implemented on an Arduino platform as firmware-based OTP generators. Experimental results demonstrate that the proposed methods offer strong security properties and are suitable for firmwarebased OTP generation. In addition, the LFSR-based method shows good performance in the NIST SP 800-22 randomness tests.

With increasing security concerns over unencrypted Mode-4 communication, the U.S. and its NATO allies are transitioning to encrypted Mode-5 to improve interoperability and security. To maintain compatibility, systems using the legacy Mark XII Mode-4 standard are being upgraded to the Mark XIIA standard for Mode-5. The primary objective is to obtain AIMS certification, which is required for compliance with the Mark XIIA Mode- 5 standard by successfully passing an assessment from AIMS PO, a U.S. DoD subsidiary. The AIMS certification process is complex, involving many individual tests across multiple stages, necessitating efficient management and organization to analyze results effectively. This paper outlines the certification procedure and proposes methods to improve the efficiency of AIMS evaluations for ground-based radar systems. For the integration test phase (AIMS-1202), similar test items are grouped based on configuration to reduce redundancy and streamline the evaluation process. In the flight test phase (AMIS-1203), test items are classified into groups according to test methods and techniques. The paper also presents strategies for planning tests under varying flight time and path constraints. This approach enhances efficiency, minimizes evaluation risks, and increases the likelihood of successful certification. The insights and methods developed will also aid in improving future AIMS-related certification processes for weapon systems.

This paper presents a novel methodology to significantly enhance the reliability of high-speed power line communication (HS-PLC) systems via deep learning (DL). We impact to propose and validate a DL-based pre-compensation scheme specially designed for the effective mitigation of impulsive noise. It utilizes a pre-trained DL model deployed at the transmitter to accurately predict the instantaneous statistical characteristics of the impulsive noise. This predictive information enables real-time pre-compensation of the transmitted signal, resulting in a substantial improvement in the received signal quality. To ensure optimal prediction accuracy, a comprehensive noise database was meticulously constructed based on the empirical characteristics of measured noise patterns. For channel modeling, the Middleton Class A interference model was adopted to accurately simulate the representative impulsive noise conditions. The performance was rigorously evaluated through bit error rate (BER) analysis. Simulation results demonstrate that the proposed DL-based technique achieves a marked reduction in BER and a significant enhancement in signal quality relative to conventional systems. The developed system model holds promising potential as a universal solution for signal integrity improvement, extending its applicability beyond HS-PLC to a wide spectrum of wired and wireless communication systems susceptible to impulsive interference.

This paper presents a Hybrid Alignment Model (HAM) based on a multi-source domain adaptation (MSDA) framework that designates directly captured, unprocessed images (Set C) as a fixed reference domain. Domain bias remains a major obstacle to generalization, particularly for models trained on visually enhanced web or academic images (Sets A and B), where Top-1 accuracy drops by up to 27.2 percentage points (pp) when transferred to real-world data (A→A = 0.988 vs. A→C = 0.716). We aim to overcome this domain bias, which severely degrades out-of-domain generalization in computer-vision–based gemstone identification. HAM builds on the 2,048-dimensional embedding of a pre-trained ResNet-50 backbone and integrates three complementary alignment strategies—maximum mean discrepancy (MMD, global), adversarial alignment (Adv., local), and contrastive alignment (Conf., class-boundary)—to mitigate domain shift at multiple levels. Decision-boundary stability is further enhanced through three auxiliary strategies—pseudo-labeling (PL), adaptive template matching (ATM), and dynamic loss weighting (DLW). Empirically, HAM achieves substantial gains over the baseline: Top-1 accuracy improves from 0.716 to 0.935 (+21.9 pp) for A→C and from 0.775 to 0.956 (+18.1 pp) for B→C, effectively mitigating domain bias and enabling robust cross-domain generalization. Explainable AI (XAI) analyses using Grad-CAM, LIME, SHAP, and Integrated Gradients (IG) confirm that the model consistently focuses on expert gemstone cues—light dispersion, surface texture, and facet boundaries— supporting interpretability and trust. To the best of our knowledge, this study presents the first empirical integration of all three core alignment mechanisms (MMD, Adv., Conf.) within a unified MSDA framework, offering an explainable and practically reliable approach for real-world gemstone identification.

