共情的定义,AI共情的相关研究与经典研究,AI个案概念化的能力
共情的概念界定、结构维度与测量方法
这组文献共同回答"共情究竟是什么、如何测量"这一基础问题:既有跨学科元定义(共情系统九大维度)与共情定义演进、跨语言术语辨析,也有对"共情七迷思"的概念澄清;既包括临床/医生共情的成分框架与操作性量表(JSE、患者—执业者双视角量表),也涵盖认知共情(观点采择)这一下位构念的可训练性检验,以及AI健康研究中共情定义与评估标准不一致问题的系统梳理。它们构成全部研究的概念与测量基础。
- The Taxonomy of Empathy: A Meta-definition and the Nine Dimensions of the Empathic System(Michaela Guthridge, M. Giummarra, 2021, Journal of humanistic psychology)
- An empathy definition at last: exposing the narcissism of small differences.(Jeremy Howick, A. Bennett-Weston, 2025, The British journal of general practice : the journal of the Royal College of General Practitioners)
- EMPATİ: TANIMI VE ÖNEMİ / EMPATHY: DEFINITION AND ITS IMPORTANCE(Elif Ersoy, F. Köşger, 2016, Osmangazi Journal of Medicine)
- Physician empathy: definition, components, measurement, and relationship to gender and specialty.(M. Hojat, J. Gonnella, T. Nasca, S. Mangione, M. Vergare, Michael Magee, 2002, American Journal of Psychiatry)
- Patient and Practitioner Perspectives on the Definition and Measurement of Therapeutic Empathy: Qualitative Study(A. Bennett-Weston, Jeremy Howick, 2025, Journal of Participatory Medicine)
- Seven Empathy Myths: Correcting Misconceptions About a Complex Construct(A. Martingano, Sara H. Konrath, D. Blanch-Hartigan, Mark Davis, Judith A. Hall, Mollie A. Ruben, Justin J. Sanders, Rachel Schwartz, 2025, Social & Personality Psychology Compass)
- Clinical empathy: definition, measurement, correlates, group differences, erosion, enhancement, and healthcare outcomes(M. Hojat, V. Maio, Charles A. Pohl, J. Gonnella, 2023, Discover Health Systems)
- Patient insights into empathy, compassion and self-disclosure in medical large language models: results from the IPALLM III study(N. Carl, S. Haggenmüller, J. Winterstein, Lisa Nguyen, C. Wies, M. J. Hetz, M. Mangold, B. Grüne, M. S. Michel, T. Brinker, F. Wessels, 2025, World journal of urology)
- Effects of perspective-taking training based on relational frame theory for cognitive empathy and emotional empathy: Differences in perspective-taking according to various theoretical approaches(Yasuhiro Ooshima, Takashi Mitamura, 2025, PLoS ONE)
- Empathy in AI for Health and Care Settings—Definition, Expression, and Measurement: Protocol for a Scoping Review(Alastair Howcroft, S. Benford, Michel F Valstar, H. Blake, 2026, JMIR Research Protocols)
共情定义的理论争议:对限制性同构匹配(RIM)界定的批判与功能性/关系性重构
这组文献属于共情定义的理论论辩而非实证测量:核心是对"限制性同构情绪匹配(RIM)"界定的批判(偏离传统概念、削弱共情的功能性、隐含动机性内容),以及将共情视为动态过程、强调关系性、他人情绪需求动态性与情境依赖的替代性概念化,并辅以对共情心理支柱的综述性梳理。它们为下文讨论"AI共情"提供非人类中心、功能导向的定义立场。
- Why We Should Reject the Restrictive Isomorphic Matching Definition of Empathy(Brett A. Murphy, S. Lilienfeld, Sara B. Algoe, 2022, Emotion Review)
- Comment: Empathy as a Flexible and Fundamentally Interpersonal Phenomenon: Comment on “Why We Should Reject the Restrictive Isomorphic Matching Definition of Empathy”(Alexandra Main, 2022, Emotion Review)
- A Mini-survey on Psychological Pillars of Empathy for Social Robots: Self-Awareness, Theory of Mind, and Perspective Taking(Elahe Bagheri, 2023, International Journal of Social Robotics)
AI/大语言模型共情能力的评估证据:系统综述与评测基准
这组文献从"能力评估"的元层面审视AI共情:系统综述汇总已有研究发现LLM已展现认知共情要素(情绪识别、情感支持),部分任务上超过人类,但存在重复套话、提示敏感、评价主观等局限;另有专门针对LLM共情评测的研究与面向共情评估的基准数据集(EmotionQueen)。共同点是提供可复用的评测方法学与AI共情能力的总体证据。
- Large Language Models and Empathy: Systematic Review(V. Sorin, D. Brin, Y. Barash, E. Konen, A. Charney, G. Nadkarni, Eyal Klang, 2023, Journal of Medical Internet Research)
- Empathy in large language models: Evaluation, enhancement, and challenges(Qian-Yi Zhou, Yaqi Cai, Ya Zhang, 2025, Advances in Psychological Science)
- EmotionQueen: A Benchmark for Evaluating Empathy of Large Language Models(Yuyan Chen, Songzhou Yan, Sijia Liu, Yue-Ze Li, Yanghua Xiao, 2024, Annual Meeting of the Association for Computational Linguistics)
LLM共情表达的实证检验与"人类—专家—AI"对比评价
这组文献以实证或综述方式将LLM生成的共情性回应/心理建议与人类及专业人员的文本进行对比评价:如共情倾听语言信号与应对自我效能、功能性共情与现象学共情的区分、LLM共情研究综述、盲评下AI与执业心理专家的共情质量、作者标签对共情归因的影响、合成医患对话中的共情评分,以及LLM是否具备认知共情。共同点是构成"AI能否共情"的证据链核心。
- Empathy on Demand: How Empathic AI Can Scale Emotional Support for Verbal Harassment(Anouk S Bergner, P. Winder, Christian Hildebrand, 2026, arXiv.org)
- Functional vs. phenomenological empathy in large language models: Rethinking artificial empathy through experimental evidence(A. Mugableh, Hissah Mohammed Alruwaili, 2026, East European Journal of Psycholinguistics)
- Large Language Models (LLMs) and Empathy - A Systematic Review(V. Sorin, D. Brin, Y. Barash, E. Konen, A. Charney, G. Nadkarni, E. Klang, 2023, medRxiv)
- Artificial intelligence vs. human expert: Licensed mental health clinicians' blinded evaluation of AI-generated and expert psychological advice on quality, empathy, and perceived authorship(Ludwig Franke Föyen, Emma Zapel, M. Lekander, E. Hedman-Lagerlöf, Elin Lindsäter, 2025, Internet Interventions)
- Human attribution of empathic behaviour to AI systems(J. Festor, Ivo Snels, B. Kleinberg, 2026, arXiv.org)
- Synthetic Patient–Physician Conversations Simulated by Large Language Models: A Multi-Dimensional Evaluation(S. A. Haider, Srinivasagam Prabha, C. A. Gomez-Cabello, S. Borna, Ariana Genovese, Maissa Trabilsy, B. Collaço, Nadia G. Wood, Sanjay Bagaria, Cui Tao, A. Forte, 2025, Italian National Conference on Sensors)
- Can large language models exhibit cognitive and affective empathy as humans?(Teng-Fei Yu, Si-Yu Pan, Caoyun Fan, Siyang Luo, Yaohui Jin, Bing-Lei Zhao, 2025, Comput. Hum. Behav. Artif. Humans)
医患沟通情境中AI共情的实证比较与"共情幻觉"争议
这组研究均在真实或类真实的医疗沟通情境中检验AI共情的效果、感知与风险:以析因实验比较聊天机器人与医生回应的共情、质量与信任评分并编码言语要素;发现癌症患者对聊天机器人回应的共情评分高于医生;利用面部分析情绪识别客观观察门诊医患情绪变化;以及围绕"医疗聊天机器人是否具有/需要共情"与"共情幻觉/同情幻觉"的评论性讨论。共同构成AI共情临床落地的证据与争议。
- What is Artificial Intelligence (AI) “Empathy”? A Study Comparing ChatGPT and Physician Responses on an Online Forum(Mollie A. Ruben, D. Blanch-Hartigan, Judith A. Hall, 2025, Journal of general internal medicine)
- Patient perceptions of empathy in physician and artificial intelligence chatbot responses to patient questions about cancer(David Chen, Kabir Chauhan, Rod Parsa, Zhihui Amy Liu, Fei-Fei Liu, E. Mak, Lawson Eng, B. Hannon, J. Croke, Andrew Hope, N. Fallah-rad, Phillip Wong, S. Raman, 2025, npj Digital Medicine)
- Emotion recognition in doctor-patient interactions from real-world clinical video database: Initial development of artificial empathy(Chih-Wei Huang, Bethany C. Y. Wu, P. Nguyen, Hsiao-Han Wang, Chih-Chung Kao, Pei-Chen Lee, A. Rahmanti, Jason C. Hsu, Hsuan-Chia Yang, Yu-Chuan Li, 2023, Computer methods and programs in biomedicine)
- Artificial empathy in healthcare chatbots: Does it feel authentic?(Lennart Seitz, 2024, Comput. Hum. Behav. Artif. Humans)
