Earticle

This study investigates the emergence of emotional dynamics and intrinsic motivation within multi-agent reinforcement learning (MARL) systems. Traditional MARL frameworks rely solely on extrinsic task rewards, which often limit exploration and adaptability. To address this, we propose an Affective-Motivated MARL (AM-MARL) framework where agents integrate curiosity-based intrinsic rewards and emotionmodulated affective feedback alongside extrinsic reinforcement. Agents operate in a continuous multiagent environment, learning through Q-learning, Actor- Critic, or Advantage Actor-Critic (A2C) methods depending on their action space. The intrinsic reward is defined as the state-prediction error between observed and expected future states, while the affective reward arises from temporal changes in emotional state and the social influence among peers. Experimental results show that incorporating intrinsic and affective rewards enhances exploration coverage, stabilizes emotional trajectories, and improves coordination efficiency compared to extrinsic-only baselines. These findings suggest that emotional feedback, when coupled with curiosity-driven intrinsic signals, fosters more humanlike adaptability, cooperative intelligence, and stable affect regulation in MARL environments.

Automated speech emotion recognition (SER) by efficient long-term temporal context modeling is a challenging task of the digital audio signal processing domain. However, by default, the recurrent neural network (RNN) is employed to incorporate the temporal dependencies in sequence to investigate the relationships among sequences and features. In this study, we design a parallel convolutional neural network (PCNN) for SER by using a squeeze and excitation network (SEnet) with the self-attention module. Additionally, we adopt the residual learning strategy in both module, SEnet and self-attention, which is further improve the performance of the network. Our proposed SER system utilizes speech spectrogram as input and extracts utterancelevel discrete features by using the PCNN model. We experimentally evaluated our proposed system by standard speech corpus, interactive emotional dyadic motion capture (IEMOCAP). The prediction result reveals the significance and robustness of the proposed PCNN system, which obtained a high recognition rate of 72.01% over state-of-the-art (SOTA) methods.

New breakthroughs were taken from the research of affective curve extracting from sequence-concentrated functional Magnetic Resonance Imaging (fMRI) images, and the emotional responses of human brain were hidden in these fMRI dataset; the purpose of this paper is to acquire critical features from fMRI images. The fMRI experiments were given by a certain theme emotion stimuli; firstly, component operations under bilateral filtering were applied for fMRI images’ morphological segmenting which reduced the computational space, for that the calculation was not based on the whole brain space. Operated by Fast Fourier Transform (FFT), fMRI images relative to functional area of human brain were pre-processed. Finally, time series based Power Spectrum Density (PSD) was founded by using an improved shape preserving fitting algorithm, and affective curves were acquired subsequently. The results showed the effectiveness of the proposed methodologies in this paper by comparing with cubic fitting and 5-th polynomial fitting operations. Experimental results also showed that this method was effective and efficient; the shaper preserving model had the lowest residual error that reflected the brain's emotional response curve adequately. The proposed methods have potential applications in the study of human-machine emotion interactions.

The main purpose of this study is to explore the potential of affective computing (AC) platforms in education through two phases of research: Phase I – platform analysis and Phase II – classification of academic emotions. In Phase I, the results indicate that the existing affective analysis platforms can be largely classified into four types according to the emotion detecting methods: (a) facial expression-based platforms, (b) biometric-based platforms, (c) text/verbal tone-based platforms, and (c) mixed methods platforms. In Phase II, we conducted an in-depth analysis of the emotional experience that a learner encounters in online video-based learning in order to establish the basis for a new classification system of online learner’s emotions. Overall, positive emotions were shown more frequently and longer than negative emotions. We categorized positive emotions into three groups based on the facial expression data: (a) confidence; (b) excitement, enjoyment, and pleasure; and (c) aspiration, enthusiasm, and expectation. The same method was used to categorize negative emotions into four groups: (a) fear and anxiety, (b) embarrassment and shame, (c) frustration and alienation, and (d) boredom. Drawn from the results, we proposed a new classification scheme that can be used to measure and analyze how learners in online learning environments experience various positive and negative emotions with the indicators of facial expressions.