The purpose of this study is to statistically examine whether there are significant differences in monthly smartphone fees among four age groups: infants, adolescents, adults, and the elderly. A one-way analysis of variance (One-way ANOVA) was employed to test whether the mean monthly fees differ significantly across these age groups. ANOVA is a statistical technique used to compare the means of three or more independent groups. In this study, the independent variable was age group (four levels), and the dependent variable was the average monthly smartphone fee. The results showed that adults had the highest average monthly fee (M = 5.20), followed by the elderly (M = 4.65), adolescents (M = 4.28), and children (M = 3.62). The ANOVA results indicated statistically significant differences among the groups (F-test, p < .05), and post-hoc analyses revealed that all groups differed significantly from each other. These findings suggest that the differences in smartphone usage patterns—parent-controlled and educational use among children, content-centered use among adolescents, functional and work-related use among adults, and basic communication use among the elderly—directly contribute to variations in communication costs. In conclusion, this study empirically demonstrates that age-related differences in smartphone usage patterns lead to significant disparities in communication expenses. The findings provide a foundational basis for designing age-specific mobile pricing plans and for developing policy measures to reduce communication costs among vulnerable groups such as adolescents and the elderly. Future research could incorporate additional socioeconomic variables (e.g., gender, income, occupation) and apply two-way ANOVA to obtain a more comprehensive understanding of factors influencing smartphone cost differences.

This paper develops the Adaptive Multisensory Integration Framework (AMIF), explaining how artificial intelligence coordinates sensory modalities to enhance human-computer interaction. Synthesizing cognitive neuroscience, AI, and HCI theories, AMIF identifies five cyclically-operating components governing multisensory integration. Critical analysis reveals fundamental limitations in temporal prediction, computational efficiency, and individual variability. Four case studies demonstrate the framework's explanatory power, showing both successes and persistent challenges in real-world systems. The framework generates testable hypotheses and design principles grounded in theoretical understanding.

This study aims to empirically examine the educational effectiveness of a Generative Artificial Intelligence(AI)- based short-term intensive Project-Based Learning (PBL) program. Motivated by the rapid emergence of Generative AI in education, the authors intend to explore whether AI-integrated PBL can serve as a viable alternative to traditional lecture-driven learning. The program was conducted with 23 university students, who used ChatGPT, Runway, Midjourney, and SUNO to complete the entire immersive video production process. A post-program survey consisting of four domains—Learning Experience, AI Utilization Competence, PBL Learning Experience, and Program Satisfaction—was administered using a five-point Likert scale. Results revealed high satisfaction across all categories, with Learning Experience (C_M = 4.36) and PBL Learning Experience (C_M = 4.36) showing the strongest outcomes. Learners particularly valued the role of Generative AI in providing practical problem-solving experiences and enabling rapid prototype production. Although some variation was observed in self-directed learning and AI confidence, overall findings indicate strong gains in engagement, creativity, and practical competency. These results support the authors’ intention to validate AI-based PBL as an immersive, efficient, and practice-oriented learning model. Moreover, the study highlights its potential for broader application in lifelong learning, reskilling, and professional technical education. The authors ultimately aim to contribute empirical evidence to the emerging discourse on AI-mediated instructional design and future-oriented educational innovation.

Recently, Large Language Models (LLMs) have led to rapid performance improvements across natural language processing (NLP) and are expanding into various domains. However, pre-trained LLMs are based on general-purpose data, which can lead to domain bias and learning inefficiencies when applied to specific domains. Consequently, efficient fine-tuning techniques that can optimize specific domains while maintaining the model's generalizability are emerging as a key research topic. To overcome these limitations, this study proposed an Adaptive Layer-wise Fine-tuning (ALF) technique that performs selective parameter updates based on layer importance. ALF calculates the weight change rate of each layer during the learning process and dynamically updates only the layers with high importance, thereby improving efficiency without performance degradation while learning only about 20-30% of the total model parameters. The results of the attention weight analysis confirmed that ALF had improved focus on semantically central words, thereby improving contextual consistency and model interpretability (interpretable fine-tuning). These results suggest that ALF is an effective approach that simultaneously improves efficiency and generalization performance compared to existing fine-tuning techniques. Future research will develop a framework for optimizing generative AI models that is both sustainable and reliable, through expansion into multimodal data, reinforcement learning-based auto-tuning, and ethical fine-tuning.

