Xinheng Song
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xinheng Song has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Bingjie Gao
Tsinghua University, The Future Laboratory, Beijing, China
4 shared publications - Zhaolin Lu
School of Design and Arts, Beijing Institute of Technology, No. 5, South Street, Zhongguancun, Haidian District, Beijing 100081, China
4 shared publications Yue Zhang
School of Architecture and Arts, Hefei University of Technology
3 shared publicationsZhaolin Lu
Industrial Manufacturing Art Innovation Design Key Laboratory of Ministry of Industry and Information Technology, Beijing Institute of Technology
3 shared publications- Haipeng Mi
Tsinghua University, Beijing, China
2 shared publications - Haiwen Song
Beijing Institude of Technology, Beijing, China
2 shared publications
Latest Publications
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Beyond Prompt-Following: Empowering Social Agency in Autistic Children through Narrative Framing
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Open AccessBeyond Prompt-Following: Empowering Social Agency in Autistic Children through Narrative Framing
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
GenAI EmoJellyfish: Investigating the Synergistic Effects of Multi-modal Interaction on Emotion Regulation
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessGenAI EmoJellyfish: Investigating the Synergistic Effects of Multi-modal Interaction on Emotion Regulation
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Fix or Fake? How Creators Negotiate Cultural Bias in Generative AI Heritage Creation
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccessFix or Fake? How Creators Negotiate Cultural Bias in Generative AI Heritage Creation
- Top Cited Papers
- 19 January 2026
- Published by Elsevier in Computers & Education
- Vol. 247
Generative AI: A double-edged sword for creative thinking learning — Evidence from facial expressions and fNIRS
- 19 January 2026
- Published by Elsevier in Computers & Education
- Vol. 247
Top Cited PapersGenerative AI: A double-edged sword for creative thinking learning — Evidence from facial expressions and fNIRS
- 15 December 2025
- Published by SAGE Publications in Information Development
The symbiosis of efficient information supply and humanistic care in the AI-driven library service systems
As library information systems become increasingly intelligent, they generate vast amounts of information. This study identifies the key factors shaping user experience in such systems. Using grounded theory, we extracted variables from interviews (N = 14) and developed a framework based on the Stimulus–Organism–Response model. A survey (N = 320) tested the hypothesized relationships through structural equation modelling, while artificial neural networks captured non-linear effects and ranked factor importance. Results show that information validity is the strongest predictor of interaction fluency. While aesthetic interface design improves Interaction Fluency, it does not significantly affect perceptions of System Support Level. These findings highlight the need to balance efficient information provision with human-centred design. Overemphasis on functionality risks diminishing engagement. Future designs should prioritize information validity for guidance tasks while also enhancing cultural and emotional experiences in both human and AI-driven library services.
- 15 December 2025
- Published by SAGE Publications in Information Development
The symbiosis of efficient information supply and humanistic care in the AI-driven library service systems
As library information systems become increasingly intelligent, they generate vast amounts of information. This study identifies the key factors shaping user experience in such systems. Using grounded theory, we extracted variables from interviews (N = 14) and developed a framework based on the Stimulus–Organism–Response model. A survey (N = 320) tested the hypothesized relationships through structural equation modelling, while artificial neural networks captured non-linear effects and ranked factor importance. Results show that information validity is the strongest predictor of interaction fluency. While aesthetic interface design improves Interaction Fluency, it does not significantly affect perceptions of System Support Level. These findings highlight the need to balance efficient information provision with human-centred design. Overemphasis on functionality risks diminishing engagement. Future designs should prioritize information validity for guidance tasks while also enhancing cultural and emotional experiences in both human and AI-driven library services.
- 4 August 2025
- Published by Emerald Publishing in Aslib Journal of Information Management
- p. 1-25
Affective computing methods for multimodal embodied AI human–computer interaction
Purpose: In the context of rapid innovation in computing technology, the Embodied Artificial Intelligence (EAI) design methods and key research issues continuously evolve. Affective computing is now becoming a core means to assits in EAI design research, enriching the emotional connotations of EAI design and promoting more natural and harmonious human–computer interaction. Design/methodology/approach: This paper adopted a systematic literature review approach and ultimately selected 36 high-quality papers related to affective computing methods in multimodal EAI human–computer interaction. The analysis is conducted on the temporal migration of research topics and the distribution of four aspects: research content, application areas, data collection form and algorithm. Findings: Three key steps were outlined in the study of affective computing methods in multimodal EAI human–computer interaction: Emotional recognition in human–computer interaction, evolutionary learning of user characteristic emotional classification, emotional analysis and behavioral feedback. Practical implications: The paper prospects the application scenarios of affective computing assistance in embodied intelligent human–computer interaction, deriving trends in enhancing emotional experience in medical service design, providing simulation training for dangerous engineering environments, supplying social design with safety perception and mitigating the socio-ethical challenges raised by the new technology. Social implications: Four challenges in the emotional data collection process were summarized: privacy protection issues, questions about the accuracy and credibility of generative design, embodiment interaction dilemmas in cross-cultural contexts and social-ethical issues surrounding affective computing assistance in EAI. Originality/value: Previous studies mainly focused on the algorithmic effects of EAI and the design of a single modality experience. This paper offers an in-depth analysis from a multimodal viewpoint, following the research framework of affective computing. It presents a thorough examination of the human-centered interaction experience facilitated by embodied intelligent design.