With the rapid advancement of generative artificial intelligence and virtual human technologies, AI-driven educational systems had demonstrated significant potential in enabling personalized instruction and intelligent interaction. However, existing systems commonly suffer from limited emotional expressiveness, inadequate knowledge response mechanisms, and weak multimodal integration. This study proposed a multimodal AI tutor system based on Unreal Engine, integrating key technologies such as the GPT-4 language model, VITS-based speech synthesis, and NVIDIA Audio2Face for facial animation. To enhance content accuracy and adaptive responsiveness, a dual knowledge graph framework was introduced, comprising a structured teaching knowledge graph and a student cognitive intent graph. The system employs MetaHuman for high-fidelity avatar modeling and leverages Live Link to establish synchronized speech-expression feedback. Experimental results validate the feasibility of the proposed AI tutor system in virtual education environments, providing a practical foundation for the development of future intelligent educational platforms.

The market globalization and the firms’ internationalization hinder the matching of the top managers’ agenda, making it difficult to meet in the same space or time. On the one hand, the appearance of Ubiquitous Group Decision Support Systems (UbiGDSS) ena-bled individuals to gather and make decisions in different spaces at different times, but on the other hand, originated problems related to the lack of human interaction. To under-stand how the arguments used can influence each of the decision-makers, what is their satisfaction regarding the decision made, and other affective issues such as emotions and mood, are some examples of that lack. In order to try to overcome this lack, we propose a theoretical model that is specially designed for agents, helping to understand the interac-tions impact on each agent and their satisfaction with the decision made.

In this paper, we propose a method to predict human emotion based on a multi-dimensional emotion model using sentimental similarity. The proposed framework predicts the user's level of emotional response to an emotional stimulus, based on a sensitivity database that consists of self-assessment manikin-based integer-scale data, rated on various object stimulations by many users. We experimented on 72 users, with 1,073 stimulation objects, based on the International Affective Picture System, and used Thayer's arousal-valence 2-dimensional emotion model to verify the proposed framework. As a result, we have confirmed that the proposed framework can predict user emotion using an arousal-valence model.

본 연구는 인간의 커뮤니케이션(communication)을 컴퓨터연산(computation) 과 동일시하고 정당화하는 일련의 공학적, 신경생리학적 접근을 비판적으로 탐색한다. 이 접근에서는 인간 감정표현의 발생원리를－인종, 종족, 문화, 성별, 교육, 사회적 훈련 등의 맥락적 요소들을 가로질러 작동하는－보편적인 인간 신체의 생리활동에 둔다. 그 대표적 연구로서 로절린드 피카드의 “정서적 컴퓨팅” 과 안토니오 다마지오의 신경생리학적 뇌연구를 논한다. 언어와 의식의 문제를도외시하는 기술적 사고(technical reasoning)의 과학레토릭에 대한 비판적 논의를 통해 본 논문은 커뮤니케이션 연구에서 문화와 경험에 대한 해석적 역량의확대와 심화를 제안한다.

This article critically examine the way in which the idea of communication is justified in terms of computation in engineering and neurophysiological research on emotions and affect. The underlying ideas of human emotions and affect emphasize the universal physiology of the human body, from which emotions and affects of different races, ethnicities, cultures, and genders arise. Technical reasoning of the sort is located in Rosalind Picard’s “affective computing” and Antonio Damasio’s brain science. Findings of my rhetoric-of- science analysis suggest the urgency of reclaiming an alternative humanistic understanding of communication.