This study aims to develop an AI-integrated educational model that incorporates generative AI into the UX design curriculum and to validate its effectiveness through Design-Based Research (DBR). The research followed the ADDIE instructional design framework—Analysis, Design, Development, Implementation, and Evaluation—using a university-level UX design course as the experimental setting. In Phase 1, learner needs and the existing curriculum were analyzed to design a project-based learning model centered on AI-supported design processes. In Phase 2, project-based classes were conducted using generative AI tools such as ChatGPT, Gemini, and Midjourney. Learning outcomes were evaluated using a structured rubric focusing on creativity, problem-solving ability, and collaboration. In Phase 3, the curriculum was refined through instructor interviews and learner feedback analysis. The findings indicate that integrating generative AI into design education significantly improved students’ problem-definition accuracy, visualization speed, and collaborative engagement. It also increased their awareness and literacy in AI utilization. This study proposes practical instructional design guidelines for incorporating AI into UX design education and provides foundational insights for developing qualitative evaluation models to support the digital transformation of design pedagogy. We aim to impact the future of design education by establishing a practical and sustainable framework that bridges generative AI technology with creative design learning.

This study investigates how generative AI prompt literacy influences creativity and design quality during the ideation stage of the UX design process. As generative AI becomes increasingly integrated into early-stage design research and idea exploration, the specificity and contextual clarity of prompts are expected to play a critical role in shaping design outcomes. A mixed-methods approach was employed, combining quantitative experimentation with qualitative interviews involving university students majoring in UX design. Participants were divided into three groups: (1) a specific prompt group, (2) an abstract prompt group, and (3) a non-AI control group. The resulting design outputs were assessed based on creativity, appropriateness, and visual completeness. Findings indicate that the group using specific prompts produced more diverse and contextually refined ideas and exhibited higher self-efficacy in utilizing AI tools. These results highlight the pivotal role of prompt literacy as a core competency in AI-integrated design education. The study offers empirical insights that can inform future models of AI-based design pedagogy and evaluation.

This study originates from the empirical confirmation of an ‘Independent Path’ phenomenon through data analysis of 556 Korean small and medium-sized manufacturing enterprises (SMEs), where smart factory adoption positively impacted safety but showed no statistically significant effect on productivity—contrary to general expectations of simultaneous improvement in both areas. Under this premise, this research develops an ‘Integrated Productivity-Safety Maturity Model (IPSM)’ and corresponding ‘Integrated Productivity-Safety Index (IPSI)’ through a systematic four-stage integration methodology, designed by deconstructing, linking, and integrating core elements of existing Smart Factory Level Diagnosis and Safety Diagnosis models operated by Korea SMEs and Startups Agency (KOSME). The IPSI adopts an innovative multiplicative structure where safety functions as a ‘sustainability multiplier’ for productivity, quantifying the impact of potential safety risks (safety debt) on productivity sustainability within a single integrated indicator, thereby moving beyond simple additive approaches. This research contributes to comprehensive smart factory maturity diagnosis and sustainable growth strategies for SMEs, offering significant academic and practical implications by presenting a novel integrated methodology that addresses existing diagnostic system limitations.