- The compassion illusion: Can artificial empathy ever be emotionally authentic?(Ajeesh K. G., Jeena Joseph, 2025, Frontiers in Psychology)
人工共情的理论模型与机器共情系统设计范式
这组文献聚焦"如何在机器中建构共情"的概念模型与工程设计范式:脆弱性驱动的具身共情以防止AI反社会、模糊逻辑支撑的多智能体认知共情决策、医疗与治疗场景中社会机器人与虚拟代理的共情增强、人机交互与具身对话智能体的共情行为建模、基于大语言模型的共情GPT架构,以及服务机器人在价值共创中的情感角色。共同点是给出人工共情的架构性、规范性设计主张。
- Preventing antisocial robots: A pathway to artificial empathy(Leonardo Christov-Moore, N. Reggente, Anthony G. Vaccaro, Felix A. Schoeller, Brock Pluimer, P. Douglas, M. Iacoboni, K. Man, António Sérgio Damásio, Jonas T. Kaplan, 2023, Science Robotics)
- Artificial Empathy and Imprecise Communication in a Multi-Agent System(Joanna Siwek, Konrad Pierzyński, Przemysław Siwek, Adrian Wójcik, Patryk Żywica, 2024, Applied Sciences)
- Artificial Empathy in Therapy and Healthcare: Advancements in Interpersonal Interaction Technologies(Tian-Yu Jia, Feiyu Pan, Xing-Chen Yang, Lin-Hong Ji, Dario Farina, Chong Li, 2025, Cyborg and Bionic Systems)
- Your Robot Will Feel You Now: Empathy in Robots and Embodied Agents(Angelica Lim, Ö. Yalçın, 2026, arXiv.org)
- Empathy-GPT: Leveraging Large Language Models to Enhance Emotional Empathy and User Engagement in Embodied Conversational Agents(Meng-Ting Shih, Ming-Yun Hsu, Sheng-Cian Lee, 2024, ACM Symposium on User Interface Software and Technology)
- Driving human-robot value co-creation in hospitality: The role of artificial empathy(Teng-Teng Zhu, Lishan Xie, Can-Mian Liu, 2025, International Journal of Hospitality Management)
AI共情的生成技术与数据集构建:从模型方法到交互实现
这组文献从技术实现角度讨论"如何工程化地生成共情":情感计算领域对共情线索提取与融入文本生成管线的综述、以情绪轨迹为中心的求助者—咨询师对话语料生成框架(EmoTrace)、支持共情互动与副语言表达的全双工语音对话大模型(JoyAI-Talker)、面向家庭服务智能体的共情式人机交互画布设计框架,以及基于对话情境与共情定义的Transformer回应生成、面向中文高语境共情的HED数据集与网络、用LLM推断用户潜在目标与基本心理需求的共情准确性度量。共同点是提供可实施的数据、模型与交互设计资源。
- A Survey on Integration of Empathy Cues Into Empathetic Text Generation(Edwin C. Montiel-Vázquez, Christian Arzate Cruz, J. R. Uresti, Miguel González-Mendoza, Randy Gomez, 2025, IEEE Access)
- EmoTrace: An Emotion Trajectory-Centered Framework for Psychological Support Dialogue Generation(Kai-Tong Weng, Lixiong Liu, Zi-Hao Liu, Bo Wang, Shi-Guang Ni, 2026, arXiv.org)
- JoyAI-Talker: Full-Duplex Speech Interactive Large Model Built for Empathetic Voice Agents(Jdg Com, 2026, arXiv.org)
- Artificial Empathy in Home Service Agents: A Conceptual Framework and Typology of Empathic Human–Agent Interactions(JooHyun Lee, Hyo-Jin Kang, 2025, Applied Sciences)
- Transformer-based Empathetic Response Generation Using Dialogue Situation and Advanced-Level Definition of Empathy(Yi-Hsuan Wang, Jia-Hao Hsu, Chung-Hsien Wu, Tsung-Hsien Yang, 2021, International Symposium on Chinese Spoken Language Processing)
- High-Context Empathy in Conversations for Large Language Models(Yu-Yan Chen, Lei Xia, Jing-Han Cao, Zhen-Dong Hou, Wei-Nan Dai, Zhi-Xu Li, 2025, International Conference on Information and Knowledge Management)
- Reading Users' Minds with LLMs: Mental Inference for Artificial Empathy in Design(Qi-Hao Zhu, Leah Chong, Maria Yang, Jianxi Luo, 2024, Journal of Mechanical Design)
人对AI共情的感知、接受度与信任:受控实验与量表评估
这组文献站在接受者一侧考察人对机器共情的主观反应:AI咨询师的高共情表达反而降低帮助性评价(提示需重新概念化"非人类共情")、共情模型越精致其准确性越低从而损害工作联盟、虚拟人共情表达应作为分级设计变量且视觉线索提升情感共情;同时包括情感智能聊天机器人使用体验的质性研究、情感识别AI对人际信任与共情影响的系统综述,以及量化技术感知共情的PETS量表开发。共同点是"共情表达并非越多越好、准确性比丰富度更关键"及感知—信任—伦理风险评估框架。
- An Empathetic AI for Mental Health Intervention: Conceptualizing and Examining Artificial Empathy(Ruosi Shao, 2023, Proceedings of the 2nd Empathy-Centric Design Workshop)
- Inaccurate Empathy is Worse than Minimal Empathy in Affective Human-Robot Interaction(Keith Rebello, Stacy Marsella, Timothy W. Bickmore, 2024, 2024 12th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW))
- The Effect of Empathic Expression Levels in Virtual Human Interaction: A Controlled Experiment(Sung Park, Daeho Yoon, Jungmin Lee, 2025, arXiv.org)
- Artificial Empathy: User Experiences with Emotionally Intelligent Chatbots(Mehdi Rostami, Shokouh Navabinejad, 2023, AI and Tech in Behavioral and Social Sciences)
- Affective Machines and Human Empathy: A Systematic Review of Emotional AI’s Impact on Social Interaction and Behavioral Trust(Irfan Ul Haq, Riffat Faizan, 2025, Scientific Societal & Behavioral Research Journal)
- Perceived Empathy of Technology Scale (PETS): Measuring Empathy of Systems Toward the User(Matthias Schmidmaier, Jonathan Rupp, Darina Cvetanova, Sven Mayer, 2024, International Conference on Human Factors in Computing Systems)
个案概念化的临床框架、能力培养与实证化个体建模
这组文献围绕"个案概念化能力"本身:论述其在临床实践与认知行为治疗研究中的核心地位、成功标准与培训价值,提供图式模式取向的具体概念化工具;并以RCT检验心理动力学个案概念化技能的可训练性与特异性,以PBAT与GIMME开展基于过程与个体化的实证概念化,以知觉因果网络(PECAN)实现个人化概念化与治疗定制,以社会效度与风险驱动取向重构ABA个案概念化。共同点是把个案概念化视为可教授、可量化、可个体化建模的临床核心能力。
- Case Conceptualization in Clinical Practice and Training(Eva Gilboa-Schechtman, 2024, Clinical Psychology in Europe)
- Case conceptualization research in cognitive behavior therapy: A state of the science review.(Michael H Easden, N. Kazantzis, 2018, Journal of Clinical Psychology)
- Using Schema Modes for Case Conceptualization in Schema Therapy: An Applied Clinical Approach(D. Edwards, 2022, Frontiers in Psychology)
- Case Conceptualization(L. Sperry, J. Sperry, 2020, No journal)
- Developing psychoanalytic case conceptualization skills through didactic teaching: A randomized controlled trial(Niccolò Fiorentino Polipo, J. Willemsen, M. Hustinx, Ariane Bazan, 2023, Psychotherapy Research)
- Toward Socially Meaningful Case Conceptualization: The Risk-Driven Approach(Rachel S. Taylor, Richard A. Colombo, M. Wallace, Benjamin Heimann, Ashton Benedickt, Allyson Moore, 2023, Behavior Analysis in Practice)
- Toward empirical process-based case conceptualization: An idionomic network examination of the process-based assessment tool(Brandon T. Sanford, Joseph Ciarrochi, Stefan G. Hofmann, Fredrick Chin, K. Gates, S. Hayes, 2022, Journal of Contextual Behavioral Science)
- How perceived causal networks can complement case conceptualization, diagnostic classification, and data-based networks: An introduction to a method for constructing personalized networks.(F. Vogel, T. Blanken, Julian Burger, J. Reichert, S. Scholten, L. Klintwall, 2025, Journal of Psychopathology and Clinical Science)
生成式AI辅助个案概念化:能力边界与伦理决策
这组文献将个案概念化推向智能化应用:一篇专题综述系统检视生成式人工智能在个案概念化中的角色、潜力和局限,另一篇引入AI"思考伙伴"支撑咨询教育中的个案概念化与伦理决策。共同点在于探讨AI能否及如何增强临床推理与概念化能力,同时警惕其准确性与伦理风险,直接对接"AI个案概念化的能力"这一主题。
- AI-assisted Case Conceptualization: Enhancing Ethical Decision-making in Graduate Counseling Education(Thang S. Tran, P. K. Rudra, J. Jacques, 2026, Journal of educational leadership in action)
- Topical review: Incorporating generative artificial intelligence into neuropsychology training: best practices, pitfalls, and recommendations for effective implementation.(Carly L. Clark, Beth A. Jerskey, Jason M. Fogler, 2026, Journal of Pediatric Psychology)