경제 불안정과 고령화, 경쟁격화 및 개인 가치관의 변화 등 사회 문제가 점점 심각해질 가능성이 있다. 이러한 상황에서 이를 해결 가능한 방안 중 하나로써 감성컴퓨팅 관련 연구가 증가하고 있다. 이에 본 연구는 감성컴퓨팅 연구 키워드를 중심으로 국내 및 글로벌 연구의 지식구조와 주요 키워드, 연구생산 현황 및 국가간 협력관계 및 주요 키워드별 네트워크 등을 파악하였다. 이를 위해 전문 학술데이터 베이스(Scopus)로부터 해당 키워드를 중심으로 논문을 검색하였으며, 서지분석과 네트워크 분석을 실시하였다. 중국과 미국이 Affective computing 분야에서 지식생산이 활발하였고, 한국은 약 10% 정도로 저조한 상황이다. 주요 키워드는 Affective computing을 중핵으로 주로 컴퓨팅 처리 및 감성분석, 인식을 분류하는 연구 및 사용자들의 모델링, 심리 분석이 주요 연구 키워드이다. 국가 간 협력구조는 중국과 미국이 가장 큰 클러스터를 형성하고 있고, 그 외에 영국, 독일, 스위스, 스페인, 캐나다 등이 협력을 주도하고 있다. 한국의 연구협력은 다양하지 않고 연구생산도 저조한 결과를 보였다. Affective computing 분야의 연구발전을 위해 미국, 중국 등 주요국과의 연구협력 강화와 연구파트너의 다양화를 위한 시사점을 결론으로 제언하였다.

Social problems, such as economic instability, aging population, heightened competition, and changes in personal values, might become more serious in the near future. Affective computing has received much attention in the scholarly community as a possible solution to potential social problems. Accordingly, we examined domestic and global knowledge structure, major keywords, current research status, international research collaboration, and network for each major keyword, focusing on keywords related to affective computing. We searched for articles on a specialized academic database (Scopus) using major keywords and carried out bibliometric and network analyses. We found that China and the United States (U.S.) have been active in producing knowledge on affective computing, whereas South Korea lags well behind at around 10%. Major keywords surrounding affective computing include computing, processing, affective analysis, research, user modeling categorizing recognitions, and psychological analysis. In terms of international research collaboration structure, China and the U.S. form the largest cluster, whereas other countries like the United Kingdom, Germany, Switzerland, Spain, and Canada have been strong collaborators as well. Contrastingly, South Korea's research has not been diverse and has not been very successful in producing research outcomes. For the advancement of affective computing research in South Korea, the present study suggests strengthening international collaboration with major countries, including the U.S. and China and diversifying its research partners.

본 연구에서는 사용자의 생체 신호를 반영한 감성 컴퓨팅 게임을 개발하였다. 생체 신호 측정을 위해 GSR(Galvanic Skin Response), FSR(Force Sensing Resistor), 온도 센서(Infrared Thermometer)를 장착한 마우스를 제작하였다. 해당 마우스를 통해 게임을 하는 사용자의 생체 신호를 비침투적으로 측정한다. 측정된 데이터는 실시간으로 처리되어 사용자의 긴장도를 3단계로 구분하며 구분된 긴장도는 반영되어 NPC(Non-Player Character)의 정서 반응과 스토리 분기의 변화를 가능하게 한다. NPC의 반응과 스토리 분기의 제작을 위해 Live 2d, Inkle Script를 사용하였다. 본 연구를 통해 사용자의 생체 신호를 이용한 감성 컴퓨팅 게임 제작에 하나의 방법론을 제시한다.

In this research, Affective computing game has been developed which reacts with a player's bio-signals. A modified computer mouse will be used to collect bio-signals by GSR, FSR, and infrared thermometer. This modified computer mouse collect human bio-signals in non-intrusive way. The collected data is complementary reflected in 3 level of tension of a player. The player's tension affects on the game and the reaction for NPC will be followed. Then this leads to plot changes individually. To let diverse NPC reaction and interactive story telling, Live 2d and Inkle Script have been used. This research can be alternative method on the game development using Affective computing.

This paper addresses the emotion computing model for software affective agents. In this paper, emotion is represented in valence-arousal-dominance dimensions instead of discrete categorical representation approach. Firstly, a novel emotion model architecture for affective agents is proposed based on Scherer's componential theories of human emotion, which is one of the well-known emotion models in psychological area. Then a fuzzy logic is applied to determine emotional statuses in the emotion model architecture, i.e., the first valence and arousal, the second valence and arousal, and dominance. The proposed methods are implemented and tested by applying them in a virtual training system for children's neurobehavioral disorders.