Emotion and Attention: When the Heart`s Eye Guides the Mind`s Eye

( Sowon Hahn ) 서울대학교 인지과학연구소 2017 Journal of Cognitive Science Vol.18 No.2

The dichotomy between “cold” cognition and “hot” emotion is deeply rooted in our understanding of the human mind. Traditional research in cognitive science often ignores the role of emotion when pursuing an information processing model. Recently, however, research has supported the critical role of emotion in many cognitive processes, including attention, memory, and decision-making. The interaction between attention and emotion is especially important in the adaptive sense because the ability to quickly respond to meaningful stimuli is essential for survival. The current review uses the distinction between top-down (goal-directed) and bottom-up (stimulus-driven) attention to investigate the relationship between emotion and attention. In particular, this review focuses on the relationship among visual selective attention and affective states, individual differences of emotional processing, and neural bases for the interaction between emotion and attention. Converging evidence shows that emotion can prioritize stimuli in the environment and that attention plays a role in controlling emotional processing.

꽃의 색채 특성 분석

한소원(HAHN, Sowon),김현경(KIM, HyunKyung),김영인(KIM, YoungIn) 한국색채학회 1999 한국색채학회 학술대회 Vol.1999 No.12

한국 성인남녀 선호의복의 색채와 색채이미지

김영인(Kim Youngin),한소원(Hahn Sowon) 한국색채학회 2000 한국색채학회 논문집 Vol.14 No.1

Enhancing Empathic Reasoning of Large Language Models Based on Psychotherapy Models for AI-assisted Social Support

이윤경,이인주,신민정,베서연,Sowon Hahn 한국인지과학회 2024 인지과학 Vol.35 No.1

Building human-aligned artificial intelligence (AI) for social support remains challenging despite the advancement of Large Language Models. We present a novel method, the Chain of Empathy (CoE) prompting, that utilizes insights from psychotherapy to induce LLMs to reason about human emotional states. This method is inspired by various psychotherapy approaches—Cognitive-Behavioral Therapy (CBT), Dialectical Behavior Therapy (DBT), Person-Centered Therapy (PCT), and Reality Therapy (RT)—each leading to different patterns of interpreting clients’ mental states. LLMs without CoE reasoning generated predominantly exploratory responses. However, when LLMs used CoE reasoning, we found a more comprehensive range of empathic responses aligned with each psychotherapy model’s different reasoning patterns. For empathic expression classification, the CBT-based CoE resulted in the most balanced classification of empathic expression labels and the text generation of empathic responses. However, regarding emotion reasoning, other approaches like DBT and PCT showed higher performance in emotion reaction classification. We further conducted qualitative analysis and alignment scoring of each prompt-generated output. The findings underscore the importance of understanding the emotional context and how it affects human-AI communication. Our research contributes to understanding how psychotherapy models can be incorporated into LLMs, facilitating the development of context-aware, safe, and empathically responsive AI.

로봇의 신뢰회복 행동이 인간-로봇 상호작용에 미치는 영향

맹호영,김환이,박재은,Sowon Hahn 한국인지과학회 2022 인지과학 Vol.33 No.4

This study aimed to confirm the effect of social and relational behavior types of robots on human cognition in human-robot interaction. In the experiment, the participants evaluated trust in robots by watching a video on the robot Nao interacting with a human, in which the robot made an error and then made an effort to restore trust. The trust recovery behavior was set as three conditions: an internal attribution in which the robot acknowledges and apologizes for an error, a condition in which the robot apologizes for an error but attributes it externally, and a non-action condition in which the robot denies the error itself and does not take any action for the error. As the result, in all three cases, the error was perceived as less serious when the robot apologized than when it did not, and the ability of the robot was also highly evaluated. These results provide evidence that human attitudes towards robots can respond sensitively depending on the robot's behavior and how they overcome errors, suggesting that human perception towards robots can change. In particular, the fact that robots are more trustworthy when they acknowledge and apologize for their own errors shows that robots can promote positive human-robot interactions through human-like social and polite behavior. 본 연구는 인간 로봇 상호작용에서 로봇의 사회적이고 관계적인 행동 유형이 인간의 인식에 끼치는 영향을 확인하고자 하였다. 이를 위한 실험에서는 연구 참여자들이 로봇 나오가 인간과 상호작용 하면서 로봇이 오류를 일으키고 신뢰회복을 위한 행동을 영상으로 시청한 후 로봇에 대한 신뢰를 평가하였다. 신뢰회복 행동은 로봇이 오류를 인정하고 사과하는 내부 귀인, 오류가 있었음을 사과하지만 외부로 귀인하는 조건, 오류 자체를 부인, 오류에 대해 아무런 사후 행동을 하지 않는 비 행동 조건으로 설정하였다. 이후 로봇에 대한 인간의 평가를 3가지 측면에서 분석하였다. 첫째, 로봇의 유능함과 정직성에 기반한 신뢰, 둘째 로봇에 대한 지각된 유능함과 정직성, 그리고 로봇의 오류로 인한 신뢰 위반에 대하여 오류의 심각성을 어떻게 지각하는지 탐색하였다. 실험의 결과는 3가지 모든 경우에서 로봇이 사과하지 않을 때보다 사과할 때 오류가 덜 심각하다고 지각하였으며 로봇에 대한 능력 또한 높이 평가하였다. 이러한 연구 결과는 로봇의 행동유형과 오류 극복 방법에 따라 로봇에 대한 인간의 태도가 민감하게 반응 할 수 있다는 근거를 제공하며 로봇에 대한 인간의 지각이 변할 수 있음을 시사한다. 특히 로봇이 스스로의 오류를 인정하고 사과하는 것이 더 신뢰를 높인다는 결과는 로봇이 인간처럼 사회적이고 매너있는 행동을 통해 긍정적인 인간 로봇 상호작용을 증진시킬 수 있음을 보여준다.