合并后的分组沿三条主线、共十个并列层级展开:(1)概念与测量层——共情的多元定义、结构维度、临床操作性量表,以及围绕限制性同构匹配(RIM)的理论争鸣;(2)AI共情层——从系统综述与评测基准、LLM共情表达的实证与\"人类—专家—AI\"对比、医患沟通情境中的临床证据与\"共情幻觉\"争议,到人工共情的理论模型与设计范式、共情生成技术与数据集构建,再到人对AI共情的感知、接受度与信任评估,形成\"能力—证据—技术—接受度\"的递进图景;(3)个案概念化层——临床框架与能力培训、实证化与个体化建模,以及生成式AI辅助个案概念化与伦理决策。三条主线在\"AI能否以及如何实现可被患者信任的共情、并有效支持临床个案概念化\"这一交汇问题上闭环,兼顾定义争议、技术实现与伦理风险。
这组文献共同回答"共情究竟是什么、如何测量"这一基础问题:既有跨学科元定义(共情系统九大维度)与共情定义演进、跨语言术语辨析,也有对"共情七迷思"的概念澄清;既包括临床/医生共情的成分框架与操作性量表(JSE、患者—执业者双视角量表),也涵盖认知共情(观点采择)这一下位构念的可训练性检验,以及AI健康研究中共情定义与评估标准不一致问题的系统梳理。它们构成全部研究的概念与测量基础。相关文献:Michaela Guthridge et. al, 2021 等 10 篇文献
这组文献属于共情定义的理论论辩而非实证测量:核心是对"限制性同构情绪匹配(RIM)"界定的批判(偏离传统概念、削弱共情的功能性、隐含动机性内容),以及将共情视为动态过程、强调关系性、他人情绪需求动态性与情境依赖的替代性概念化,并辅以对共情心理支柱的综述性梳理。它们为下文讨论"AI共情"提供非人类中心、功能导向的定义立场。相关文献:Brett A. Murphy et. al, 2022 等 3 篇文献
这组文献从"能力评估"的元层面审视AI共情:系统综述汇总已有研究发现LLM已展现认知共情要素(情绪识别、情感支持),部分任务上超过人类,但存在重复套话、提示敏感、评价主观等局限;另有专门针对LLM共情评测的研究与面向共情评估的基准数据集(EmotionQueen)。共同点是提供可复用的评测方法学与AI共情能力的总体证据。相关文献:V. Sorin et. al, 2023 等 3 篇文献
这组文献以实证或综述方式将LLM生成的共情性回应/心理建议与人类及专业人员的文本进行对比评价:如共情倾听语言信号与应对自我效能、功能性共情与现象学共情的区分、LLM共情研究综述、盲评下AI与执业心理专家的共情质量、作者标签对共情归因的影响、合成医患对话中的共情评分,以及LLM是否具备认知共情。共同点是构成"AI能否共情"的证据链核心。相关文献:Anouk S Bergner et. al, 2026 等 7 篇文献
这组研究均在真实或类真实的医疗沟通情境中检验AI共情的效果、感知与风险:以析因实验比较聊天机器人与医生回应的共情、质量与信任评分并编码言语要素;发现癌症患者对聊天机器人回应的共情评分高于医生;利用面部分析情绪识别客观观察门诊医患情绪变化;以及围绕"医疗聊天机器人是否具有/需要共情"与"共情幻觉/同情幻觉"的评论性讨论。共同构成AI共情临床落地的证据与争议。相关文献:Mollie A. Ruben et. al, 2025 等 5 篇文献
这组文献聚焦"如何在机器中建构共情"的概念模型与工程设计范式:脆弱性驱动的具身共情以防止AI反社会、模糊逻辑支撑的多智能体认知共情决策、医疗与治疗场景中社会机器人与虚拟代理的共情增强、人机交互与具身对话智能体的共情行为建模、基于大语言模型的共情GPT架构,以及服务机器人在价值共创中的情感角色。共同点是给出人工共情的架构性、规范性设计主张。相关文献:Leonardo Christov-Moore et. al, 2023 等 6 篇文献
这组文献从技术实现角度讨论"如何工程化地生成共情":情感计算领域对共情线索提取与融入文本生成管线的综述、以情绪轨迹为中心的求助者—咨询师对话语料生成框架(EmoTrace)、支持共情互动与副语言表达的全双工语音对话大模型(JoyAI-Talker)、面向家庭服务智能体的共情式人机交互画布设计框架,以及基于对话情境与共情定义的Transformer回应生成、面向中文高语境共情的HED数据集与网络、用LLM推断用户潜在目标与基本心理需求的共情准确性度量。共同点是提供可实施的数据、模型与交互设计资源。相关文献:Edwin C. Montiel-Vázquez et. al, 2025 等 7 篇文献
这组文献站在接受者一侧考察人对机器共情的主观反应:AI咨询师的高共情表达反而降低帮助性评价(提示需重新概念化"非人类共情")、共情模型越精致其准确性越低从而损害工作联盟、虚拟人共情表达应作为分级设计变量且视觉线索提升情感共情;同时包括情感智能聊天机器人使用体验的质性研究、情感识别AI对人际信任与共情影响的系统综述,以及量化技术感知共情的PETS量表开发。共同点是"共情表达并非越多越好、准确性比丰富度更关键"及感知—信任—伦理风险评估框架。相关文献:Ruosi Shao et. al, 2023 等 6 篇文献
这组文献围绕"个案概念化能力"本身:论述其在临床实践与认知行为治疗研究中的核心地位、成功标准与培训价值,提供图式模式取向的具体概念化工具;并以RCT检验心理动力学个案概念化技能的可训练性与特异性,以PBAT与GIMME开展基于过程与个体化的实证概念化,以知觉因果网络(PECAN)实现个人化概念化与治疗定制,以社会效度与风险驱动取向重构ABA个案概念化。共同点是把个案概念化视为可教授、可量化、可个体化建模的临床核心能力。相关文献:Eva Gilboa-Schechtman et. al, 2024 等 8 篇文献
这组文献将个案概念化推向智能化应用:一篇专题综述系统检视生成式人工智能在个案概念化中的角色、潜力和局限,另一篇引入AI"思考伙伴"支撑咨询教育中的个案概念化与伦理决策。共同点在于探讨AI能否及如何增强临床推理与概念化能力,同时警惕其准确性与伦理风险,直接对接"AI个案概念化的能力"这一主题。相关文献:Thang S. Tran et. al, 2026 等 2 篇文献
总计 57 篇相关文献
No abstract available
This article presents a brief history of the Jefferson Empathy Project, provides a definition of clinical empathy, and describes the Jefferson Scale of Empathy (JSE), the most frequently used instrument for measuring clinical empathy in health professions students and healthcare practitioners. Also, we highlight selected empirical findings from our team and other national and international researchers on correlates of clinical empathy showing significantly positive associations between scores on the JSE and measures of personal qualities that are conducive to relationship building. Findings include group differences on the JSE scores by gender, race/ethnicity, academic background, specialty interest, and clinical competence, as well as empirical findings on erosion and enhancement of clinical empathy in health professions students and practitioners. Additionally, significant associations are reported between physicians’ scores on the JSE, and pertinent/tangible healthcare outcomes in diabetic patients in the U.S. and Italy; plausible explanations are offered for psychosocial and neurobiological mechanisms possibly involved in the link between physician empathy and healthcare outcomes. These findings suggest that empathy must be placed in the realm of evidence-based medicine, considered an essential element of overall professional competence, assessed for admission decisions to medical schools and postgraduate training programs, and incorporated into professional development of all health professionals-in-training and in-practice.
No abstract available
Abstract Background Most definitions of therapeutic empathy are based on practitioners’ perspectives, and few account for patients’ views. Therefore, we do not understand what therapeutic empathy means to patients. Given that therapeutic empathy involves a relationship between patients and practitioners, the underrepresentation of the patient voice threatens to undermine the validity of therapeutic empathy definitions and subsequently, how the concept is measured, taught, and practiced. Objective The aim of the study is to explore the perspectives of patients and practitioners on the definition of therapeutic empathy and how it should therefore be measured. Methods A qualitative study, underpinned by a social constructivist stance, was conducted. Patients and practitioners were purposively sampled from a medical school and a school of health care to represent a diversity of lived experiences and health care professions. In-depth, semistructured interviews were undertaken, and the data were analyzed using reflexive thematic analysis. Data collection ceased upon reaching meaning saturation. Results In total, 16 participants (8 patients and 8 practitioners) were interviewed in June and July 2024. Reflexive thematic analysis generated three overarching themes that synthesize the views of patients and practitioners on therapeutic empathy and how it should be measured: (1) therapeutic empathy involves the practitioner showing the patient (that they are interested in the patient as a person, that they are actively listening, that they understand, that they are emotionally engaged, and that they are responding to their needs), (2) context matters (eg, the clinical scenario, time, and the patient), and (3) short, simple scales are a pragmatic approach to measurement. Conclusions Patients and practitioners have similar views about what empathy is and define therapeutic empathy as involving the practitioner demonstrating specific attitudes and behaviors to their patients. These attitudes and behaviors should be included in interventions to enhance therapeutic empathy and in measures of the concept. However, contextual factors may influence the expression of therapeutic empathy in practice. The findings highlight the need for, and can inform the development of, a short therapeutic empathy scale that allows the comparison of scores between patients, practitioners, students, and observers.
