• 学术搜索
  • 科研智能体
    • Research Labs
    • AI 阅读
    • AI 文库
    • 深度研究
    • 学者亮点
  • 学术资源
    • AI2000
    • 期刊/会议
    • 学者库
    • 学术API
      MCP&Skill
    • 溯源树
    • 数据集
  • 知识沉淀
    • 学术空间
订阅小程序
旧版功能
aminer vip
开通会员低至0.73元/天
一次搞定AI科研
立即登录
  • English
  • 联系方式
    Computers in Human Behavior: Artificial Humans

    Computers in Human Behavior: Artificial Humans

    JournalISSN 2949-8821eISSN 2949-8821

    年发文量

    研究主题

    论文(385)

    排序
    1Robot Social Anxiety As a Distinct Psychological Construct: Relations to Human Social Anxiety and Development of the Multidimensional Social Robot Anxiety Scale
    Te-Chi Huang,Ritsuko Iwai,Takatsune Kumada

    Anxiety toward social robots is becoming increasingly relevant as they enter everyday environments; however, the psychological structure of this anxiety and its relation to interpersonal social anxiety remain unclear. This study developed and validated the Multidimensional Social Robot Anxiety Scale (MSRAS) and examined whether robot social anxiety (RSA) constituted a distinct psychological construct via two large-scale surveys (N1 = 1957; N2 = 2393). Based on Dennett's framework, factor-analytic results identified four dimensions: Physical Danger, Social Cognitive Ability, Information Security, and Social Anxiety toward Robots. These were supported by a bifactor structure in which RSA retained substantial unique variance, indicating that it was conceptually and empirically distinct from other robot-related anxieties. Furthermore, structural equation modeling revealed that general human social anxiety significantly predicted RSA, suggesting that RSA and human social anxiety (HSA) share certain psychological characteristics. However, RSA diverged from HSA in its socio-emotional correlates: it had markedly weaker associations with self-perceived social skills and perceived social support. The association between RSA and need to belong was weaker than that with social phobia but comparable to that with social interaction anxiety, providing only partial support for the predicted attenuation of belongingness-related concerns. These findings suggest that RSA and HSA share certain characteristics while exhibiting systematic differences, potentially reflecting the perceived limitations of robots' social-evaluative and emotional capacities. Furthermore, the MSRAS is a psychometrically robust tool for examining individual differences in robot-related anxiety and advancing theoretical models of anxiety in human–robot interactions.

    2026
    引用
    AI阅读
    加入学术空间
    2From Virtual Pets to AI Companions: A Relational-Design Framework for Human–Artificial Bonds
    Fernando Coura Assis

    Artificial companions occupy a hybrid position between interactive systems and social partners, yet research remains divided across virtual pets, social robots, relational agents, mixed-reality animals, and generative AI companions. This critical integrative review asks which design properties can make sustained human–artificial relationships possible without assuming that relational significance depends on human likeness or technical sophistication. It develops a five-dimensional relational-design framework: temporal continuity; legible need and vulnerability; contingent responsiveness; social presence and anthropomorphic interpretation; and recurrent care and support. A central revision to the framework is that care is not necessarily directed from the human toward the artificial entity. In many relational-agent, health, educational, and companionship contexts, the human is the recipient of support, and the system must detect, interpret, and respond to human needs. The framework therefore treats care as directional and potentially reciprocal, while distinguishing companion relationships from training simulations in which a virtual patient or student is primarily an object of professional practice. Drawing on research in virtual pets, social responses to computers, relational agents, human–robot interaction, digital possessions, and AI companions, the review connects each dimension to identifiable literature and illustrates its operation across different system types. It also examines well-being, asymmetrical power, vulnerable populations, loss, and platform governance. The framework is a conceptual tool rather than a validated measurement instrument; its value lies in generating comparative hypotheses and design questions that can be tested longitudinally and across contexts.

    2026
    引用
    AI阅读
    加入学术空间
    3Enhancing AI Explainability and Transparency Through the Integration of Explainable AI and Generative AI
    Ching-Hua Chuan,Wan-Hsiu Sunny Tsai, Jeongwon Yang

