THE PROJECT RATIONALE
The project aims at developing a research agenda and methods of research and support for struggling with depression as manifested via conversational and behavioral practices in mediated realms. Today, AI contributes depression detection and prevention in two crucial aspects: 1) analytical AI instruments help in detecting depression markers in mediated communication, both by already-diagnosed patients and people who experience symptoms but lack guidance in how to address their potential disease properly; 2) generative AI instruments are helpful in organizing support via generation of proper conversational and visual guidance, as well as answering typical questions on the disease for relatives (e.g., parents or spouses) of a potential patient.
The first-stage detection of many mental illnesses, including depression, is aggravated by their popular misunderstanding. The ‘social distortion’ of depression is twofold: First, the symptoms of clinical depression are not taken seriously by the prospective patients and their relatives until the disease takes a severe form; second, on the contrary, many mood manifestations that are, indeed, not related to clinical depression (e.g., in adolescents) are called ‘depressive’ without proper grounds. Both distortions may lead to additional burdens to the healthcare systems, as severe forms of disease are treated longer and with lower success rates, while multiple non-grounded addresses of medical personnel by patients distract efforts from the patients in real need. In Russia and China, though, the social cultures prevent the latter (as mental diseases may be considered shameful) but lead to an even bigger distortion, which is uneducated self-treatment. As a result, depression on its early stages rarely gets curated properly, despite significant efforts invested by the state healthcare systems into enlightening of populace and promoting correct depression-related behaviors by patients and their family members.
One of the windows of opportunities here is created by AI. First, today, a large part of social communication is mediated and conducted via social media. As the current Western research shows, the fixed nature and peculiar shape of mediated communication may help in better tracing individual shifting towards depression, as well as its clinical manifestations. However, this research is practically absent in non-Latin-language countries; also, it often stops at the stage of finding markers of depression, which does not convert into prevention and support practices. Meanwhile, generative AI becomes more capable of sophisticated individualized guidance of early-stage enquiries on mental diseases; creation of genAI-based help&support services would allow for instruments available for mass groups of people in terms of guidance on proper behavior, including family relations, addressing medical personnel, and patient support during the disease. Thus, uniting analytical and generative AI instruments for, respectively, detection and guidance of patients may create easier, non-socially-bullied paths to proper early detection and treatment of depression.
Therefore, the Russian-Chinese working group aims at employing analytical and generative AI in creating a detection-to-support chain for early stages of depression. This demands:
-detailed (seminar- and/or round-table based) discussion on a cross-country research agenda in socially-mediated depression detection-to-support, which would involve experts in psychology, media research, AI development, and beyond (if necessary);
-creation of the road map for the elaboration of a detection-to-support AI-based system;
-testing available analytical AI instruments for Chinese and Russian, including development and/or fine-tuning of AI models, creating test datasets, and elaboration on quality assessment criteria for each of the two languages;
-creation of a joint application to Russian Science Foundation/National Science Foundation of China and/or alternative research foundations.
METHODS AND APPROACHES
The general design of the 'big' project (2025-2027) goes from fundamental research to practical applications. Thus, the methods that are planned to be employed include:
- 'opinion tree' pipeline for mapping user discussions based on a combination of three analytical AI methods of textual analysis, including BERT-based and Transformer-based transfer learning methods, as tuned for behavior detection;
- abstractive summarization of user texts;
- AI-assisted content analysis;
- generative AI-based recommender prototype for fine-tuning, as based upon the results of analytical research (textual analysis and/or focus groups with users);
- quality evaluation methods against pre-set academic and operational objectives.
The 2024 project will be focusing on testing the available methods/AI instruments for Chinese and Russian, including the abovementioned BERT- and Transformer-based architectures. Other activities are focused on literature reviewing, organization of discussion, and interaction with scholars outside the working group.
PREVIOUS RESEARCH AND OTHER PROJECT PREMISES
Project sustainability is ensured by the following capacities and results demonstrated earlier.
1. The research profiles and long-term research and academic leadership experience of the Principal Investigators (including national-level grant projects in both countries), with 100+ Scopus/WoS-indexed papers altogether, most of them directly fitting into the project theme. The number of projects led by Profs Zhang and Bodrunova is 10+ altogether, including major projects funded by Russian Science Foundation and National Natural Science Foundation of China.
2.The interdisciplinary constellation of the research groups from both sides, including experts in communication, social psychology, artificial intelligence, and social media data collection. Thus, the presented research groups include two department heads, well-known experts in the field, as well as already-awarded young scholars from both universities.