COVID-19 시대 대규모 감정일기에서 관찰된 조망수용능력과 심리적, 정서적 특성의 관계: 군집분석을 사용한 접근

김지원,정윤원,이윤경,Sowon Hahn 한국건강심리학회 2023 한국심리학회지 건강 Vol.28 No.1

본 연구에서는 인간에게 공감할 수 있는 AI모델을 구축하기 위해 수집된 대규모 COVID-19 감정 일기 데이터를 기반으로, 제3자가 분류한 조망수용수준과 심리적, 정서적 특성의 관계를 확인하였다. 먼저 연구참여자 5,403명의 심리적 특성을 측정하기 위해 활용된 단축형 자기보고식 척도의 신뢰도를 측정하였다. 그 결과 신경과민성, 우호성, 외로움, 공감, 지각된 스트레스의 문항이 일부 선별되었다. 이후, 군집분석을 통해 연구참여자를 세 개의 조망수용 군집으로 나누고 각 군집간 심리적 특성의 차이를 살펴보았다. 높은 조망수용능력 군집은 낮은 조망수용능력 군집보다 높은 수준의 우호성, 공감 수준을 보였고, 보다 낮은 일기스트레스와 낮은 부정정서 수준을 나타냈다. 낮은 조망 수용능력 군집은 세 집단 중 가장 높은 수준의 신경과민성, 외로움, 지각된 스트레스, 일기스트레스, 그리고 높은 부정정서 수준을 보였다. 조망수용능력이 확인되지 않은 미결정 군집은 가장 낮은 신경과민성, 외로움, 지각된 스트레스와 부정정서 수준을 보였다. 미결정 집단을 제외한 조망수용수준과 심리적, 정서적 특성의 관계는 선행연구에서 시사된 바와 일치하였다. 마지막으로 논의에서는 본 연구의 의의와 빅데이터를 활용한 정신건강개입 발전 가능성에 대한 시사점을 제시하였다. Using large data collected during COVID-19, the present study explored how perspective-taking levels of diary writers evaluated by the third party were associated with their psychological characteristics. We analyzed the perspective-taking levels of the 5,403 participants who wrote diary entries on their thoughts and emotions, along with their self-reported psychological characteristics. Cluster analysis revealed three distinct perspective-taking profiles. We examined differences in psychological traits between groups. The higher perspective-taking group exhibited higher agreeableness, higher trait empathy, lower diary stress, and less unpleasant feelings than the lower perspective-taking group. Neuroticism, loneliness, perceived stress, diary stress, and unpleasant feelings were the highest in the lower perspective-taking group. Neuroticism, loneliness, perceived stress, and unpleasant feelings were the lowest in the undetermined group. Our findings mirrored the previously documented links between self-reported perspective-taking abilities and psychological qualities. Applications for large-scale text-based mental health therapies were also discussed in this study.

Effect of Feedback Styles in Robotic Fitness Coaching on Exercise Intensity

Yoon Kyung Lee,Yongha Park,Minjung Shin,Sowon Hahn 한국HCI학회 2024 한국HCI학회 논문지 Vol.19 No.2

Intensive exercise effectively prevents adult diseases, such as cardiovascular disease; however, many people find it challenging to keep this routine due to many situational and financial constraints. Recently, a growing interest has been in deploying social robots as fitness coaches in exercise sessions. This study investigated the effectiveness of real-time feedback provided by robot coaches in increasing exercise intensity. Participants were randomly assigned to either an empathic feedback group, focusing on emotional support, or a factual feedback group, offering information and advice only. All participants engaged in physical exercises (e.g., stretching and treadmill walking), receiving feedback on their walking speed and posture at 20-second intervals for 10 to 30 minutes. By analyzing variations in heart rate and treadmill speed, the study assessed the impact of the robot coach’s intervention on changes in exercise intensity. The results indicated that participants receiving factual feedback experienced significant heart rate and walking speed increases, maintaining higher walking speeds for extended periods. However, the post-experiment interview results revealed that while many participants preferred factual feedback, they also highlighted the importance of the robot’s emotional support (e.g., encouragement) in enhancing the overall exercise experience. In conclusion, the current study suggests that while informative feedback effectively improves actual exercise intensity, emotional support from a robot coach is also crucial for fostering long-term effects through continuous interaction.