No abstract available
Background “Empathy” is widely discussed in health and care settings and is increasingly claimed as an attribute of artificial intelligence (AI) systems (eg, socially assistive robots and chatbots), but the term is used inconsistently across the literature. In research on AI in these settings, it is often unclear what authors mean by “empathic AI,” what systems do that is intended to be empathic, and how empathy is assessed. This matters because perceived empathy can shape users’ experience of AI-mediated support and their willingness to engage with these systems. Objective This study aims to map how empathy is defined, operationalized, and evaluated in peer-reviewed AI research in health and care settings and to describe interactional design features commonly reported in systems perceived as more empathic. Methods This protocol outlines a scoping review following Joanna Briggs Institute guidance and is reported in accordance with PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews). We use “AI” as an umbrella term and will extract and classify each system’s type (eg, rule-based or large language model–based). We will search PubMed (MEDLINE), Embase, PsycInfo, CINAHL, Scopus, IEEE Xplore, and the ACM Digital Library databases. Two reviewers will screen titles and abstracts using ASReview and full texts by using Rayyan. We will extract study characteristics, empathy definitions and framing, empathy-related system behaviors and design features, and evaluation methods, and synthesize findings thematically. Results This scoping review forms a part of the first author's doctoral research, funded by an Engineering and Physical Sciences Research Council studentship from October 2025. Pilot searches were conducted on January 20, 2026; full searches and synthesis are planned for 2026, with publication anticipated in 2027. The review will produce (1) a summary of how empathy is defined in AI research in health and care settings, (2) a grouped list of the main empathic interactional behaviors and design features described, and (3) an overview of how empathy is measured across studies. Where studies report empathy ratings, we will summarize which features are most commonly present in higher-rated systems within comparable contexts. Conclusions The review will provide a clearer picture of what researchers mean by “AI empathy” in health and care settings and what system features are most commonly used when trying to build it. These findings may help guide the development of more empathic AI systems. International Registered Report Identifier (IRRID) PRR1-10.2196/93078
A growing cadre of influential scholars has converged on a circumscribed definition of empathy as restricted only to feeling the same emotion that one perceives another is feeling. We argue that this restrictive isomorphic matching (RIM) definition is deeply problematic because (1) it deviates dramatically from traditional conceptualizations of empathy and unmoors the construct from generations of scientific research and clinical practice; (2) insistence on an isomorphic form undercuts much of the functional value of empathy from multiple perspectives of analysis; and (3) combining the opposing concepts of isomorphic matching and self-other awareness implicitly requires motivational content, causing the RIM definition to implicitly require the kind of non-matching emotional content that it explicitly seeks to exclude.
The conceptual diversity in the definition of empathy has resulted in descriptions of a highly heterogeneous collection of related phenomena, causing confusion as to what empathy actually is. Some of this heterogeneity arises due to disparate viewpoints across different disciplines. Capturing this transdisciplinary construct and arriving at a clear and unambiguous overarching definition of empathy will help provide a clearer outline of the fundamental dimensions of empathy, and will facilitate greater consistency in research and discussion of empathy across and between a range of disciplines. An inductive conceptual content analysis of the existing definitions of empathy was undertaken to distil the common higher order and lower order components of empathy definitions that have been used in the literature since 1980. A total of 146 definitions of empathy were sourced from a sample of 506 publications. Nine overarching dimensions were identified within the 146 definitions, including empathy as a catalyst, function, process, outcome, affective state, cognitive state, involving self and other, leading to a behaviour, and occurring in a specific context. The resultant meta-definition of empathy is “the ability to experience affective and cognitive states of another person, while maintaining a distinct self, in order to understand the other.” The results reveal empathy as a complex series of processes that we argue should be considered an “empathic system” given its multidimensional nature.
I strongly agree with the criticisms of the restrictive isomorphic matching (RIM) definition of empathy made by Murphy, Lilienfeld, and Algoe (2022), and largely agree with their conceptualization of empathy as a dynamic process best defined by its function. In this commentary, I extend this argument by emphasizing the relational, interpersonal aspects of empathy. It is my view that in order to understand the functions of empathy, we must take into account not only the internal experience of the individual empathizing, but also the individual (or group) whose perspective the empathizer is attempting to take. I highlight how the emotional needs of others are dynamic and require flexible adaptation and underscore the role of context in appreciating the function of empathy.
This study proposes an approach to transformer-based empathetic response generation using dialogue situation and advanced-level definition of empathy. First, SBERT is adopted to extract the dialog situation vector from the user's historical sentences. A BERT-based emotion detector, a topic detector and an information estimator are constructed for empathy-related feature extraction. The change of the emotional valance and the textual information gain, obtained from the emotion detector and the information estimator, are used for adversarial training of the transformer-based empathetic response generator. The loss function of the transformer is defined to measure how good the expected response in terms of fluency and empathy. The EmpatheticDialogues corpus was adopted for system training and evaluation on empathetic response generation. According to the experimental results, the BLEU score was increased to 7.821 after considering the dialogue situation feature and empathy definition, outperforming the comparison models. In terms of human subjective evaluation, three evaluation results of empathy, relevance and fluency for the proposed system are better than that for the baseline model.
The term empathy has become a buzzword in recent decades, and the concept has received both scholarly attention and has also been the focus of public interest, professional trainings, and policy initiatives. However, misconceptions about its nature persist. Our aim is to rectify these misunderstandings by highlighting claims about empathy that have been empirically refuted. We address seven myths about empathy: #1 People mean the same thing when they say “empathy,” #2 Empathy increases burnout, #3 Empathy cannot be measured, #4 Empathy comes effortlessly, #5 Empathy cannot be learned, #6 More recent generations lack empathy, #7 Women are naturally more empathic. These myths, selected due to their considerable implications, often contain a grain of truth but are usually exaggerated or misapplied, typically by generalizing findings from one narrow empathy definition to all empathy constructs. The term “empathy” is an umbrella term encompassing lower‐order constructs like compassion, personal distress, emotional congruence, perspective taking, and accurate interpersonal perception. We specify for which lower‐order empathy constructs each myth holds, for which constructs it is debunked (based on empirical evidence), and for which lower‐order constructs sufficient or consistent evidence exists to offer a conclusive verdict. We illuminate the complexities involved in discussing and studying empathy while debunking these prevalent misunderstandings. Our goal extends beyond merely refuting these myths; we strive to avert their potentially harmful impact on policies and society.
In recent years, affective computing has seen a surge in empathetic response generation algorithms. These incorporate different empathy-related concepts to increase their perceived behavioral empathy. This survey aims to define the predominant research directions and offer structure to the field through the lens of empathy cues. In this definition, we include the concepts, factors, and components of interpersonal communication that relate to the expression of empathy. We describe the four general methods taken for extracting empathy cues from multiple sources. Subsequently, we explore three major strategies used to incorporate these cues into the text generation pipeline. Then, we present the methods used by the most prominent algorithms for evaluating their results and present a novel hierarchical view of performance across the field. Finally, we present areas of opportunity for novel approaches.
Perspective-taking has garnered considerable attention across various psychological domains and is often conceptualized as cognitive empathy within the organizational model of empathy. Relational frame theory, a theoretical framework elucidating human behavior and language, addresses perspective-taking and offers corresponding training protocols. However, previous studies have not examined the relationship between training and perspective-taking, defined as cognitive empathy in the organizational model of empathy. This study investigated the impact of perspective-taking training rooted in the relational frame theory on perspective-taking, as conceptualized by the organizational model of empathy. To this end, 43 participants underwent a training regimen and completed pre- and post-training assessments using questionnaires that measure organizational empathy. The post-training scores for the subscales measuring cognitive empathy increased significantly (p = 0.031); however, The increase in the score remained significant trend after adjusting for training performance. The scores for some subscales of emotional empathy also increased after training. The limitations of this study include a bias in the sample and discrepancies in the results between scales measuring the same concept. Nonetheless, these results provide scope for debate regarding the impact of perspective-taking training under relational frame theory on perspective-taking within the organizational model of empathy. While considering the limitations of the obtained results, this study discusses the relationship between perspective taking based on relational frame theory and the organizational model of empathy, including emotional empathy.