The rapid expansion of electric vehicles and energy storage systems has led to an unprecedented accumulation of end-of-life lithium-ion batteries (LIBs), creating both environmental challenges and safety hazards. This study proposes an AI-based risk estimation framework that integrates real-time sensor data and machine learning analytics to enhance the safety and efficiency of LIB recycling processes. The framework comprises four key layers—data acquisition, preprocessing, AI risk estimation, and decision support—and applies a hybrid Deep Neural Network (DNN) and Gradient Boosted Tree (GBT) model to predict potential hazards across mechanical dismantling, thermal pretreatment, and hydrometallurgical recovery stages. A pilot-scale experimental setup was constructed, collecting over 10,000 time-series samples of temperature, gas emission rate, pH, pressure, and current under varying recycling conditions. The proposed model achieved 98% accuracy, 0.97 AUC, and a 45% reduction in mean time-to-alarm (MTTA) compared with rule-based monitoring systems, providing proactive early warnings 2–3 minutes before hazardous events. Explainability analysis using SHAP values identified temperature deviation (ΔT) and gas emission rate (G) as the dominant contributors to risk prediction, jointly explaining over 60% of model variance. The system maintained stable performance under noisy and incomplete data, confirming its applicability for real-time industrial deployment. Overall, this work demonstrates that AIdriven predictive intelligence can significantly improve safety, sustainability, and operational reliability in LIB recycling, supporting the transition toward smart, circular, and Industry 5.0–aligned recycling systems.

Traditional recommendation systems face limitations associated with data scarcity and structural complexity. In contrast, graph-based methods can effectively leverage extensive relational information. In this study, we examine the core principles of dynamic clustering and LightGCN to enhance the recommendation performance of artificial intelligence (AI) agents. Based on our findings, we propose an embedding propagation and aggregation strategy incorporating dynamic clustering to extend and refine the LightGCN architecture. Experimental results demonstrate significant performance improvements compared with the conventional LightGCN across key evaluation metrics, confirming the effectiveness of the proposed method for next-generation, AI-agent-based recommendation services.

The rapid evolution of Large Language Models (LLMs) has accelerated the creation of advanced AI agent frameworks capable of automating complex tasks across multiple domains. However, the diversity of available frameworks presents significant challenges for developers in selecting appropriate platforms for their specific needs. We designed this study to provide a systematic comparative analysis of five major AI agent frameworks— LangChain, AutoGen, CrewAI, OpenDevin, and SuperAGI—to guide framework selection and identify key characteristics distinguishing each platform. We evaluate these frameworks based on multiple criteria, including architectural design, integration capabilities, developer experience, and scalability. Our methodology combines analysis of official documentation, hands-on experimentation, and assessment of community feedback to provide comprehensive insights. The analysis identifies significant trade-offs between flexibility and simplicity, with each framework demonstrating distinct strengths in particular application contexts. LangChain offers maximum flexibility for custom implementations, AutoGen simplifies multi-agent coordination, CrewAI provides intuitive team-based orchestration, OpenDevin specializes in software development automation, and SuperAGI delivers comprehensive platform capabilities. We present practical guidance for developers and researchers seeking suitable frameworks for their projects and highlight emerging trends toward standardization in the AI agent ecosystem, contributing to more informed decision-making in framework adoption.

The purpose of this study is to establish a systematic auditing methodology for ensuring the reliability of AI systems. To achieve this, we reviewed the definition of AI systems and the necessity of trustworthy AI, and specified the scope and targets of auditing based on the full life cycle of the AI system. As a result of the research, we proposed an AI System Auditing Framework consisting of six phases: Requirements Definition, Data Collection and Preprocessing, Model Development and Training, Deployment and Operation, Maintenance and Continuous Improvement, and Decommissioning and Archiving. For each phase, specific verification items were derived, including verification of suitability for purpose, data quality and bias checks, monitoring of the training process and explainability assessment, performance and stability verification, inoperation security and user trust evaluation, and personal information destruction and transition management.

In semiconductor manufacturing, the Clean and Coat (C&C) process critically influences product yield, yet its internal operation is difficult to observe directly. To address this limitation, we propose a patch-based region segmentation and recommendation framework for analyzing camera sensor images captured during C&C operations. The proposed method divides monitoring images into uniform patch regions suitable for AI-based image analysis and automatically recommends regions of interest that are most sensitive to abnormal phenomena such as wafer wobbling. A client–server monitoring system was implemented, where the frontend provides live or recorded video playback and the backend performs patch analysis using a FastAPI-based module. Experimental validation using a mock-up setup demonstrated that the system can accurately detect wafer boundaries, group surface brightness levels, and rank candidate patches in real time. The proposed framework effectively reduces operator dependence, improves consistency in region-of-interest configuration, and enhances the reliability of real-time C&C process monitoring. This study provides a foundation for integrating intelligent vision algorithms into semiconductor manufacturing environments.

