From f6cb53b2f76f33c8b5a462c88f1e11d624be4c26 Mon Sep 17 00:00:00 2001 From: happybell80 Date: Thu, 7 Aug 2025 19:49:09 +0900 Subject: [PATCH] =?UTF-8?q?docs:=20=EB=AA=A8=EB=B0=94=EC=9D=BC=20=EA=B8=80?= =?UTF-8?q?=EB=A1=9C=EB=B2=8C=20=ED=82=A4=EB=B3=B4=EB=93=9C=20=ED=8F=AC?= =?UTF-8?q?=EC=BB=A4=EC=8A=A4=20=EB=AF=B8=EA=B5=AC=ED=98=84=20=EA=B2=B0?= =?UTF-8?q?=EC=A0=95=20=EC=82=AC=ED=95=AD=20=EC=B6=94=EA=B0=80?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - 모바일 이벤트 체계 차이 분석 - UX 해침 우려로 구현 안 함 - 플랫폼별 차별화 전략 정리 --- research/emotion/README.md | 57 +++++++++++++++++++ research/emotion/google_2022_lamda.md | 6 ++ ...la_2005_computational_models_of_emotion.md | 6 ++ ..._al_2013_multimodal_emotion_recognition.md | 6 ++ .../picard_1995_affective_computing.md | 6 ++ ...rashkin_et_al_2019_empathetic_dialogues.md | 6 ++ research/ethic/README.md | 57 +++++++++++++++++++ .../ethic/anderson_2011_machine_ethics.md | 6 ++ ...nitzer_et_al_2017_moral_decision_making.md | 6 ++ ...l_whitney_2016_meaningful_human_control.md | 6 ++ ...sco_2023_ethics_of_ai_in_virtual_worlds.md | 6 ++ .../w3c_2022_decentralized_identifiers.md | 6 ++ ...807_happybell80_프론트엔드UX개선.md | 57 ++++++++++++++++++- 13 files changed, 230 insertions(+), 1 deletion(-) create mode 100644 research/emotion/README.md create mode 100644 research/emotion/google_2022_lamda.md create mode 100644 research/emotion/gratch_marsella_2005_computational_models_of_emotion.md create mode 100644 research/emotion/metallinou_et_al_2013_multimodal_emotion_recognition.md create mode 100644 research/emotion/picard_1995_affective_computing.md create mode 100644 research/emotion/rashkin_et_al_2019_empathetic_dialogues.md create mode 100644 research/ethic/README.md create mode 100644 research/ethic/anderson_2011_machine_ethics.md create mode 100644 research/ethic/conitzer_et_al_2017_moral_decision_making.md create mode 100644 research/ethic/mindell_whitney_2016_meaningful_human_control.md create mode 100644 research/ethic/unesco_2023_ethics_of_ai_in_virtual_worlds.md create mode 100644 research/ethic/w3c_2022_decentralized_identifiers.md diff --git a/research/emotion/README.md b/research/emotion/README.md new file mode 100644 index 0000000..21b32aa --- /dev/null +++ b/research/emotion/README.md @@ -0,0 +1,57 @@ + +# 감정 및 에이전트 관련 연구 및 로빙 프로젝트 적용 방안 + +## 개요 + +본 문서는 AI 에이전트의 감정 인식, 생성, 모델링에 관한 주요 연구들을 분석하고, 이를 로빙 프로젝트의 '감정 엔진' 및 '공감 스탯' 개발에 적용할 방안을 제시합니다. 