No abstract available
No abstract available
No abstract available
No abstract available
Artificial Empathy in Therapy and Healthcare: Advancements in Interpersonal Interaction Technologies
The healthcare sector is challenged by critical workforce shortages, and this is causing an urgent need for innovative technologies to support or augment human roles. Although much of the research effort has focused on support and training of functional tasks, the emotional impacts that humans bring to the loop have often been overlooked. This gap is particularly pressing in healthcare and therapy, where empathy and emotional support are central to patient well-being. Unlike machines, humans possess a unique capacity for empathy, connecting emotionally with others and providing the essential support that fosters healing. Bridging this gap requires integrating affective elements, such as empathy, into therapeutic systems, which is the key to improving their effectiveness. This review explores groundbreaking techniques that integrate interpersonal interactions within therapy and healthcare, focusing on multiplayer games that strengthen real-time social connections, alongside social robots and virtual agents designed to simulate human-like affective interactions. Using artificial intelligence, these technologies aim to replicate complex human dynamics and foster artificial empathy, thus revolutionizing how we deliver care and support.
As artificial intelligence (AI) technology advances, there has been a diversification of home service functions and services, as well as a change in the applied technologies and functions. This is due to the fact that the needs and expectations vary depending on the purpose of performing the task in the same environment. Although interactions with AI often occur in the home environment, which is a personal space, there is a need for research that examines interactions in consideration of the concept of empathy. This study thus aims to identify previous studies that examine the interaction between users and technology and to systematize the elements of interaction that can be considered based on intelligent agents that are often used in the home environment. To this end, a framework was established to examine multifaceted elements through research that shows that the interaction between technology and users should be natural, with sophisticated psychological anthropomorphism. This study analyzed the literature for the establishment of an artificial empathy interaction system and presented an initial framework. Subsequently, we proceeded to the application of authentic industry cases to the framework, with the objective of ascertaining the feasibility of mapping groups exhibiting analogous trends. This process culminated in the categorization of these cases into three distinct types, alongside the identification of the empathy interaction elements that should be given consideration for each category. Moreover, we identified additional components necessary for the formulation of the final framework and elements that were deemed to be superfluous. Thereafter, we initiated the refinement process to elaborate the framework. The final framework is “Empathic HAX (human-agent interactions) Canvas”, which is designed to examine the necessity of empathic interaction between users and AI agents in the home service domain and to determine the optimal design for such interaction. The significance of this study lies in the creation of a framework that has not previously existed, and the presentation of a design tool that is highly likely to be used both academically and practically.
Artificial intelligence chatbots can draft empathetic responses to cancer questions, but how patients perceive chatbot empathy remains unclear. Here, we found that people with cancer rated chatbot responses as more empathetic than physician responses. However, differences between patient and physician perceptions of empathy highlight the need for further research to tailor clinical messaging to better meet patient needs. Chatbots may be effective in generating empathetic template responses to patient questions under clinician oversight.
This paper introduces a novel artificial intelligence model that integrates artificial empathy into the decision-making processes of collaborative agent systems. The existing models of collaborative behaviors, especially in swarm applications, lack the aspect of empathy, known to improve cooperation in human teams. Emphasizing both cognitive and emotional aspects of empathy, the introduced model navigates communication uncertainties and ambiguities, transforming these challenges into opportunities for learning and adaptation in dynamic environments. A significant feature of this model is its handling of imprecision through fuzzy logic, using fuzzy similarity measures in the decision process. The main objective of the presented research is to introduce a new model for improving cooperativeness in multi-agent systems with the use of cognitive empathy. Future research focus on implementing the model on physical platform and optimize the artificial empathy algorithms in the decision-making module.
BACKGROUND AND OBJECTIVE The promising use of artificial intelligence (AI) to emulate human empathy may help a physician engage with a more empathic doctor-patient relationship. This study demonstrates the application of artificial empathy based on facial emotion recognition to evaluate doctor-patient relationships in clinical practice. METHODS A prospective study used recorded video data of doctor-patient clinical encounters in dermatology outpatient clinics, Taipei Municipal Wanfang Hospital, and Taipei Medical University Hospital collected from March to December 2019. Two cameras recorded the facial expressions of four doctors and 348 adult patients during regular clinical practice. Facial emotion recognition was used to analyze the basic emotions of doctors and patients with a temporal resolution of 1 second. In addition, a physician-patient satisfaction questionnaire was administered after each clinical session, and two standard patients gave impartial feedback to avoid bias. RESULTS Data from 326 clinical session videos showed that (1) Doctors expressed more emotions than patients (t [326] > = 2.998, p < = 0.003), including anger, happiness, disgust, and sadness; the only emotion that patients showed more than doctors was surprise (t [326] = -4.428, p < .001) (p < .001). (2) Patients felt happier during the latter half of the session (t [326] = -2.860, p = .005), indicating a good doctor-patient relationship. CONCLUSIONS Artificial empathy can offer objective observations on how doctors' and patients' emotions change. With the ability to detect emotions in 3/4 view and profile images, artificial empathy could be an accessible evaluation tool to study doctor-patient relationships in practical clinical settings.
Background The use of artificial intelligence for psychological advice shows promise for enhancing accessibility and reducing costs, but it remains unclear whether AI-generated advice can match the quality and empathy of experts. Method In a blinded, comparative cross-sectional design, licensed psychologists and psychotherapists assessed the quality, empathy, and authorship of psychological advice, which was either AI-generated or authored by experts. Results AI-generated responses were rated significantly more favorable for emotional (OR = 1.79, 95 % CI [1.1, 2.93], p = .02) and motivational empathy (OR = 1.84, 95 % CI [1.12, 3.04], p = .02). Ratings for scientific quality (p = .10) and cognitive empathy (p = .08) were comparable to expert advice. Participants could not distinguish between AI- and expert-authored advice (p = .27), but perceived expert authorship was associated with more favorable ratings across these measures (ORs for perceived AI vs. perceived expert ranging from 0.03 to 0.15, all p < .001). For overall preference, AI-authored advice was favored when assessed blindly based on its actual source (β = 6.96, p = .002). Nevertheless, advice perceived as expert-authored was also strongly preferred (β = 6.26, p = .001), with 93.55 % of participants preferring the advice they believed came from an expert, irrespective of its true origin. Conclusions AI demonstrates potential to match expert performance in asynchronous written psychological advice, but biases favoring perceived expert authorship may hinder its broader acceptance. Mitigating these biases and evaluating AI's trustworthiness and empathy are important next steps for safe and effective integration of AI in clinical practice.
Given the accelerating powers of artificial intelligence (AI), we must equip artificial agents and robots with empathy to prevent harmful and irreversible decisions. Current approaches to artificial empathy focus on its cognitive or performative processes, overlooking affect, and thus promote sociopathic behaviors. Artificially vulnerable, fully empathic AI is necessary to prevent sociopathic robots and protect human welfare. Description To prevent AI agents from making harmful decisions, they may need embodied, vulnerability-driven artificial empathy.
BACKGROUND Artificial intelligence (AI) chatbots may be an asset to patient-provider communication, but not enough is known about how patients respond and how chatbots answer patients' questions. OBJECTIVE How perceptions of empathy, quality, trust, liking, and goodness vary by both the actual and perceived source of responses to patient questions (chatbot vs. actual physician). We also coded and compared key verbal elements in chatbot and physician responses. DESIGN This cross-sectional experimental study used chatbot and physician responses from Ayers et al. (2023) in a 2 (actual source: chatbot vs. physician) × 2 (perceived source: chatbot vs. physician) factorial design. PARTICIPANTS U.S.-based, English-speaking participants were recruited online (N = 1454). MAIN MEASURES Participants rated responses on empathy, quality, trust, liking, and goodness. Verbal content of the chatbot and physician responses was independently coded by trained research assistants to identify elements contributing to higher empathy ratings by participants. KEY RESULTS Replicating Ayers et al. (2023), participants rated chatbot responses as more empathic than physician responses (Cohen's d = 0.56, p < 0.001). Chatbot responses received higher empathy ratings than physician responses regardless of what participants were told about authorship (ηp2 = 0.60, p < 0.001). Empathy ratings were higher when participants thought the response was physician-authored, whether it was or not (ηp2 = 0.17, p < 0.001). Participant ratings of quality, trust, liking, and goodness followed the same pattern as empathy. Chatbot responses contained more coder-rated validation, reassurance, and non-judgmental language and were less rushed and more structured than physician responses (Cohen's d = 0.32 to 1.82, p's < 0.01). CONCLUSIONS AI-generated responses, with human oversight, could enhance computer-mediated clinical communication, although patient awareness of AI contributions may reduce perceptions of empathy. Identification of the specific verbal elements in AI-generated responses could augment communication and increase perceptions of empathic care.
The ability to express empathy is recognized as a key factor for non-human agents to provide effective support and companionship. However, the conceptualization of empathy goes against their non-human nature, as it is grounded in unique human experiences. Therefore, the question remains: can humans accept non-human agents expressing empathy, and if so, how? In a laboratory experiment, 123 participants engaged in a Zoom-based counseling process with either an AI-driven or a human-operated counselor who expressed high or low empathy during empathetic counseling to address individual stressors. The results suggest that overall, empathetic counseling increased counseling quality only when it came from a human-operated counselor. AI-provided empathetic counseling backfired and was rated as less helpful. As we further examine different aspects of empathy, we found that perceiving two sub-dimensions of empathetic support from AI predicted better counseling quality: offering attentive comments and providing hope. This study concludes that it is necessary to reconsider the non-human nature of AI and conceptualize artificial empathy accordingly.
