New roles of research data infrastructure in research paradigm evolution

In previous decades, national scientific data centers distributed across scientific domains were a prevalent pattern in most countries. Since the 1990s, the US, UK and Germany had constructed and enhanced their national science data center systems in reliance on national institutions (Wang & Wang, 2022). They served as the pipelines and key nodes to facilitate the aggregation of research data from national laboratories, research institutions, universities, research communities, researchers and other innovation bodies across various disciplines. For its duty, the national scientific data center ultimately achieved vertical connectivity of data within a certain domain.

However, data from a single discipline is inadequate to respond to the complex scientific problems. Scientists often require diverse data, different analyses and an integrated research data platform across various disciplines to enhance their comprehensive understanding of the natural phenomena, and to address the increased complexity of socio-economic development challenges. There is an increasing number of problem-oriented RDIs, rather than research domain-oriented ones, in the global RI landscape (Figure 3). They are pioneering the breakthrough of traditional research fields boundaries with the trend of interdisciplinarity and the evolution of the fourth research paradigm, while also overcoming existing geographical and organizational information barriers.

Figure 3.

These research data or data-driven infrastructures coordinate integrated data and knowledge across various domains of natural sciences, including natural science and social science. This progresses gradually. For example, PANdata, the Photon and Neutron data infrastructure initiative, brought together 13 major European analytical RIs to establish a fully integrated, seamless, pan-European, information infrastructure supporting the scientific process in 2011 (Bicarregui et al., 2015). This initiative developed a common data format and management, offering a unity framework for the data generated by synchrotron radiation light sources, free electron laser facilities, laser light sources, and spallation neutron sources. It meant that researchers could take advantage of experimental data from more than one analytical infrastructure without replication, in order to create a more detailed, broader spectrum portrayal of the material world, to break down the barriers of experimental means. Research Infrastructure for EnvIRonmental Exposure assessmeNt in Europe (EIRENE RI) was included the ESFRI Roadmap for 2021 (ESFRI, 2021). This distributed data-driven infrastructure aims to support a large-scale, wide ranging research into the interdisciplinary assessment of environmental determinants of health. It will harmonize integrated services for data, knowledge and tools across a cross-cutting area of environmental and analytical chemistry, biology and toxicology, epidemiology, environmental and human exposure and risk assessment, pharmacokinetics, and geospatial modelling. But it mainly across the domains of natural science. Next, we use the Intergovernmental Panel on Climate Change (IPCC) as an example of the cross-cutting between natural science and social science. The object of IPCC was to provide governments with scientific information that they could use to develop climate policies (Dong & Zhang, 2014). Its Data Distribution Centre offered the climate, environment and socio-economic data from the past and also in scenarios projected into the future, and shown a good practice from research data, to information, knowledge, decision, finally to policy at international level (Stockhause & Lautenschlager, 2022).

With the evolution of the fourth paradigm, two notable branches have gained increasing significant: X-info, for the perception, collection and analysis of data and information, and Comp-X, for computing and simulation (Zhang et al., 2019). Research data has become a new type of innovation capital. Some scientific activities do not necessarily rely on physical experimental observation facilities or instruments from the outset, but rather on the accumulation of data collected for other purpose, usually called data reuse. As a one-station, unified, comprehensive digital research environment, RDIs will serve as the primary interface for direct interaction with researchers. The network and flexible nature of science leads to a significant transformation of the scientific service mode.

The role of X-info was demonstrated in the evolution of traditional RIs. The responses of analytical RI during and post COVID-19 era provide clues about the transformation of the digital and knowledge service environment of infrastructures. The restriction on travel during the COVID-19 era necessitated more standardized and digitalized procedures for remote user access across infrastructures, as well as enhanced interaction and collaboration between users and technical experts, and a comprehensive overview of information and knowledge to facilitate digital sample handling (LEAPS, 2021). These upgrades enabled the flexible transmission of scientific data to researchers without requiring on-site operation, and established a permanent capability.

Thanks to intelligent computing, RDIs with a scaling effect have become a goldmine for scientific discoveries. In recent years, AI scientists and robot scientists emerged in various disciplines, which force the data and knowledge service intelligently (Sun & Han, 2021). ChatGPT, Large Language Models (LLM) molded AI research assistants, and might be capable of works from fine-tuning language, extracting references and metadata from papers, to helping evaluate the trustworthiness of findings by aggregating the heuristics of many experts (Cramer, 2021; Davies et al., 2021; OECD, 2023; Zhao et al., 2023). This trend not only occurred in disciplines with a high level of digitization, good data accumulation and relatively well-defined problems, but also in finding solutions to complex problems. The European Integrated Infrastructure for Social Mining and Big Data Analytics (SoBigData++), has created a platform for designing and implementing of large-scale social science data experiments, serving as a digital academic laboratory to understand complex, globally interconnected societies.