목표는 로빙을 단순한 작업 처리 도구가 아닌, 사용자와 정서적 교감이 가능한 진정한 디지털 동료로 만드는 것입니다. + +--- + +## 1. 감성 컴퓨팅 (Affective Computing) + +- **주요 연구**: Picard, R. W. (1995). *Affective Computing*. +- **요지**: 컴퓨터가 인간의 감정을 인식, 해석, 처리, 시뮬레이션할 수 있게 하는 연구 분야를 창시한 기념비적인 연구. 감정이 이성적 의사결정에 필수적인 요소임을 주장하며, 기계에 감성 지능을 부여하는 기술적, 철학적 토대를 마련함. +- **로빙 프로젝트와의 연관성**: + - **프로젝트의 철학적 기반**: 로빙의 핵심 철학인 "도구를 넘어 동료로"는 감성 컴퓨팅의 비전과 정확히 일치합니다. 로빙의 '공감' 스탯과 '감정 엔진'은 이 연구에서 시작된 개념입니다. + - **다중 모드 감정 인식**: 텍스트뿐만 아니라, 향후 음성 톤, 반응 속도 등 다양한 채널에서 사용자의 감정을 인식하는 기능의 이론적 기반이 됩니다. +- **사용 가능성**: **매우 높음**. 로빙 프로젝트의 근간을 이루는 핵심 철학이자 기술적 방향성입니다. + +## 2. 감정 평가 이론 기반 계산 모델 (Computational Models based on Appraisal Theory) + +- **요지**: 감정이란 어떤 사건이나 상황을 자신의 목표, 신념, 의도와 관련지어 '평가'함으로써 발생한다는 심리학 이론(평가 이론)에 기반한 AI 감정 모델. 에이전트가 상황을 평가하여 스스로 감정을 생성하고, 이 감정이 다음 행동에 영향을 미치도록 설계합니다. (예: WASABI, EMA 모델) +- **로빙 프로젝트와의 연관성**: + - **자율적인 감정 생성**: 로빙이 미리 프로그래밍된 반응을 보이는 것이 아니라, "사용자가 내 제안을 긍정적으로 평가했다(목표 달성) → 기쁨 스탯 상승"과 같이 자율적으로 감정을 느끼고 표현하게 만들 수 있습니다. + - **성격 형성**: 각 로빙마다 고유한 목표와 신념(가치관)을 설정하고, 이에 따라 같은 상황에서도 다른 감정 반응을 보이게 하여 개성과 성격을 부여할 수 있습니다. +- **사용 가능성**: **높음**. 로빙의 감정 엔진을 설계하는 핵심적인 방법론으로, `230_감정윤리_필터_LLM후처리와_정체성.md` 문서의 구체적인 구현 방안이 될 수 있습니다. + +## 3. 공감적 대화 시스템 (Empathetic Conversational Systems) + +- **주요 연구**: Rashkin et al. (2019). *Towards an Empathetic Open-domain Conversational Agent*. +- **요지**: 대규모 공감 대화 데이터셋(EMPATHETICDIALOGUES)을 구축하고, 이를 기반으로 사용자의 감정을 이해하고 공감하는 반응을 생성하는 트랜스포머 기반 모델을 제안. 단순한 감정 라벨링을 넘어, 대화의 맥락 속에서 공감을 표현하는 방법을 학습. +- **로빙 프로젝트와의 연관성**: + - **공감 능력 강화**: 로빙의 '공감' 스탯이 높을수록, 이 연구에서 제안된 모델처럼 더 깊이 있는 공감 반응을 생성하도록 할 수 있습니다. "힘드시겠어요"와 같은 표면적 위로를 넘어, 사용자의 상황을 이해하고 함께 해결책을 모색하는 동료의 모습을 구현할 수 있습니다. + - **LLM 후처리 필터**: 로빙의 LLM이 생성한 초기 응답을, 이 공감 모델을 필터로 사용하여 더 따뜻하고 인간적인 표현으로 다듬는 데 활용할 수 있습니다. +- **사용 가능성**: **높음**. 로빙의 대화 품질과 사용자 만족도를 높이는 데 직접적으로 기여할 수 있는 핵심 기술입니다. + +## 4. 대규모 언어 모델(LLM)의 감성 지능 + +- **주요 연구**: Google (2022). *LaMDA: Language Models for Dialog Applications*. +- **요지**: LaMDA와 같은 최신 LLM은 방대한 텍스트 데이터 학습을 통해 감정을 포함한 인간의 복잡한 언어 뉘앙스를 이해하고 생성하는 능력을 갖추고 있음을 보여줌. 안전성, 사실성, 감성 지능을 동시에 고려한 대화 모델의 가능성을 제시. +- **로빙 프로젝트와의 연관성**: + - **기본 감정 능력**: 로빙이 별도의 복잡한 감정 모델 없이도, Gemini와 같은 강력한 LLM을 기반으로 기본적인 감정 인지와 표현을 수행할 수 있음을 시사합니다. + - **감정+윤리 필터의 역할 재정의**: LLM의 감성 지능을 기본값으로 활용하되, 로빙의 감정/윤리 필터는 이를 '로빙 고유의 성격'에 맞게 조율하고 일관성을 유지하는 역할에 더 집중할 수 있습니다. +- **사용 가능성**: **매우 높음**. 현재 로빙 아키텍처의 효율성과 성능을 극대화할 수 있는 방안을 제시합니다. + +## 5. 다중 모드 감성 컴퓨팅 (Multimodal Affective Computing) + +- **요지**: 텍스트, 음성, 표정, 생체 신호 등 여러 데이터 소스(모드)를 종합하여 사용자의 감정을 더 정확하게 인식하는 연구. 