This study aims to explore user experiences with emotionally intelligent chatbots, focusing on their perceived empathy, satisfaction with interactions, and the impact on user perception. Additionally, it seeks to identify the main challenges and future expectations users have towards these AI systems. Employing a qualitative research design, this study collected data through semi-structured interviews with 28 participants who had interacted with emotionally intelligent chatbots in various contexts. Thematic analysis was conducted to identify main themes, categories, and concepts within the data, providing insights into users' perceptions and experiences. The analysis revealed eight main themes: Perceived Empathy, Interaction Satisfaction, Trust and Security, Human-Like Interaction, User Adaptation, Impact on User Perception, Barriers to Engagement, and Future Expectations. These themes encompass categories such as Emotional Understanding, Contextual Sensitivity, Engagement Level, Privacy Concerns, Learning Curve, AI Capabilities, Technological Limitations, and Improvement Suggestions. Participants valued chatbots' ability to understand and adapt to their emotional states but highlighted challenges in achieving authentic empathy and expressed concerns over privacy and data security. Emotionally intelligent chatbots hold promise for enhancing user experiences through artificial empathy. However, the authenticity of empathy, coupled with ethical considerations such as privacy and security, presents significant challenges. Future developments should focus on improving the genuineness of empathetic responses, ensuring ethical use of AI, and addressing users' concerns to fully realize the potential of emotionally intelligent chatbots.
In human-centered design, developing a comprehensive and in-depth understanding of user experiences–empathic understanding–is paramount for designing products that truly meet human needs. Nevertheless, accurately comprehending the real underlying mental states of a large human population remains a significant challenge today. This difficulty mainly arises from the trade-off between depth and scale of user experience research: gaining in-depth insights from a small group of users does not easily scale to a larger population, and vice versa. This paper investigates the use of Large Language Models (LLMs) for performing mental inference tasks, specifically inferring users' underlying goals and fundamental psychological needs (FPNs). Baseline and benchmark datasets were collected from human users and designers to develop an empathic accuracy metric for measuring the mental inference performance of LLMs. The empathic accuracy of inferring goals and FPNs of different LLMs with varied zero-shot prompt engineering techniques are experimented against that of human designers. Experimental results suggest that LLMs can infer and understand the underlying goals and FPNs of users with performance comparable to that of human designers, suggesting a promising avenue for enhancing the scalability of empathic design approaches through the integration of advanced artificial intelligence technologies. This work has the potential to significantly augment the toolkit available to designers during human-centered design, enabling the development of both large-scale and in-depth understanding of users' experiences.
The personalization of psychopathology through the use of personalized symptom networks appears to be a promising approach for gaining deeper insights into the development and maintenance of mental disorders. One way to create such networks is by using the perceived causal networks (PECAN) method. In this method, respondents are systematically asked to quantify how their symptoms are causally linked. Answers are then visualized, either for the individual or aggregated for a group, as a directed network. PECAN can represent causal relations irrespective of their timescales and requires no data-hungry estimations. The following guidelines are intended to assist clinicians and researchers in the creation of personalized networks using the PECAN method. These networks can facilitate case conceptualization and personalization of treatments for individual patients and the description of groups of patients, revealing recurring feedback loops and central symptoms. Additionally, recommendations are provided regarding the procedures to be employed in the selection of nodes, assessment of edges, and visualization of the data. Furthermore, the potential for evaluating the reliability, validity, and clinical usefulness, as well as strengths, limitations, and future challenges of PECAN, is discussed. We conclude with an overview of the challenges of PECAN and a research agenda that highlights opportunities to improve the still very young method and implement it in clinical research and practice. (PsycInfo Database Record (c) 2025 APA, all rights reserved).
Case conceptualization is central to the success of the therapeutic process. However, integrative case conceptualization research has lagged behind research on integrating therapeutic intervention techniques. A successful case conceptualization provides (a) a dynamic, context-sensitive, yet parsimonious model of the client’s functioning; (b) relevant treatment targets and associated assessment procedures; and (c) a treatment plan including intervention phases and potential obstacles. Success in case conceptualization is a core clinical competency goal for trainees in clinical psychology and a career-long learning goal even for expert clinicians. Emerging technological trends and the formation of adversarial collaborative teams may assist research on the utility of well-constructed case conceptualizations.
Background: Syndromal classification has failed to produce a progressive science of case conceptualization for mental and behavioral health issues. An idiographic application of processes of change can provide a viable empirical functional analytic alternative if it could be linked to an idionomic approach , modeling idiographic effects first, and retaining nomothetic findings if they improve idiographic fit. Method: The present study examined this possibility by using the Process-Based Assessment Tool (PBAT), a new assessment tool linked to the Extended Evolutionary Meta-Model (EEMM) of Process-Based Therapy. The PBAT and items assessing common clinical outcomes were assessed repeatedly in 50 individuals in an experience sampling format over a 35 day period yielding at least 60 measurement occasions per person. These data were then analyzed in an idionomic fashion using Group Iterative Multiple Model Estimation (GIMME). Results: Analyses showed that the PBAT related to common clinical outcomes for virtually all participants in the individual complex networks identified by GIMME. Data showed that relationships had to be studied using an idionomic approach because participants ’ responses violated the ergodic assumptions underlying classical normative statistics. No overall group patterns were found. Subgroup relations did emerge for three common outcomes (sadness, anxiety, and life satisfaction) but most process to outcome relationships were idiographic. Idiographic networks were interpretable, however, using the broadened psychological flexibility approach of the EEMM. Conclusion: Idionomic network analysis of processes of change may provide a replicable form of empirical functional analysis and process-based case conceptualization.
Abstract Objective: Time-limited didactic interventions have been shown to be effective in developing “generic” case conceptualization skills. The objective of this study is to test whether similar interventions can be used to develop case conceptualization skills that are “specific” to a treatment modality. Method: University psychology students were randomized to a target (n = 62) or a control group (n = 62). The target group received a training on psychoanalytic case conceptualization skills based on the newly-developed operators model. The control group received a training on generic case conceptualization skills based on the well-established 5 Ps model. Results: The students’ self-efficacy for case conceptualization significantly increased in both groups. However, students in the target group reported a significantly greater increase in psychoanalytic case conceptualization skills and in their ability to make clinical inferences. The teaching method, as well as the case conceptualization models, were acceptable to students. However, the 5 Ps model was significantly more acceptable to students than the operators model. Conclusions: This is the first RCT to provide evidence that psychoanalytic case conceptualization skills can be developed through didactic teaching and that they constitute a specific set of skills that are not developed by learning generic case conceptualization skills.
The Behavior Analysis Certification Board (BACB) Ethics Code states that “behavior analysts should put compliance with the law and clients’ interests first by actively working to maximize desired outcomes and minimize risk ” (emphasis added; BACB, 2020 , p. 5). In turn, board certified practitioners must approach the case conceptualization process in applied behavior analysis (ABA) with respect to minimizing risks to an improved quality of life (QoL). As such, ABA services must be based on an understanding of risk —risk to ensuring desired outcomes . The purpose of the current article is two-fold (1) revisit social validity and propose features of socially meaningful case conceptualization, and (2) introduce a corresponding structured risk-driven approach to ABA service delivery. A primary aim is to equip all stakeholders with readily accessible practice-related supports—ensuring clients’ rights to effective services towards an improved QoL.
This article is situated within the framework of schema therapy and offers a comprehensive and clinically useful list of schema modes that have been identified as being relevant to conceptualizing complex psychological problems, such as those posed by personality disorders, and, in particular, the way that those problems are perpetuated. Drawing on the schema therapy literature, as well as other literature including that of cognitive behavior therapy and metacognitive therapy, over eighty modes are identified altogether, categorized under the widely accepted broad headings of Healthy Adult, Child modes, Parent modes and coping modes which are, in turn, divided into Surrender, Detached/Avoidant, and Overcompensator. An additional category is included: Repetitive Unproductive Thinking. This draws attention to the recognition by metacognitive therapists that such covert behaviors play a significant role in amplifying distress and perpetuating a range of psychological problems and symptoms. In addition to the modes themselves, several concepts are defined that are directly relevant to working with modes in practice. These include: default modes, blended modes, mode suites and mode sequences. Attention is also drawn to the way in which Child modes may be hidden “backstage” behind coping modes, and to the dyadic relationship between Child modes and Parent modes. Also relevant to practice are: (1) the recognition that Critic voices may have different sources and this has implications for treatment, (2) the concept of complex modes in which several submodes work together, and (3) the fact that in imagery work and image of a child may not represent a Vulnerable Child, but a Coping Child. The modes and mode processes described are directly relevant to clinical practice and, in addition to being grounded in the literature, have grown out of and proved to be of practical use in conceptualizing my own cases, and in supervising the cases of other clinicians working within the schema therapy framework.