3.The positions of the Principal Investigators as editorial board members and special issue editors in major Q1/Q2 journals in the project-related fields of science. Thus, Wei Zhang is an editorial board member at Frontiers in Public Health, Frontiers in Psychiatry, and Frontiers in Sociology. Svetlana Bodrunova is a member of boards at World of Media, Global Media and Communication, Central European Journal of Communication, and Future Internet.
4. Availability of research facilities, including those at the SPbU Scientific Park. Moreover, the research group possesses patented instruments (9 patents altogether) for the research methods stated above, as well as unique and well-tested software for data collection and analysis already proven in previous research.
RESULTS AND IMPACT
Academic and social impact:
I.This collaboration corresponds with national and university strategic priorities on both sides. In particular, HUST’s mission is to conduct research on people’s life and health and healthy China promotion; one of its priorities till 2030 is the interdisciplinary development of Medicine, AI and big data analytics. For SPbU, the respective priorities are 2030-1.1 ‘Digital technologies, artificial intelligence, new materials’ and 2030-1.2 ‘Personalized and high-tech healthcare, genetics, and pharmacology.’
II.The project will start linking early detection of depression with publicly available social media data to the necessary popular knowledge of depression, support of queries, and long-standing assistance during the disease. This may critically enhance the struggle with depression in both countries, breaking the ‘glass wall’ between AI-based detection of depression manifestations and actual guidance and help to affected people.
III.Developing a kick-off joint project will surely help in integrating research agendas in the designated area in China and Russia, thus making joint efforts much more efficient and independent from Western research, while efficiently using their previous findings.
IV.The joint project will integrate students, including the PhD students and early-career scholars, from at least three thematic areas (medicine/psychology, computer science/AI, and communication), into an interdisciplinary real-world project.
V.If successful on next stages (RSF/NSFC application), the project promises creation of safer communication environments, easier pathways to treatment, and more sophisticated support of the affected people, thus having a long-term impact upon the levels of early detection of depression manifestations.
VI.In addition, the collaboration will helps boost HUST’s and SPbU’s international reputation and academic impact as expert centres for depression detection, prevention, and support.
Conceptual and methodological output:
I.A road map for establishment of Russian and Chinese online services for depression detection-to-support.
II.Fine-tuned analytical AI models for Russian and Chinese created for detection of depression markers and/or conversational patterns telling of depression. This may include instruments developed/fine-tuned in both universities and/or modified algorithms on well-known computing languages and in well-developed packages (Python, R, or others).
Academic output:
III.Submission and receiving an R&R status for two joint publications in international journals/book series, not lower than Q2 (both aimed at Q1).
IV.Submission of a joint NSFC/RSF application for 3 years.
V.At least two interdisciplinary workshops conducted, with their reports becoming parts / guiding materials for the road map.
VI.At least two (we aim at four) public speeches on the issues within the project.
VII.At least three (we aim at four) PhD students involved directly into the research on the topic.
USE OF RESEACH RESULTS IN TEACHING
The results of the 2024 project will have limited impact upon teaching; however, the methods elaborated will surely be integrated into at least five courses taught on three Master programs at SPbU ('Media communications', 'AI in media and communications', and 'Robotics and AI'), as well as in two courses taught in HUST (their tentative titles will be available later this year). Moreover, some of the available instruments have already been integrated into teaching, but in a preliminary and non-user-friendly formats (due to lack of user interfaces). The project will help develop more student-friendly options for this software and its applications.
ACTIVITIES DURING 2024
I. Preparing a joint proposal for Russian Science Foundation / National Science Foundation of China (or National Natural Science Foundation of China) for 2025-2027.
II. Co-organising and co-delivering two seminars/workshops, one featuring road map creation and one focusing on data collection and methods (one at HUST, and one at SPBU).
III. In parallel to the workshops, delivering a series (at least two, but we aim at four) public/open lectures on digitalized depression, AI-based detection of mental illnesses, and methods of data collection and AI-based analysis.
IV. Preparing joint publications on methods of depression detection and AI-based support instruments, including communicative (conversational and visual) ones, which would involve at least three PhD students (we aim at four, two at HUST and two at SPBU). Holding interim (monthly or more frequent) online meetings for publication preparation. We aim at submitting two publications to the leading publishers by October 2024.
V. Co-organizing two HUST-SPBU PhD/graduate students’ online workshops, focusing on addressing digital health problems from an interdisciplinary approach.
VI. Co-participation at two international conferences (accepted for 2024/2025).
APPLICATION FOR FURTHER RESEARCH
The 2024 project is a preparatory one for the joint proposal for the RSF/NSFC grant competition.
Options for other applications and funding include:
1) National Natural Science Foundation of China;
2) Presidential grants for academic schools in Russia;
3) funding via the SPbU Center for International Media Research.