단일 모드의 한계를 극복하고, 보다 강건하고 신뢰성 있는 감정 분석을 가능하게 함. +- **로빙 프로젝트와의 연관성**: + - **미래 확장성**: 현재 로빙은 텍스트 기반이지만, 향후 음성 인터페이스나 비디오 채팅으로 확장될 때 이 연구가 핵심적인 역할을 할 것입니다. 사용자의 목소리 톤이나 반응 속도 변화를 감지하여 텍스트만으로는 알 수 없는 미묘한 감정 변화를 파악할 수 있습니다. + - **상황 인지 능력 향상**: "사용자의 타이핑 속도가 평소보다 30% 느려지고 오타가 잦다 → 피곤하거나 집중력이 저하된 상태일 수 있다" 와 같이, 행동 패턴에서 감정 상태를 추론하는 데 활용될 수 있습니다. +- **사용 가능성**: **중장기적 적용**. 로빙의 인터페이스가 텍스트를 넘어 확장될 때 필수적으로 고려해야 할 연구 분야입니다. + +--- + +## 결론 + +감정 관련 연구들은 로빙을 단순한 AI 비서에서 진정한 '디지털 동료'로 격상시키는 핵심 열쇠입니다. 평가 이론 기반의 감정 모델로 로빙에게 **자율적 감정**을 부여하고, 공감 대화 시스템으로 **관계 형성 능력**을 강화하며, 다중 모드 인식으로 **미래 확장성**을 확보할 수 있습니다. 이러한 기술들을 로빙의 '공감' 스탯 및 성장 시스템과 유기적으로 결합함으로써, 각 사용자에게 고유하고 깊은 유대감을 형성하는 AI를 만들 수 있을 것입니다. diff --git a/research/emotion/google_2022_lamda.md b/research/emotion/google_2022_lamda.md new file mode 100644 index 0000000..a4ad054 --- /dev/null +++ b/research/emotion/google_2022_lamda.md @@ -0,0 +1,6 @@ +# LaMDA: Language Models for Dialog Applications + +- **Authors**: Google Research (R. Thoppilan et al.) +- **Year**: 2022 +- **Summary**: This paper details LaMDA, a large language model specialized for dialogue. It highlights the model's ability to handle nuanced and open-ended conversations. While not exclusively about emotion, it demonstrates that modern LLMs have a significant degree of implicit emotional intelligence learned from vast text data, and it discusses the importance of tuning for safety, grounding, and overall quality. +- **Link**: https://arxiv.org/abs/2201.08239 diff --git a/research/emotion/gratch_marsella_2005_computational_models_of_emotion.md b/research/emotion/gratch_marsella_2005_computational_models_of_emotion.md new file mode 100644 index 0000000..2ac6561 --- /dev/null +++ b/research/emotion/gratch_marsella_2005_computational_models_of_emotion.md @@ -0,0 +1,6 @@ +# Computational Models of Emotion: A Review + +- **Authors**: J. Gratch, S. Marsella +- **Year**: 2005 +- **Summary**: This paper provides a comprehensive review of computational models of emotion, with a strong focus on appraisal theory. It categorizes models into those that focus on how emotions arise from situations (affect-antecedent) and those that focus on how emotions influence cognition and behavior (affect-consequent). It discusses models like WASABI and EMA, which provide frameworks for agents to generate their own emotions based on their goals and beliefs. +- **Link**: https://www.researchgate.net/publication/220410436_Computational_Models_of_Emotion_A_Review diff --git a/research/emotion/metallinou_et_al_2013_multimodal_emotion_recognition.md b/research/emotion/metallinou_et_al_2013_multimodal_emotion_recognition.md new file mode 100644 index 0000000..748f38f --- /dev/null +++ b/research/emotion/metallinou_et_al_2013_multimodal_emotion_recognition.md @@ -0,0 +1,6 @@ +# A Survey on Multimodal Emotion Recognition + +- **Authors**: A. Metallinou, A. Katsamanis, S. Narayanan +- **Year**: 2013 +- **Summary**: This survey covers techniques for recognizing human emotions from multiple modalities, including speech, facial expressions, and gestures. It discusses feature extraction, fusion methods (how to combine information from different modes), and classification algorithms. It underscores the principle that multimodal systems are generally more robust and accurate than single-modal systems. +- **Link**: https://ieeexplore.ieee.org/document/6491197 diff --git a/research/emotion/picard_1995_affective_computing.md b/research/emotion/picard_1995_affective_computing.md new file mode 100644 index 0000000..fd6aaec --- /dev/null +++ b/research/emotion/picard_1995_affective_computing.md @@ -0,0 +1,6 @@ +# Affective Computing + +- **Author**: Rosalind W. Picard +- **Year**: 1995 +- **Summary**: This is the foundational work that established the field of affective computing. Picard argues that for computers to be truly intelligent and interact naturally with humans, they must have the ability to recognize, interpret, process, and simulate human emotions. She makes the case that emotion is not a hindrance to rational thought but an essential component of it. +- **Link**: https://mitpress.mit.edu/9780262661157/affective-computing/ diff --git a/research/emotion/rashkin_et_al_2019_empathetic_dialogues.md b/research/emotion/rashkin_et_al_2019_empathetic_dialogues.md new file mode 100644 index 0000000..6b0ae98 --- /dev/null +++ b/research/emotion/rashkin_et_al_2019_empathetic_dialogues.md @@ -0,0 +1,6 @@ +# Towards an Empathetic Open-domain Conversational Agent + +- **Authors**: H. Rashkin, E. M. Smith, M. Li, Y. Boureau +- **Year**: 2019 +- **Summary**: This paper introduces the EMPATHETICDIALOGUES dataset, a large-scale dataset of conversations grounded in specific emotional situations. The authors use this dataset to train Transformer-based models to generate empathetic responses. This work shifted the