    This study examines how generative AI (GenAI)-mediated explanations of automated decision-making (ADM) shape users’ justice perceptions. It investigates how GenAI advances conversational XAI by enabling more open-ended and dynamically generated explanatory exchanges and examines their implications for the ethical evaluation of AI decisions beyond traditional usability and explainability outcomes. The study focuses on AI-driven loan evaluation contexts, where concerns about fairness, transparency, and accountability are especially salient. It foregrounds the perspective of imposed users, individuals affected by AI decisions without their volition. A between-subjects online experiment was conducted with three conditions: no explanation, traditional static explainable AI (XAI), and interactive explanations powered by generative AI (GenAI). Grounded in Organizational Justice Theory and the Human–AI Interaction Model, the findings demonstrate that GenAI-powered interactive XAI (GenXAI) enhances users’ sense of agency, message contingency, and sensemaking processes, which were associated with their justice perceptions. Sensemaking was the most consistent antecedent across all four justice dimensions, whereas sense of agency was associated only with procedural and informational justice and message contingency with distributive, informational, and interpersonal justice. GenXAI also showed a negative direct association with distributive justice alongside a positive indirect association through sensemaking. By illuminating the psychological mechanisms through which explanations about AI decisions shape ethical evaluations of AI systems, this study shows that explanation is not merely about providing information, but an inherently interactive and communicative process that can empower users’ sensemaking of AI decisions.

    2026
    引用
    AI阅读
    加入学术空间
    4Programming Task Performance and Perceived Cognitive Engagement Across Three LLM Support Configurations
    Michail Giannakos, Henrik Giil Liisberg, Martin Kvalsund, Chaofan Wang,Evangelos Niforatos

    Large language models (LLMs) can assist students with programming tasks, but performance achieved with assistance does not establish independent learning. This exploratory mixed-methods study compared three configurations: a programming environment without LLM assistance, a code-feedback assistant intended to limit direct solutions, and unrestricted LLM chat. Forty-five university students with prior introductory programming experience were randomly assigned to one configuration to complete a Python task. Data comprised submitted code, interaction measures, questionnaires, and interviews with 15 students from the two LLM conditions. Unrestricted chat yielded higher task scores than both restricted feedback and the no-LLM environment. Both LLM configurations received higher usefulness ratings than the baseline environment. Interview accounts described differences in perceived cognitive engagement, including working through hints and relying on generated solutions, alongside difficulties using the restricted interface. Because the configurations also differed in interface, conversational flexibility, and contextual information, their differences cannot be attributed solely to answer restrictions. The findings provide context-specific evidence about assisted task performance and reported experience.

    2026
    引用
    AI阅读
    加入学术空间
    5Finding Romance with AI: Comparing Human and AI Chatbot Intimacy and Relationships
    Annette M. Masterson,Xin Ye,Lionel P. Robert

    The expanding role of artificial intelligence (AI) in interpersonal communication challenges traditional understandings of intimacy. This study rigorously compares how distinct dimensions of intimacy (emotional, intellectual, sexual, and social) are associated with loneliness, relationship satisfaction, and well-being across AI chatbot, long-distance human, and in-person human relationships. Drawing on a survey of 463 participants (155 AI companion users, 153 in long-distance human relationships, and 155 in in-person human relationships), we found comparable levels of intimacy across all conditions, illustrating strong relationships both between humans and between humans and chatbots. In particular, emotional, intellectual, sexual, and social intimacy are all significantly present within chatbot relationships. We also uncover a complex and paradoxical landscape. Emotional intimacy in AI companion relationships correlated positively with loneliness, despite positively contributing to relationship satisfaction—a stark contrast to its negative association with loneliness in human relationships. However, the association between loneliness and emotional intimacy is no longer significant when social support is accounted for. Intellectual intimacy emerged as a stronger predictor of well-being and satisfaction in chatbot and long-distance relationships than in in-person relationships, suggesting a unique value for cognitive companionship. Recreational intimacy proved non-transferable to AI contexts. These findings indicate that while AI companion relationships can mirror the structure of human intimacy, providing satisfaction and intellectual engagement, they may not fully replicate their function, correlating with lower loneliness or providing holistic social connection. This research advances the literature by identifying the boundaries of AI intimacy and highlighting the ethical imperative for responsible AI design.

    2026
    引用
    AI阅读
    加入学术空间
    立即登录,查看全部 385 篇论文

    高被引作者

    作者引用发文
    Lennart Nacke1612
    Reza Hadi Mogavi1612
    Fatma Gizem Karaoglan Yılmaz1593
    Ramazan Yılmaz1491
    Chao Deng1431
    Sujit Gujar1431
    Peng Yuan Zhou1431
    Simone Bassanelli1431
    Pan Hui1431
    Ahmed Tlili1431

    高产作者

    作者引用发文
    Tenzin Doleck1025
    Matthieu J. Guitton35
    Andree Hartanto255
    Laura M Vowels374
    Markus Langer144
    Markus Huff54
    Nicolas Spatola174
    Zohar Elyoseph24
    Birgit Lugrin03
    Nadia Said43

    相关期刊

    暂无数据
    暂无数据