Affective computing scholarship and the rise of China: a view from 25 years of bibliometric data

Artificial intelligence (AI) is rapidly transforming various aspects of life, with affective computing, or emotional AI, playing a significant role. This field combines computer science, engineering, psychology, and neuroscience to model and analyze human emotions. While the US initially dominated affective computing research, China is now a major contributor. This study investigates the global landscape of affective computing research over the past 25 years, focusing on national contributions, international collaborations, and thematic evolution. The study challenges the narrative of a new Cold War between the US and China in AI research, highlighting the significant collaborations that exist despite political tensions and ethical concerns surrounding data privacy and algorithmic bias.

The literature review implicitly examines existing research on affective computing, referencing works by Rosalind Picard (who coined the term), and highlighting the commercial applications of the technology across various sectors, including education, entertainment, and the workplace. It also explores the ethical implications and potential negative consequences of emotion-sensing AI, such as increased technostress, bias, and lack of data privacy regulations, particularly in non-Western contexts. Existing literature on algorithmic management, data privacy legislation (e.g., GDPR), and the impact of cultural norms on data collection are reviewed.

This study employed bibliometric analysis using the Web of Science database. The search focused on English-language publications containing the keyword 'affective computing' from 1995 to 2020. A total of 3448 articles were included in the analysis after removing duplicates and 2021 publications. The open-source bibliometrix package in R was used to conduct the analysis, examining publication trends, major contributors (countries and authors), collaboration networks, and thematic evolution. The analysis included descriptive statistics, co-citation analysis, and co-word analysis to identify key themes and their evolution over time.

The annual growth rate of publications in affective computing is 11.36%, indicating exponential growth. The US initially dominated the field (1995-2005), but China has significantly increased its output and citation impact in recent years (2016-2020). Two major collaborative networks emerged: a US/Asia-Pacific cluster (US, China, Singapore, Japan, India) and a European cluster (Germany, UK, Netherlands). A significant shift in research focus is observed, moving away from mental illness diagnosis towards commercially driven applications such as smart city design and workplace management. The ensemble method, combining symbolic and sub-symbolic AI, is a prominent technique. Notable absences include Russia, due to geopolitical factors, economic challenges, and language barriers, though recent private sector initiatives show promise. Thematic analysis shows a transition from mental health applications (1995-2005) to sentiment analysis (2006-2015) and finally, to more commercially-focused applications (2016-2020), with an increasing use of multimodal techniques.

The findings challenge the notion of a technological Cold War between the US and China. The substantial collaboration between researchers from both countries demonstrates ongoing knowledge exchange. The shift in research focus reflects the commercial viability of affective computing, but also raises concerns regarding ethical implications and potential biases. The lack of strong data privacy regulations in some regions exacerbates these concerns. The absence of Russia highlights the impact of political and economic factors on research output. Future research should focus on cross-cultural differences in emotion expression, the development of more robust and bias-free algorithms, and the ethical implications of increasingly pervasive emotion-sensing technologies.

Affective computing exhibits exponential growth, with China emerging as a leading force. International collaborations are crucial, challenging geopolitical narratives. A shift from mental health applications towards commercial applications is evident. Future work needs to address ethical concerns and cross-cultural nuances in emotion recognition.

The study is limited to English-language publications indexed in Web of Science, potentially overlooking relevant research in other languages. The focus on the 'affective computing' keyword might exclude studies using alternative terminology. The bibliometric analysis relies on publication counts and citations, which may not fully capture the quality or impact of individual research projects.