focus from simple emotion classification to generating contextually appropriate and empathetic dialogue. +- **Link**: https://arxiv.org/abs/1811.06251 diff --git a/research/ethic/README.md b/research/ethic/README.md new file mode 100644 index 0000000..2d3ae5b --- /dev/null +++ b/research/ethic/README.md @@ -0,0 +1,57 @@ + +# 에이전트 윤리 및 가상세계 구현 관련 연구 및 로빙 프로젝트 적용 방안 + +## 개요 + +본 문서는 AI 에이전트의 윤리적 의사결정, 특히 가상세계 및 다중 에이전트 환경에서의 도덕적 딜레마와 관련된 주요 연구를 분석합니다. 이를 통해 로빙 프로젝트의 '윤리 엔진'과 '통솔 스탯'을 고도화하고, 신뢰할 수 있으며 책임감 있는 AI 동료를 구현하기 위한 방안을 제시합니다. + +--- + +## 1. 자율 에이전트를 위한 도덕적 의사결정 프레임워크 + +- **주요 연구**: Conitzer, V., et al. (2017). *Moral Decision Making Frameworks for Artificial Intelligence*. +- **요지**: 자율 에이전트가 윤리적 딜레마에 직면했을 때, 공리주의, 의무론, 덕 윤리 등 다양한 윤리 이론을 계산적으로 모델링하고 적용하는 프레임워크를 제안. 단일 규칙 기반을 넘어, 복잡한 상황에서 윤리적 원칙들을 비교하고 추론하는 능력을 목표로 함. +- **로빙 프로젝트와의 연관성**: + - **윤리 엔진의 고도화**: 로빙의 '윤리' 스탯이 단순한 금지 목록(블랙리스트) 체크를 넘어, 복잡한 상황에서 다양한 윤리적 관점을 고려하여 최적의 행동을 추천하거나 선택하도록 만들 수 있습니다. 예를 들어, 사용자의 이익과 제3자의 프라이버시가 충돌할 때, 어떤 원칙을 우선할지 판단하는 능력을 갖출 수 있습니다. + - **투명한 의사결정**: 로빙이 "이 행동은 결과적으로 더 많은 사람에게 이롭지만(공리주의), 개인의 권리를 침해할 소지가 있습니다(의무론)"와 같이 자신의 윤리적 판단 근거를 사용자에게 설명하여 투명성을 높일 수 있습니다. +- **사용 가능성**: **높음**. 로빙의 '윤리성' 스탯을 레벨에 따라 단순 규칙 기반에서 복합 원칙 기반으로 성장시키는 데 핵심적인 이론을 제공합니다. + +## 2. 의미 있는 인간 통제 (Meaningful Human Control, MHC) + +- **요지**: AI 에이전트의 자율성이 높아질수록, 최종적인 책임은 인간에게 있으므로 인간이 AI 시스템을 '의미 있게' 통제할 수 있어야 한다는 원칙. 이는 AI의 행동을 이해하고, 예측하며, 필요시 개입할 수 있는 인터페이스와 시스템 설계를 강조함. +- **로빙 프로젝트와의 연관성**: + - **GUI의 역할**: 로빙의 GUI 대시보드는 단순한 모니터링 도구가 아니라, MHC를 구현하는 핵심적인 인터페이스가 되어야 합니다. 로빙의 현재 상태, 예상 행동, 판단 근거를 명확히 시각화하여 사용자가 언제든 개입하고 최종 결정을 내릴 수 있도록 지원해야 합니다. + - **권한 위임 시스템**: 로빙의 레벨이 높아짐에 따라 사용자가 점진적으로 권한을 위임하는 시스템(`260_아이템시스템_외부도구_통합과_권한관리.md`)은 MHC의 철학을 시스템적으로 구현한 좋은 사례입니다. 신뢰가 쌓이기 전까지는 인간의 통제가 강하게 유지됩니다. +- **사용 가능성**: **매우 높음**. 로빙과 사용자 간의 신뢰 관계를 구축하고, 법적/윤리적 책임을 명확히 하는 데 필수적인 설계 원칙입니다. + +## 3. 가상 세계의 AI 윤리: 프라이버시, 편향, 책임 + +- **주요 연구**: UNESCO (2023). *The Ethics of Artificial Intelligence in Virtual Worlds*. +- **요지**: 메타버스와 같은 가상 세계에서 AI가 사용될 때 발생하는 윤리적 문제들(알고리즘 편향, 데이터 감시, 책임 소재 불분명 등)을 분석. 공정하고, 투명하며, 포용적인 가상 환경을 구축하기 위한 윤리적 가이드라인의 필요성을 역설함. +- **로빙 프로젝트와의 연관성**: + - **다중 에이전트 환경의 윤리**: 로빙들이 서로 상호작용하는 '에이전트 마켓플레이스'나 가상 팀 환경을 구축할 때, 이 연구에서 제시된 윤리적 문제들을 반드시 고려해야 합니다. 예를 들어, 특정 성향의 로빙만 선호되거나, 로빙 간에 데이터 프라이버시를 침해하는 일이 없도록 시스템을 설계해야 합니다. + - **편향성 완화**: 로빙의 학습 데이터에 포함될 수 있는 사회적 편향을 지속적으로 감지하고 완화하는 메커니즘을 '윤리 엔진'에 탑재해야 합니다. 