- 4 August 2025
- Published by Emerald Publishing in Aslib Journal of Information Management
- p. 1-25
Affective computing methods for multimodal embodied AI human–computer interaction
Purpose: In the context of rapid innovation in computing technology, the Embodied Artificial Intelligence (EAI) design methods and key research issues continuously evolve. Affective computing is now becoming a core means to assits in EAI design research, enriching the emotional connotations of EAI design and promoting more natural and harmonious human–computer interaction. Design/methodology/approach: This paper adopted a systematic literature review approach and ultimately selected 36 high-quality papers related to affective computing methods in multimodal EAI human–computer interaction. The analysis is conducted on the temporal migration of research topics and the distribution of four aspects: research content, application areas, data collection form and algorithm. Findings: Three key steps were outlined in the study of affective computing methods in multimodal EAI human–computer interaction: Emotional recognition in human–computer interaction, evolutionary learning of user characteristic emotional classification, emotional analysis and behavioral feedback. Practical implications: The paper prospects the application scenarios of affective computing assistance in embodied intelligent human–computer interaction, deriving trends in enhancing emotional experience in medical service design, providing simulation training for dangerous engineering environments, supplying social design with safety perception and mitigating the socio-ethical challenges raised by the new technology. Social implications: Four challenges in the emotional data collection process were summarized: privacy protection issues, questions about the accuracy and credibility of generative design, embodiment interaction dilemmas in cross-cultural contexts and social-ethical issues surrounding affective computing assistance in EAI. Originality/value: Previous studies mainly focused on the algorithmic effects of EAI and the design of a single modality experience. This paper offers an in-depth analysis from a multimodal viewpoint, following the research framework of affective computing. It presents a thorough examination of the human-centered interaction experience facilitated by embodied intelligent design.
- Open Access
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Mozualization: Crafting Music and Visual Representation with Multimodal AI
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessMozualization: Crafting Music and Visual Representation with Multimodal AI
- Open Access
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
“Do You Need the Sage's Tea or the Friend's Cola” Exploring the Differential Healing Effects of Generative AI Conversational Styles
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open Access“Do You Need the Sage's Tea or the Friend's Cola” Exploring the Differential Healing Effects of Generative AI Conversational Styles
- Open Access
- 11 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Chatbot and Fatigued Driver: Exploring the Use of LLM-Based Voice Assistants for Driving Fatigue
This study explores the application of Large Language Model (LLM)-based voice assistants, such as ChatGPT-4, in mitigating passive driving fatigue, while aiming to enhance driving performance and safety. Employing an empirical approach with a driving simulator and the LLM-based voice assistant "Driver Mate," the study focuses on conversation complexity and frequency as independent variables, and utilizes electroencephalogram (EEG), driving simulator data, and scales for measures. The findings reveal the effectiveness of in-vehicle LLM-based voice assistants, highlighting that low-complexity, high-frequency conversation is optimal for driver alertness and acceptance, while low-complexity, low-frequency interactions significantly improve driving performance. This study innovatively investigates the role of LLM-based voice assistants in alleviating driving fatigue, offering practical suggestions for future in-vehicle systems.
- 11 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessChatbot and Fatigued Driver: Exploring the Use of LLM-Based Voice Assistants for Driving Fatigue
This study explores the application of Large Language Model (LLM)-based voice assistants, such as ChatGPT-4, in mitigating passive driving fatigue, while aiming to enhance driving performance and safety. Employing an empirical approach with a driving simulator and the LLM-based voice assistant "Driver Mate," the study focuses on conversation complexity and frequency as independent variables, and utilizes electroencephalogram (EEG), driving simulator data, and scales for measures. The findings reveal the effectiveness of in-vehicle LLM-based voice assistants, highlighting that low-complexity, high-frequency conversation is optimal for driver alertness and acceptance, while low-complexity, low-frequency interactions significantly improve driving performance. This study innovatively investigates the role of LLM-based voice assistants in alleviating driving fatigue, offering practical suggestions for future in-vehicle systems.