Recently, various deep learning-based approaches have been studied to interpret complex human actions. A significant limitation of conventional methods is their reliance on a fixed graph structure, which restricts their ability to capture discriminative receptive fields. In this paper, we design a deformable spatio-temporal attention-based graph convolutional neural network (deformable STA-GCN) that dynamically identifies informative joints and temporal patterns. The proposed framework integrates two key modules, a deformable spatial attention (DeSA) module and a deformable temporal attention (DeTA) module. The DeSA module is integrated to dynamically identify joints closely related to motion, and the DeTA module is incorporated to identify informative temporal receptive fields. Within each module, an attention mechanism is employed to selectively emphasize the sampled features, enabling the network to extract the most discriminative spatiotemporal information. Experimental results on the NTU RGB+D benchmark dataset demonstrate that the proposed model achieves higher accuracy than the comparison model.

This study proposes a fire disaster prevention system that analyzes CCTV video streams to detect multi-crowd movements in real time and evaluate potential risk levels. The system identifies early warning signs by assessing crowd density, movement speed, and directional changes, enabling rapid detection of hazardous situations in densely populated environments. The proposed method extracts motion vectors and behavioral features from sequential video frames to compute a risk index that reflects scene instability. Experimental results using synthetic and real CCTV data demonstrate that the model effectively distinguishes normal, alert, and evacuation states. These findings indicate that the system can provide timely early warnings and support prompt decision-making during fire-related emergencies.

In this study, I apply the Selected Query Detection Transformer, a powerful transformer-based framework for real-time bird detection. This research aims to improve the reliability of bird detection to address bird strikes, a serious threat to aviation safety. The proposed model was trained using the CUB-200-2011 dataset and subsequently fine-tuned on a real-world bird surveillance dataset collected near airport runways to enhance adaptability to real environments. SQ-DETR employs a layer-wise adaptive query pruning mechanism that dynamically removes low-importance object queries during decoding, thereby reducing redundant computations while preserving detection accuracy. Experimental results demonstrate that SQ-DETR outperforms YOLOv8-L, achieving 2.5% higher mean Average Precision and reducing computational cost by approximately 18%, with an AP₅₀ of 90.0 and AP₇₅ of 87.2. Qualitative analysis further shows that SQ-DETR more accurately detects small or partially occluded birds in complex airport scenes compared to YOLO. Overall, SQ-DETR effectively balances precision and efficiency, providing a practical and scalable framework for real-time bird surveillance and bird-strike prevention systems. This study highlights the potential of Transformer-based architectures to enhance safety and operational reliability in modern aviation monitoring environments.

We designed a multifunctional neural network to jointly perform semantic segmentation and object boundary extraction. Our goal was to enhance boundary precision, especially for thin and complex facial structures. The model is built on a pre-trained ResNet101 backbone and incorporates ASPP and transposed convolutions to handle objects across multiple scales. A parallel branch structure enables simultaneous learning of semantic regions and boundary details. To further improve visual clarity, we introduce an RoI Tanh-warping technique that selectively distorts the background while preserving the natural appearance of the target region. We also apply a progressive layer unfreezing strategy to allow smooth adaptation to new tasks while retaining key pre-trained features. Experimental results confirm that our method delivers superior boundary accuracy, achieving mean F1 scores of 93.40 and 93.42 on the validation and test datasets. These findings demonstrate that the proposed approach provides both quantitative gains and visually coherent segmentation results.

We designed a boosting-based tourist attraction recommendation system that integrates theme classification and satisfaction prediction into a single pipeline. Using AI Hub and KMA datasets, we preprocessed tourist destination information and vectorized destination names with Char2Vec. XGBoost was applied for theme classification, achieving high accuracy, while Gradient Boosting regression was used for satisfaction prediction with winsorizing to ensure stability. Experimental results show that the proposed model outperformed other baseline algorithms in both classification and regression tasks. The system visualizes regional theme distributions through Geo and Choropleth Maps, enabling users to explore personalized recommendations intuitively. These results demonstrate that our integrated pipeline can serve as a foundation for future AI-driven personalized tourism recommendation platforms.