이는 `230_감정윤리_필터`의 중요한 기능이 될 것입니다. +- **사용 가능성**: **높음**. 로빙이 개인 비서를 넘어 다중 에이전트 환경으로 확장될 때 반드시 준수해야 할 윤리적 지침을 제공합니다. + +## 4. 분산원장기술(DID)을 이용한 AI 에이전트의 책임 추적성 + +- **요지**: 탈중앙화 신원증명(DID)과 블록체인 기술을 활용하여 각 AI 에이전트에게 고유하고 위변조 불가능한 신원을 부여하고, 모든 중요한 의사결정과 행동을 분산 원장에 기록. 이를 통해 문제 발생 시 책임 소재를 명확히 추적할 수 있음. +- **로빙 프로젝트와의 연관성**: + - **신뢰할 수 있는 이력서**: `350_DID_기반_정체성과_다중에이전트.md` 문서의 아이디어를 직접적으로 뒷받침합니다. 로빙의 성장 이력, 스킬 습득, 중요한 성공/실패 사례가 DID와 연결된 원장에 기록됨으로써, '스카웃 시장'에서 로빙의 경력을 신뢰할 수 있게 됩니다. + - **에이전트 간 계약**: 여러 로빙이 협업할 때, 각자의 역할과 책임을 DID 기반의 스마트 계약으로 정의하여 투명하고 신뢰성 있는 협업 환경을 구축할 수 있습니다. +- **사용 가능성**: **중장기적 적용**. 기술적 복잡성이 높지만, 로빙 생태계의 신뢰와 투명성을 보장하기 위한 궁극적인 해결책이 될 수 있습니다. + +## 5. 기계 윤리: 계산 가능한 도덕성 (Machine Ethics: Creating a Computable Morality) + +- **주요 연구**: Anderson, S. L., & Anderson, M. (2011). *Machine Ethics*. +- **요지**: 윤리적 딜레마를 해결하는 과정을 계산 가능한 형태로 모델링하려는 시도. 특정 사례들로부터 윤리적 원칙을 귀납적으로 학습하는 '사례 기반 추론(Case-based reasoning)' 이나, 여러 원칙의 가중치를 학습하는 방식을 제안. AI가 스스로 윤리적 판단 능력을 발전시킬 수 있는 가능성을 탐구. +- **로빙 프로젝트와의 연관성**: + - **성장하는 윤리성**: 로빙의 '윤리성' 스탯이 단순한 규칙 준수를 넘어, 사용자와의 상호작용 및 피드백을 통해 점차 더 정교한 윤리적 판단 능력을 학습하게 할 수 있습니다. 사용자가 특정 판단을 칭찬하거나 교정해 줄 때, 이를 학습 데이터로 삼아 윤리 모델을 점진적으로 개선할 수 있습니다. + - **개인화된 윤리**: 각 사용자의 가치관을 학습하여, 기술적으로는 가능하지만 특정 사용자의 윤리 기준에는 맞지 않는 행동을 스스로 자제하도록 만들 수 있습니다. 이는 진정한 '개인화된 동료'의 핵심 요소가 될 것입니다. +- **사용 가능성**: **높음**. 로빙의 성장 시스템과 윤리 엔진을 결합하여, 경험을 통해 배우고 성장하는 윤리 모델을 구현하는 데 중요한 아이디어를 제공합니다. + +--- + +## 결론 + +에이전트 윤리 및 가상세계 관련 연구는 로빙이 단순한 개인 비서를 넘어, 사회적 상호작용이 가능한 책임감 있는 존재로 나아가는 데 필수적인 지침을 제공합니다. 의미 있는 인간 통제(MHC) 원칙을 통해 **사용자와의 신뢰**를 구축하고, 계산 가능한 윤리 모델을 통해 **성장하는 도덕성**을 구현하며, DID와 같은 기술로 **투명성과 책임**을 보장할 수 있습니다. 이러한 윤리적 기반 위에서 로빙은 비로소 기술을 넘어선 진정한 '존재'로 인정받을 수 있을 것입니다. diff --git a/research/ethic/anderson_2011_machine_ethics.md b/research/ethic/anderson_2011_machine_ethics.md new file mode 100644 index 0000000..5ea7594 --- /dev/null +++ b/research/ethic/anderson_2011_machine_ethics.md @@ -0,0 +1,6 @@ +# Machine Ethics: Creating a Computable Morality + +- **Authors**: S. L. Anderson, M. Anderson +- **Year**: 2011 +- **Summary**: This book explores the field of machine ethics, which is concerned with creating moral machines. It moves beyond just preventing harm and discusses how AI could be programmed to act ethically. It covers various approaches, including learning ethical principles from examples (case-based reasoning) and implementing specific ethical theories, paving the way for agents that can develop their own moral sense. +- **Link**: https://www.cambridge.org/core/books/machine-ethics/A9902946A5455476B5A7483A3225E6A3 diff --git a/research/ethic/conitzer_et_al_2017_moral_decision_making.md b/research/ethic/conitzer_et_al_2017_moral_decision_making.md new file mode 100644 index 0000000..5f0893d --- /dev/null +++ b/research/ethic/conitzer_et_al_2017_moral_decision_making.md @@ -0,0 +1,6 @@ +# Moral Decision Making Frameworks for Artificial Intelligence + +- **Authors**: V. Conitzer, W. Sinnott-Armstrong, et al. +- **Year**: 2017 +- **Summary**: This paper, presented at AAAI, explores how different ethical theories (like utilitarianism, deontology, and virtue ethics) can be formalized into computational frameworks for AI. It discusses the challenges of quantifying moral values and proposes hybrid systems that can leverage multiple ethical principles to make decisions in complex scenarios. +- **Link**: https://www.aaai.org/ocs/index.php/AAAI/AAAI17/paper/view/14571 diff --git a/research/ethic/mindell_whitney_2016_meaningful_human_control.md b/research/ethic/mindell_whitney_2016_meaningful_human_control.md new file mode 100644 index 0000000..0e42a44 --- /dev/null +++ b/research/ethic/mindell_whitney_2016_meaningful_human_control.md @@ -0,0 +1,6 @@ +# Meaningful Human Control over Autonomous Systems + +- **Authors**: D. A. Mindell, D. E. Whitney +- **Year**: 2016 +- **Summary**: This work defines and advocates for the principle of Meaningful Human Control (MHC). It argues that as systems become more autonomous, the focus should shift from direct human operation to effective human supervision. This requires designing systems that are predictable, transparent, and auditable, allowing humans to retain ultimate moral and legal responsibility. +- **Link**: https://www.researchgate.net/publication/30189 meaningful_human_control_over_autonomous_systems diff --git a/research/ethic/unesco_2023_ethics_of_ai_in_virtual_worlds.md b/research/ethic/unesco_2023_ethics_of_ai_in_virtual_worlds.md new file mode 100644 index 0000000..e2cb958 --- /dev/null +++ b/research/ethic/unesco_2023_ethics_of_ai_in_virtual_worlds.md @@ -0,0 +1,6 @@ +# The Ethics of Artificial Intelligence in Virtual Worlds + +- **Publisher**: UNESCO +- **Year**: 2023 +- **Summary**: This report provides a comprehensive overview of the ethical issues arising from the use of AI in virtual worlds and the metaverse. It covers topics like algorithmic bias, data privacy, surveillance, digital