No abstract available
No abstract available
Large Language Models (LLMs) exhibit remarkable capabilities across various downstream tasks, including empathetic dialogues. However, a non-trivial question arises: Do they possess high-context empathy and can they generate emotional interactions with humans? High-context empathy, which tends to be more indirect and concise like Chinese-style empathy, differs from the current empathy capabilities of LLMs. These capabilities are predominantly low-context empathy, which is often direct and lengthy, resembling English-style empathy. In this paper, We first construct a comprehensive Chinese High-context Empathy Dialogue dataset (HED), which consists of emotional, role-based emotional, personality-based emotional, and role-personality-based emotional dialogues. Next, we explore whether LLMs have high-context empathy in conversations. After that, we propose an innovative High-context Empathy Network (HEN) to improve LLMs' capabilities in generating high-context empathetic responses. Our empirical study demonstrates that there is much room for LLMs in generating high-context empathetic responses, and the proposed HEN can not only significantly improve LLMs' capabilities in generating high-context empathetic responses, but also has positive effects for LLMs in solving similar sentiment-related tasks.
Background Empathy, a fundamental aspect of human interaction, is characterized as the ability to experience another being’s emotions within oneself. In health care, empathy is a fundamental for health care professionals and patients’ interaction. It is a unique quality to humans that large language models (LLMs) are believed to lack. Objective We aimed to review the literature on the capacity of LLMs in demonstrating empathy. Methods We conducted a literature search on MEDLINE, Google Scholar, PsyArXiv, medRxiv, and arXiv between December 2022 and February 2024. We included English-language full-length publications that evaluated empathy in LLMs’ outputs. We excluded papers evaluating other topics related to emotional intelligence that were not specifically empathy. The included studies’ results, including the LLMs used, performance in empathy tasks, and limitations of the models, along with studies’ metadata were summarized. Results A total of 12 studies published in 2023 met the inclusion criteria. ChatGPT-3.5 (OpenAI) was evaluated in all studies, with 6 studies comparing it with other LLMs such GPT-4, LLaMA (Meta), and fine-tuned chatbots. Seven studies focused on empathy within a medical context. The studies reported LLMs to exhibit elements of empathy, including emotions recognition and emotional support in diverse contexts. Evaluation metric included automatic metrics such as Recall-Oriented Understudy for Gisting Evaluation and Bilingual Evaluation Understudy, and human subjective evaluation. Some studies compared performance on empathy with humans, while others compared between different models. In some cases, LLMs were observed to outperform humans in empathy-related tasks. For example, ChatGPT-3.5 was evaluated for its responses to patients’ questions from social media, where ChatGPT’s responses were preferred over those of humans in 78.6% of cases. Other studies used subjective readers’ assigned scores. One study reported a mean empathy score of 1.84-1.9 (scale 0-2) for their fine-tuned LLM, while a different study evaluating ChatGPT-based chatbots reported a mean human rating of 3.43 out of 4 for empathetic responses. Other evaluations were based on the level of the emotional awareness scale, which was reported to be higher for ChatGPT-3.5 than for humans. Another study evaluated ChatGPT and GPT-4 on soft-skills questions in the United States Medical Licensing Examination, where GPT-4 answered 90% of questions correctly. Limitations were noted, including repetitive use of empathic phrases, difficulty following initial instructions, overly lengthy responses, sensitivity to prompts, and overall subjective evaluation metrics influenced by the evaluator’s background. Conclusions LLMs exhibit elements of cognitive empathy, recognizing emotions and providing emotionally supportive responses in various contexts. Since social skills are an integral part of intelligence, these advancements bring LLMs closer to human-like interactions and expand their potential use in applications requiring emotional intelligence. However, there remains room for improvement in both the performance of these models and the evaluation strategies used for assessing soft skills.
No abstract available
No abstract available
Emotional empathy, the ability to understand and respond to others’ emotions, is essential for effective communication. We propose Empathy-GPT, featuring embodied conversational agents with empathic capacity. To address the limitations of rule-based conversational agents, we leverage contextual understanding and adaptation capabilities of large language models (LLMs) to coordinate multiple modalities (e.g., agent’s tone, body movements, and facial expressions). To enhance user engagement in human-agent communication, agents dynamically respond to users’ voices and facial expressions, providing contextually empathic responses.
Background: Data accessibility remains a significant barrier in healthcare AI due to privacy constraints and logistical challenges. Synthetic data, which mimics real patient information while remaining both realistic and non-identifiable, offers a promising solution. Large Language Models (LLMs) create new opportunities to generate high-fidelity clinical conversations between patients and physicians. However, the value of this synthetic data depends on careful evaluation of its realism, accuracy, and practical relevance. Objective: To assess the performance of four leading LLMs: ChatGPT 4.5, ChatGPT 4o, Claude 3.7 Sonnet, and Gemini Pro 2.5 in generating synthetic transcripts of patient–physician interactions in plastic surgery scenarios. Methods: Each model generated transcripts for ten plastic surgery scenarios. Transcripts were independently evaluated by three clinically trained raters using a seven-criterion rubric: Medical Accuracy, Realism, Persona Consistency, Fidelity, Empathy, Relevancy, and Usability. Raters were blinded to the model identity to reduce bias. Each was rated on a 5-point Likert scale, yielding 840 total evaluations. Descriptive statistics were computed, and a two-way repeated measures ANOVA was used to test for differences across models and metrics. In addition, transcripts were analyzed using automated linguistic and content-based metrics. Results: All models achieved strong performance, with mean ratings exceeding 4.5 across all criteria. Gemini 2.5 Pro received mean scores (5.00 ± 0.00) in Medical Accuracy, Realism, Persona Consistency, Relevancy, and Usability. Claude 3.7 Sonnet matched the scores in Persona Consistency and Relevancy and led in Empathy (4.96 ± 0.18). ChatGPT 4.5 also achieved perfect scores in Relevancy, with high scores in Empathy (4.93 ± 0.25) and Usability (4.96 ± 0.18). ChatGPT 4o demonstrated consistently strong but slightly lower performance across most dimensions. ANOVA revealed no statistically significant differences across models (F(3, 6) = 0.85, p = 0.52). Automated analysis showed substantial variation in transcript length, style, and content richness: Gemini 2.5 Pro generated the longest and most emotionally expressive dialogues, while ChatGPT 4o produced the shortest and most concise outputs. Conclusions: Leading LLMs can generate medically accurate, emotionally appropriate synthetic dialogues suitable for educational and research use. Despite high performance, demographic homogeneity in generated patients highlights the need for improved diversity and bias mitigation in model outputs. These findings support the cautious, context-aware integration of LLM-generated dialogues into medical training, simulation, and research.
Purpose: Empathy, a cornerstone of human interaction, is a unique quality to humans that Large Language Models (LLMs) are believed to lack. Our study aims to review the literature on the capacity of LLMs in demonstrating empathy Methods: We conducted a literature search on MEDLINE up to July 2023. Seven publications ultimately met the inclusion criteria. Results: All studies included in this review were published in 2023. All studies but one focused on ChatGPT-3.5 by OpenAI. Only one study evaluated empathy based on objective metrics, and all others used subjective human assessment. The studies reported LLMs to exhibits elements of empathy, including emotions recognition and providing emotionally supportive responses in diverse contexts, most of which were related to healthcare. In some cases, LLMs were observed to outperform humans in empathy-related tasks. Conclusion: LLMs demonstrated some aspects of empathy in variable scenarios, mainly related to healthcare. The empathy may be considered cognitive empathy. Social skills are a fundamental aspect of intelligence, thus further research is imperative to enhance these skills in AI.
No abstract available
Although large language models (LLMs) are increasingly used in emotionally sensitive contexts, it remains unclear whether their responses reflect genuine empathy or functional simulation. This study employed a mixed-methods experimental design to compare empathetic responses from human participants (n = 100) with those generated by LLMs under three conditions: text-only, anthropomorphic cue, and multimodal with cue. Using ten standardized emotional vignettes, 200 independent raters evaluated all responses with the Consultation and Relational Empathy (CARE) measure. Quantitative results showed that LLM-generated responses were consistently rated as more empathetic than human responses, with empathy scores increasing across enhanced design conditions. Qualitative analysis using the Empathic Communication Coding System (ECCS) revealed that LLMs relied heavily on surface validation and supportive strategies, while demonstrating limited contextual probing and a lack of experiential grounding compared to human responses. These findings highlight a critical distinction between perceived empathy and phenomenological empathy. While LLMs can effectively simulate empathic communication and achieve high ratings due to consistency and alignment with social expectations, they lack the contextual depth and experiential understanding that characterize human empathy. This study contributes to the literature by moving beyond performance-based evaluations and clarifying the structural differences between human and artificial empathy. The results have important implications for the use of LLMs in healthcare, education, and counseling, where perceived empathy may enhance user experience but also raises concerns about over-trust and anthropomorphic attribution. Disclosure Statement The authors reported no potential conflicts of interest.