This paper synthesizes the status quo—namely the non-recognition of copyright for purely AI-generated results in Korea—emerging information-disclosure obligations, and policy alternatives such as provenance labeling, watermarking, and text-and-data-mining (TDM) governance, and presents an actionable blueprint for regulators and creators. Building on doctrinal analysis of Korea’s Copyright Act and pending National Assembly amendment texts (2024–2025), policy guidance on registration of AI-assisted works, and a comparative scan of U.S., Canadian, Indian, and other international positions, the study clarifies how human authorship remains a baseline requirement while AI-generated material increasingly permeates cultural production. First, it reconstructs the normative debates for and against protection of AI-generated works, including arguments about data ownership, the public domain, and the equivalence and contribution of AI outputs to cultural development. Second, it examines global trends in AI art creation and authorship through key cases and technologies such as Naruto v. Slater, Google DeepDream, GAN/CAN-based artistic systems, and the U.S. Copyright Office’s refusal to register A Recent Entrance to Paradise. Third, it identifies four core issues for Korean law: the non-protection of purely AI-generated works, the threshold of human contribution in AI-assisted works, the legality of training datasets and TDM uses, and the balance between creator protection and innovation incentives. On this basis, the paper proposes a disclosure-and-accountability regime for Korea comprising standardized AI-use labeling and machine-readable provenance metadata, registration gatekeeping via human-contribution statements and logs, TDM governance using opt-out or collective licensing mechanisms, and the application of ordinary copyright liability rules to infringing AI outputs. Tables and conceptual figures compare policy options and summarize expected effects on transparency, legal certainty, and industrial predictability. The study concludes that a disclosure-first model which preserves the human-authorship baseline while improving transparency and training-data governance offers a pragmatic near-term path for aligning AI innovation with sustainable cultural and creative industries.

This study explored how artificial intelligence (AI) could be integrated into hair design education. The Beauty industries undergo significant changes driven by advancement of AI. Hair design education that is traditionally rooted in hands-on techniques and creative expression provides benefits of enhancing students’ study process and creative scope by integrating AI tools. It offers opportunities for hair design education by enabling style simulation, personalized learning, and immediate feedback. AI offers promising potential for interactive, efficient learning, fostering creativity, and trend recognition when complemented by traditional education tools. This study searched methods of enhancing the learning outcomes in the hair design education by using AI. We referenced course curricula related to hair design that applied AI in fields such as makeup, cosmetic package design, and virtual hairstyle fitting, and refined them to draw meaningful implications. We delivered a valuable outcome by developing a curriculum that incorporates AI. We made an important and distinctive contribution compared to previous research, and we provided a foundation for further discussion on the benefits, limitations, and possibilities of AI utilization in hair design education.

Recent shifts toward digitally mediated learning have made Learning Management Systems (LMSs) essential, yet many platforms still function mainly as one-way channels for distributing content. This limitation restricts timely exchanges between instructors and learners, which in turn reduces engagement and weakens students’ sense of progress. To respond to these issues, the present study proposes a redesign of the LMS architecture grounded in three key principles of Human–Computer Interaction (HCL): Feedback, Affordance, and Adaptivity. The proposed framework was examined through a Delphi-based expert review to assess both its technical feasibility and its capacity to support more interaction-oriented system behaviors. Findings indicate that embedding HCL principles into the LMS structure substantially strengthens real-time feedback mechanisms and contributes to more interactive learning experiences.

This study examines Daejeon's “Kumssi Family” as an affective nudge for civic participation, leveraging the role of "fun" in user experience (UX) (Nudge [1], Emotional Design [2]). Analysis identified three core strategies: Visual-Emotional Nudge, Participatory Nudge, and Experiential Diffusion Nudge. Achieving KRW 1.6 billion in goods sales, a Presidential Award at the 21st Korea Local Government Management Expo [3], and a Brand Excellence Award [4], the project demonstrates that fun successfully transforms public characters into affective UX platforms that reinforce civic empathy.