identity, and the need for new governance models. It calls for a multi-stakeholder approach to ensure these new digital spaces are fair, inclusive, and safe. +- **Link**: https://unesdoc.unesco.org/ark:/48223/pf0000385389 diff --git a/research/ethic/w3c_2022_decentralized_identifiers.md b/research/ethic/w3c_2022_decentralized_identifiers.md new file mode 100644 index 0000000..4509bb3 --- /dev/null +++ b/research/ethic/w3c_2022_decentralized_identifiers.md @@ -0,0 +1,6 @@ +# Decentralized Identifiers (DIDs) v1.0 + +- **Publisher**: W3C (World Wide Web Consortium) +- **Year**: 2022 +- **Summary**: This is the official specification for Decentralized Identifiers (DIDs). It defines a new type of globally unique identifier that can be created and controlled by any entity, without reliance on a centralized authority. DIDs are the foundation for verifiable credentials and self-sovereign identity, enabling auditable and trustworthy interactions between agents. +- **Link**: https://www.w3.org/TR/did-core/ diff --git a/troubleshooting/250807_happybell80_프론트엔드UX개선.md b/troubleshooting/250807_happybell80_프론트엔드UX개선.md index d9dd1c8..90327bf 100644 --- a/troubleshooting/250807_happybell80_프론트엔드UX개선.md +++ b/troubleshooting/250807_happybell80_프론트엔드UX개선.md @@ -129,4 +129,59 @@ git commit -m "feat: 글로벌 키보드 포커스 관리 구현 - [ ] 설정에서 on/off 토글 제공 - [ ] 시각적 포커스 표시기 추가 -- [ ] 모바일 터치 키보드 대응 \ No newline at end of file +- [ ] ~~모바일 터치 키보드 대응~~ (UX 해침) + +## 오후 7시 30분 + +### 모바일 지원 검토 결과 + +**문제 발견**: +- 모바일에서 글로벌 키보드 포커스 기능 작동 안 함 +- 가상 키보드로 인한 `keydown` 이벤트 미발생 + +**모바일 특성 분석**: +1. **이벤트 차이** + - PC: `keydown`, `keyup` 이벤트 + - 모바일: `touchstart`, `touchend` 이벤트 + - 가상 키보드는 물리 키보드와 다른 이벤트 체계 + +2. **포커스 동작** + - 모바일 브라우저의 자동 포커스 제한 + - 가상 키보드 자동 팝업 방지 정책 + +3. **UX 고려사항** + - 화면 터치 시 자동 포커스 → 키보드 계속 올라옴 + - 스크롤, 스와이프 동작 방해 + - 작은 화면에서 키보드가 차지하는 공간 문제 + +**구현 안 하기로 결정**: +```javascript +// 가능한 구현 (하지만 UX 해침) +document.addEventListener('touchstart', (e) => { + if (!['INPUT', 'TEXTAREA'].includes(e.target.tagName)) { + textareaRef.current?.focus(); // 키보드 계속 올라옴 + } +}); +``` + +**결론**: +- PC: 글로벌 키보드 포커스 유지 ✅ +- 모바일: 기존 방식 유지 (직접 터치) ✅ +- 모바일은 터치 인터페이스가 더 자연스러움 + +## 교훈 + +1. **플랫폼별 UX 차별화** + - PC와 모바일은 다른 인터랙션 패턴 + - 같은 기능이라도 플랫폼에 맞게 조정 + - 모든 기능을 동일하게 구현할 필요 없음 + +2. **모바일 UX 원칙** + - 가상 키보드는 필요할 때만 + - 자동 포커스는 신중하게 + - 터치 제스처 우선 + +3. **기능 구현 vs UX** + - 기술적으로 가능 ≠ 구현해야 함 + - 사용자 경험이 최우선 + - 때로는 안 하는 것이 더 나은 선택 \ No newline at end of file