As artificial intelligence (AI) systems become increasingly embedded in everyday human interaction, a new frontier – “emotional AI or affective computing” that is transforming how machines perceive and respond to human emotions. This systematic review synthesizes recent empirical evidence (2015-2025) examining the behavioral, psychological and societal impacts of emotion-recognizing AI in human trust, empathy and social interaction. Using the PRISMA methodology, the review analyses 50 empirical studies across domains including healthcare, education, customer service and human–machine collaboration. The results reveal a dual pathway: while emotional AI can foster improved human–machine cooperation and emotional well-being through adaptive empathy simulation, it simultaneously poses risks of privacy violations, cultural bias and emotional manipulation. Drawing on frameworks from social cognition and behavioral trust theory, we show that human trust in machine-mediated empathy is shaped more by perceived sincerity and relational cues than by raw algorithmic accuracy. The review concludes with ethical and practical recommendations.
Simulated empathy is valuable for conversational agents and robots in emotionally sensitive domains such as counseling, healthcare, and customer service. However, the appropriateness of different types of empathy for different task domains is under-explored. We describe the results from a study comparing three approaches to the automatic generation of empathic responses for a conversational robot that conducts a depression screening conversation with a user. The computational models of empathy incorporate both rule-driven, theory-based responses with data-driven responses from a transformer model. Our primary finding i s t hat a sour empathic response models increase in sophistication their empathic accuracy decreases, resulting in more erroneous responses. Empathic inaccuracy, in turn, led to significantly lower ratings o f working a lliance, a measure of user trust in and liking of the robot.
Affective computing improves rapidly, allowing systems to process human emotions. This enables systems such as conversational agents or social robots to show empathy toward users. While there are various established methods to measure the empathy of humans, there is no reliable and validated instrument to quantify the perceived empathy of interactive systems. Thus, we developed the Perceived Empathy of Technology Scale (PETS) to assess and compare how empathic users perceive technology. We followed a standardized multi-phase process of developing and validating scales. In total, we invited 30 experts for item generation, 324 participants for item selection, and 396 additional participants for scale validation. We developed our scale using 22 scenarios with opposing empathy levels, ensuring the scale is universally applicable. This resulted in the PETS, a 10-item, 2-factor scale. The PETS allows designers and researchers to evaluate and compare the perceived empathy of interactive systems rapidly.
As artificial intelligence (AI) is increasingly integrated into counselor education, a significant gap exists between the rapid adoption of technology and the development of structured pedagogical frameworks for ethical decision-making. This article introduces a multi-component framework to support ethical AI integration in graduate counseling education, comprising three key elements: (1) a case conceptualization model with AI assistance, (2) Morgan (SC), an intentionally imperfect AI school counseling thought partner trained on ethical standards and decision-making models, and (3) structured protocols for ethical decision-making aligned with American Counseling Association and National Board for Certified Counselors guidance. The framework is designed to center human judgment, cultural responsiveness, and professional values, leveraging AI to prompt reflection, generate alternative perspectives, and highlight ethical and contextual considerations. The framework is conceptually situated within a graduate-level internship course and informs instructional design and reflective learning practices. This article presents a practice-informed model for ethical AI integration.
No abstract available
Verbal harassment is a growing source of psychological stress for people around the world. It occurs both online and offline and relies on language to demean, threaten, or discredit its targets. Unlike other stressors such as loss or uncertainty, verbal harassment aims at silencing its targets by eroding their sense of being heard and weakening their perceived ability to respond. Many individuals lack access to adequate and timely support, however, when they experience such harassment. People increasingly turn to conversational artificial intelligence (AI) such as ChatGPT or dedicated AI companions for emotional support, raising questions about whether it can facilitate the same psychological benefits as actual human empathy. We focus on online contexts as a prevalent application of verbal harassment. We develop and test a psychological framework identifying three key linguistic signals of empathic listening (perspective-taking, emotional validation, and action orientation), that together restore a sense of feeling heard and enhance coping in the context of verbal harassment. We find that LLMs consistently produce language exhibiting stronger empathic-listening markers than human non-experts and trained mental health professionals, promoting more approach-oriented (vs. avoidance-oriented) coping strategies. A subsequent behavioral study shows that these linguistic signals boost recipients'sense of feeling heard and increase their coping self-efficacy. These findings reveal how specific linguistic features create empathic connections between humans and advanced conversational AI and can enhance people's psychological resilience. Our results highlight the potential for AI to serve as a scalable source of emotional support, especially when human support is unavailable or insufficient.
Artificial intelligence systems increasingly generate text intended to provide social and emotional support. Understanding how users perceive empathic qualities in such content is therefore critical. We examined differences in perceived empathy signals between human-written and large language model (LLM)-generated relationship advice, and the influence of authorship labels. Across two preregistered experiments (Study 1: n = 641; Study 2: n = 500), participants rated advice texts on overall quality and perceived cognitive, emotional, and motivational empathy. Multilevel models accounted for the nested rating structure. LLM-generated advice was consistently perceived as higher in overall quality, cognitive empathy, and motivational empathy. Evidence for a widely reported negativity bias toward AI-labelled content was limited. Emotional empathy showed no consistent source advantage. Individual differences in AI attitudes modestly influenced judgments but did not alter the overall pattern. These findings suggest that perceptions of empathic communication are primarily driven by linguistic features rather than authorship beliefs, with implications for the design of AI-mediated support systems.
The fields of human-robot interaction (HRI) and embodied conversational agents (ECAs) have long studied how empathy could be implemented in machines. One of the major drivers has been the goal of giving multimodal social and emotional intelligence to these artificially intelligent agents, which interact with people through facial expressions, body, gesture, and speech. What empathic behaviors and models have these fields implemented by mimicking human and animal behavior? In what ways have they explored creating machine-specific analogies? This chapter aims to review the knowledge from these studies, towards applying the lessons learned to today's ubiquitous, language-based agents such as ChatGPT.
As artificial intelligence (AI) systems become increasingly embedded in everyday life, the ability of interactive agents to express empathy has become critical for effective human-AI interaction, particularly in emotionally sensitive contexts. Rather than treating empathy as a binary capability, this study examines how different levels of empathic expression in virtual human interaction influence user experience. We conducted a between-subject experiment (n = 70) in a counseling-style interaction context, comparing three virtual human conditions: a neutral dialogue-based agent, a dialogue-based empathic agent, and a video-based empathic agent that incorporates users'facial cues. Participants engaged in a 15-minute interaction and subsequently evaluated their experience using subjective measures of empathy and interaction quality. Results from analysis of variance (ANOVA) revealed significant differences across conditions in affective empathy, perceived naturalness of facial movement, and appropriateness of facial expression. The video-based empathic expression condition elicited significantly higher affective empathy than the neutral baseline (p<.001) and marginally higher levels than the dialogue-based condition (p<.10). In contrast, cognitive empathy did not differ significantly across conditions. These findings indicate that empathic expression in virtual humans should be conceptualized as a graded design variable, rather than a binary capability, with visually grounded cues playing a decisive role in shaping affective user experience.
We present JoyAI-Talker, a full-duplex speech dialogue system that delivers robust foundation model capabilities while empowering empathetic interaction and voice agent intelligence. JoyAI-Talker adopts a modular Thinker-Talker architecture and further implements a unified speech-text joint training pipeline to mitigate the common"cognitive degradation"bottleneck, thereby largely preserving the model's core textual reasoning, STEM, and logical capabilities while extending them to speech-based interaction. For expressive speech synthesis, the Talker module employs a text-controllable generation paradigm that enables natural-language instructions to flexibly control vocal attributes and localized paralinguistic events, such as laughter and sighs, supporting more expressive and fine-grained speech responses. To enhance conversational empathy, we introduce the Persona-Adaptive Empathetic Response (PAER) framework. PAER employs a hierarchical cognitive pipeline to extract non-verbal speaker cues, such as gender, age, and emotional state, from raw input audio, incorporate them into the Thinker's CoT reasoning, and generate context-adaptive responses that align semantically appropriate text with fine-grained control over utterance-level expressiveness and localized paralinguistic events, including sighs, speaking rate, and volume. We further integrate Joy-Duplex, a state-driven, plug-and-play full-duplex framework that functions as an efficient gating engine for real-time turn control. Extensive evaluations show that JoyAI-Talker achieves highly competitive performance on foundational T2T and S2T benchmarks. In full-duplex evaluation, the system reaches a high response rate of 0.88 under user interruptions while maintaining an extremely low false-trigger rate under background speech, demonstrating its readiness for fluid and natural speech dialogue.
Using large language models (LLMs) to assist psychological counseling is an important task in the field of natural language processing. The construction of high-quality psychological support dialogue corpora serves as a critical foundation for training counseling-oriented conversational models. However, existing data generation approaches generally suffer from several limitations, including emotionally stable seekers, limited variation in emotional dynamics, and a high degree of compliance with counselors'guidance. These issues result in LLM that lack the capability to effectively respond to emotionally unstable scenarios. In addition, counselor responses are typically driven by problem-solving objectives, thereby overlooking the role of emotion-focused interaction, which are essential in psychological counseling. To address these gaps, we propose EmoTrace, a multi-turn dialogue corpus generation framework centered on modeling seekers'emotional trajectories. we construct seekers'cognitive profile and introduce a seeker module with emotional schemas and an associated activation mechanism, a counselor module, and an emotional trajectory control module, thereby enhancing the layering of the seeker's emotional expression and the counselor's targeted empathic expression. Experimental results demonstrate that the proposed method outperforms existing approaches in terms of emotional richness and empathy quality.