Although Busan-Gyeongnam is a region with a high influx of tourists, the existing design of public facilities in urban railway systems has been primarily developed for general Korean users. Consequently, foreign tourists and transportation-vulnerable individuals often face inconvenience due to differences in language, culture, and physical abilities. In particular, the lack of consistency in the signage system and unclear foreignlanguage labeling have intensified confusion in wayfinding. Accordingly, this study aims to apply the seven principles of universal design to diagnose the current state of guidance design and public facilities at Busan- Gyeongnam KTX stations, and to propose integrated system design standards that ensure all users can access and use these spaces safely and conveniently. As part of the research process, the study examined the signage systems and the installation status of accessibility facilities for transportation-vulnerable users in key areas such as station entrances, concourses, and platforms of the Busan urban metro. In particular, the study focused on analyzing, from the perspective of universal design, the appropriateness of multilingual labeling, potential confusion in line color symbolism, and problems related to the placement and construction of guidance facilities. Based on the results of the current condition analysis and the integration of universal design principles, this study presents improvement strategies for signage systems and public facilities that consider the movement scenarios of foreign tourists. Ultimately, this research contributes to establishing a sustainable design environment in which urban railway stations, as public spaces, embrace the diversity of all users.

This study aims to design barrier-free and eco-friendly children's play facilities that can be used by both people with and without disabilities without any restrictions. Currently, most play facilities in Korea are inaccessible to the socially vulnerable, and the proportion of accessible, inclusive playgrounds is only around 0.03%, necessitating significant improvements. To address this, we propose user-centered, inclusive spaces and customized play equipment utilizing cutting-edge technologies such as AI, 3D printing, and the Internet of Things (IoT). The study analyzed international safety standards—including those in the US and Germany— domestic case studies, and field surveys to derive four key design principles: accessibility, interactivity, sensory experience diversification, and sustainability. Through design simulations and practical application, we highlight integrated functionality, such as wheelchair-accessible ramps, safety-equipped equipment, and devices that stimulate auditory and tactile senses. The design methods and guidelines presented serve as a foundation for supporting the emotional and social development of children with and without disabilities. In conclusion, this study proposes practical strategies and implementation guidelines that can contribute to the expansion of inclusive play environments and the advancement of social inclusion. We hope to directly impact the development of inclusive play spaces by fostering continuous participation from users and collaboration.

While the development of Information and Communication Technology acts as a catalyst for social development and economic growth, the digital divide is presenting a threat by leaving some countries behind in the race to global competitiveness. Although numerous attempts have been made to analyze the digital divide in various global regions, there are few the detailed studies for the Asian countries. This study is to analyze and examine the level of the digital divide using digital achievement in this region. In order to this analysis, we propose a flexible framework for benchmarking countries’ achievement in ICT diffusion by suggesting a composite Digital Achievement Index (DAI), which is a modified index from the previous studies to target 34 Asian countries. The DAI is a composite index of these five sub-indexes-digital availability, digital mobile adoption, digital affordability, digital speed and internet skills. These five sub-indexes, in addition to those presented in previous studies, reflect factors such as the availability of relevant data in the Asian region and the availability of wireless digital technologies and services. The DAI provides a more comprehensive picture of where the countries stand in their evolution toward narrowing digital divide by measuring each country’s current digital achievement relative to other countries and proposes an important benchmark for assessing country-specific needs.

This paper proposes and validates a reproducible generative AI framework for re-presenting and artistically extending Korean traditional folding screens in digital media. A curated set of 70 high-resolution minhwa landscape images is processed with CLIP-based near-duplicate pruning at a cosine threshold of 0.985 and ViT-based tagging with manual refinement. The model is fine-tuned with LoRA on Stable Diffusion v1.5, and a parameter search identifies an effective regime of 25 sampling steps and a CFG scale of 7.0 to 7.5. Subtle image-to-video motion with a fixed camera preserves contemplative aesthetics while enhancing immersion, yielding three exhibited artifacts: mountain, water, and an integrated scene. The study evidences a shift in digital heritage from preservation to generative extension and provides empirical support for LoRA-based style transfer in media art, while noting limitations in dataset size and genre scope and outlining future work in real-time interactivity and VR/AR deployment.