로그 데이터를 활용한 노년층의 스마트폰 사용 연구

박기범(Gibeom Park),현우창(Woochang Hyun),한소원(Sowon Hahn),서봉원(Bongwon Suh) 한국HCI학회 2018 한국HCI학회 학술대회 Vol.2018 No.1

노년층 스마트폰 사용자의 경우, 스마트폰 관련 개념에 대한 이해가 높지 않음으로 인하여 자신의 스마트폰 활용 방법 및 과정을 상세하게 설명하지 못하는 경우가 종종 발생한다. 이는 노년층의 스마트폰 활용 연구를 수행하는 연구자들에게 정확하고 상세한 정보를 전달하는 것을 어렵게 하여, 조사된 결과가 실제 노년층의 스마트폰의 활용 정도를 정확하게 평가하지 못하게 되는 장애물로 작용할 수 있다. 이러한 장애를 극복하고자, 스마트폰 애플리케이션의 활용을 기록하는 앱 로거를 개발하여 노년층 사용자의 애플리케이션 사용 행태를 분석하는 방법을 제시하였다. 앱 로거는 사용자의 별다른 행위 없이도 애플리케이션의 사용 정보를 백그라운드에서 수집하여 서버로 전달할 수 있도록 개발하였다. 앱 로거를 활용하여 Android OS 기반 스마트폰을 사용하는 만 60 세 이상의 노년층 사용자 총 61 명의 애플리케이션 활용 데이터를 성공적으로 수집하였으며, 수집된 데이터를 분석한 결과 연구 참여자들의 스마트폰 활용 행태를 정확하게 분석할 수 있었다.

Effects of ICT Ability of the Elderly People for the Acceptability of Conversational Agent

Minjee Kim,Na Kyoung You,Yoonki Kim,Sowon Hahn,Myung Hwan Yun 대한인간공학회 2017 대한인간공학회 학술대회논문집 Vol.2017 No.4

Objective: This research is an analysis of how ICT ability are affected in each acceptance of factors in conversational agent, to invigorate the use of conversational agent for elderly people. Background: With the rapid increase in the population of elderly people, the ability of having independence and participating in social activities are one of the main issues to provide solutions for the problems of the aged. The control method for conversational agent is intuitive as it does not require an input device. This allows easier adaptation of interaction method even for the elderly people with less experience with ICT. Method: We recorded eight different conversation scenarios formulated for this research in advance, using the text-to-speech program. The laboratory was set to allow conversational interaction by transmitting the voice converted to the speaker through computer, upon hearing the articulations of the participants over a partition. The ICT ability of the participants, and acceptance of conversational agent was measured through a survey by seven-point Likert scale. Results: In order to enhance the acceptance of conversational agent among elderly people, it is required to develop a new design technique allowing elderly people to form trust towards conversational agent by perceiving it as a social presence. Moreover, it is necessary to consider the importance of elderly people to acknowledge the sociability and usefulness of conversational agent. Conclusion: In this research we compared various factors of acceptance among elderly people, utilizing conversational agent. Conversational agent would henceforth function as an opportunity not only for elderly people, but also for the context and users with difficulty controlling devices with certain body parts.

장노년층 스마트폰 활용능력에 따른 체질량지수 차이

김준식 ( Joon-sik Kim ),김정운 ( Jung-woon Kim ),한소원 ( Sowon Hahn ),김연수 ( Yeon-soo Kim ) 대한스포츠의학회 2018 대한스포츠의학회지 Vol.36 No.4

Purpose: The purpose of this study was to compare the difference of body mass index (BMI) to smart phone proficiency in men and women over the age of 60. Methods: Patients were divided into three groups with high (n=33), average (n=34), and low (n=33) smart phone proficiency. Fitness characteristics related to smart phone usage were evaluated by measuring cardiorespiratory endurance, grip strength, eye-hand coordination. As well, smart phone proficiency was evaluated by a self-reported questionnaire and a smart phone usability task that was composed of two categories: usage of the smartphone device itself and usage of phone applications. The differences in BMI of the subjects was analyzed by analysis of covariance adjusting for independent variables including age, smartphone usage period, eye-hand coordination, education and income. Results: There was a significant difference in BMI among the three groups after adjustment of age, eye-hand coordination, smartphone usage period, education and income. The results showed that the self-reported questionnaire showed a significant difference in BMI between high proficiency and low proficiency groups (high 24.88±2.46, low 23.37±2.56; p=0.037). Smart phone usability test results also showed a significant difference in BMI among the three groups (high 25.18±2.58, low 23.15±2.6; p=0.000 and high 25.18±2.58, middle 23.57.7±1.69; p=0.010). Conclusion: Our results suggest that high smart phone proficiency shows increased BMI in the elderly. This study suggests that people over the age of 60 who have high smartphone proficiency should be cautious of an increased BMI score.