With AI scientists (Boiko et al., 2023; Koscher et al., 2023; Merchant et al., 2023; Romera-Paredes et al., 2023) working in digital, autonomous labs (Szymanski et al., 2023), the RIDs will serve as the interface for providing efficient, flexible and intelligent knowledge services in physical or virtual digital academic laboratories.

The high-end characteristics of its communication platform were primarily evident in the efficient communication pattern, the aggregation of elite talents, and the delivery of high-quality content within academic discourse (Figure 5).

Figure 4.

Figure 5.

Firstly, the problem-oriented RDIs suggest a highly efficient communication pattern, fostering the unrestricted dissemination of knowledge for the benefit of researchers, scholars, students, and, more broadly, society worldwide, without barriers, during the implementation of mission-driven research endeavors. In Europe, Open scholarly communication in the European Research Area for Social Sciences and Humanities (OPERAS) constituted a distributed RI, aiming to promote Open Science by establishing a scholarly communication system with specific forms (monographs, critical editions, and edited bibliographies, amongst others). Meanwhile, communication was not limited to cross-geographical boundaries or cross-disciplines boundaries, but also transcended time. Scientific big data, historical or generated in real time, can be transformed into robust information and knowledge through simulation, modelling and analysis, bridging the gap between past, now and future. LLMs currently served as a new medium, reshaping the way and promoting the efficiency in how researchers interact with knowledge itself, in order to solve the pronounced conflicts between the ever-increasing volume of knowledge and the limited capacity of the human brain (Zhang et al., 2023).

Secondly, RDIs have formed scientific communities fostering the convergence of high-caliber talents and wisdom. They served as pivotal avenues for the generation of increased opportunities and the flow of knowledge, fostering in-depth collaborative efforts on a global scale.

Thirdly, RDIs will provide well-organized, high-quality, credible research and communication content, based on knowledge-based data curation (Sakai et al., 2019). On the one hand, open access journals, open data, data journals, traditional academic journals, academic conferences and other knowledge repositories would be organized in a manner that was more readily understandable and usable by AI, to discovery, integration and synthesis (Carter et al., 2023), and harmonized with whole lifecycle in physical RIs (Bunakov & Matthews, 2013). On the other hand, RDIs focus on not only the fine outputs, like academic articles report series etc., but also the whole process of experiments, including automated collection and aggregation, standardization and traceability of scientific research data with quality control. They will be remarkable and credible infrastructures in data-driven research paradigm.

Openness of research data was a prevalent requirement for innovation activities in the fourth research paradigm. However, there exist conflicts between openness and ethics due to various rights and demands of stakeholders, such as privacy, intellectual property, etc. (Li et al., 2022). Legal data ingestion, integration, retrieval and replication, transparency, fairness, citation as academic norms were hotly debated and widely considered in research data curation themes (Elliott et al., 2023; Kraemer, et al., 2021; Li & Shen, 2024; Tong et al., 2022). As central nodes in the cyber space, RDIs should play a crucial role in conducting curation and governance effectively, while balancing the needs of scientific openness and ethics. As research data converges and is reused on a larger scale and in more deeply connected contexts, issues of safety and ethics will manifest in more diverse forms. RDIs will shoulder more responsibility for balancing the openness and ethics, as well as addressing security considerations. They will capitalize fully on data resources, technical tools, standards and protocols, and working systems.

In recent years, new concepts and frameworks have been proposed and implemented in governance practice. In healthcare, the introduction of the Personal Health Train, which involves the submission of questions and algorithms to the data-centered infrastructures and networks (Beyan et al., 2020), is considered as an alternative to directly sharing basic data. This approach is considered to address the privacy risks associated with distributed analytics scenarios (van Soest et al., 2018; Welten et al., 2022).

Moreover, advanced information technologies such as blockchain, secure multiparty computation, and trusted execution environments are gaining traction in addressing research data ethics and safety issues, ensuring the integrity of each step of data workflows in RDIs. Oak Ridge National Laboratory developed a secure computation platform that allocates research data generated by infrastructures or instruments to provide a safe collaboration environment and capabilities for training a differential privacy enabled deep learning network on the super-computer Summit, both within and outside of the laboratory (Yoginath et al., 2021). Chinese National Marine Data Center launched a blockchain platform for data transmission in the Science and Technology Program in 2022. A model for data management and certification was constructed, and designed smart contracts such as automated control of the remittance process, data certification, etc., so as to ensure authenticity, consistency, completeness, and traceability of the data files and documents.