TY - JOUR
AU - Wu, Peiyi
AB - Abstract With the rise of artificial intelligence (AI), data are widely viewed as the “new oil”. However, data substantially differ from conventional resources in the sense that they are important not only for production but also for knowledge development and public policymaking. This article explores whether and how data reshape government–industry–university relations in the era of AI. Taking China’s AI innovation system as a case, this article investigates the dynamics of actor relations in the business subsystem, knowledge subsystem, and regulatory subsystem. The change of the fundamental input from physical resources to virtual data in AI innovation systems has significantly transformed the relations among industry, state, and academia, and digital platforms are playing an increasingly important role in business value creation, knowledge generation, and regulation formation due to their control of valuable data and frontier expertise in the context of uncertainty. 1. Introduction In the digital era, data have been regarded as the “new oil” (Parkins, 2017). Global internet protocol traffic has grown from 100 GB per day in 1992 to 46,600 GB per second in 2017, and it is expected to quadruple by 2022 (UNCTAD, 2019). The Fourth Industrial Revolution is strongly associated with digital technologies, such as artificial intelligence (AI), blockchain, big data, internet of things (IoT), and cloud computing, of which data are the fuel in most. Tremendous value has been captured with the progress of data mining technologies, and data acquisition has become a key area of business competition (Troisi et al., 2018). It is not surprising to see that the list of the worlds’ most valued companies is dominated by giant digital platforms (Kenney and Zysman, 2020). In addition, data are becoming more important for knowledge generation and public policy. The process of collecting, organizing, and analysing data is the key exercise in the generation of scientific knowledge (Foster et al., 2018). In the area of public administration, big data have been regarded as a “game-changer” (Maciejewski, 2017). There is an increasing reliance on data for public policymaking (Lane, 2016). Critical resources that are important for innovation but are difficult to imitate or transfer from one actor to another can largely affect actors’ strategies and meso-level system structure (Markard and Truffer, 2008). However, the existing innovation system approaches are largely hardware-focused (Weber and Truffer, 2017) and rarely examine how the change in foundational input from physical materials to virtual data influences system-level transformations. Data inherently differ from physical resources. Firstly, data are never exhausted. More value can be exploited when one type of data is linked to others, creating new data (Rumbold and Pierscionek, 2018). Secondly, while legal ownership of a thing usually implies the ability to exclude others from its possession or use, data are characterized by non-rivalry and can be used simultaneously by different users. Legally, data per se cannot be owned (ibid.). Data are often co-produced by those who collect the data and those who are measured. Moreover, data have externalities, that is, the data created by one individual may contain information about others (Weber, 2017; Agrawa et al.,2019 (Arenal et al., 2020)). Finally, due to the network effect of data in the platform economy (Rochet and Tirole, 2006), valuable data are usually controlled by a few digital platforms. This article aims to reveal how data, as a new critical resource, shape actor relations in data-driven innovation systems. Under different levels of aggregations, actors can refer to either specific individuals, individual organizations, or general categories of organizations (Avelino and Wittmayer, 2016). In this article, we focus on the relations between different types of organizations involved in the generation, diffusion, and commercialization of innovations. These actors usually include, but not limited to, universities, research institutes, investors, firms, government, intermediaries, and end-users, with each taking a different division of labour in an innovation system. At a higher level of aggregation, the “triple helix” literature argues that the strategic interactions and collaboration between the state, academia, and industry play a central role in defining the types and structures of an innovation system (Etzkowitz and Leydesdorff, 2000). Here, the industry is a broad institutional sphere consisting of many business actors involved in customer value creation. In different innovation systems, there are various actor relations, which are the way how actors interact (e.g. compete, exchange, substitute, command, and collaborate) (Malerba, 2005) and exert impacts on each other. As data are gaining a more essential role in production, innovation, and governance, it is reasonable to expect that industry, state, and academia, might experience transformations in terms of their roles and relations in data-driven innovation systems. Actor relations in innovation systems are contingent on territorial contexts (Lundvall, 2007; Ranga and Etzkowitz, 2013), as well as sector-specific characteristics (Malerba and Nelson, 2011). We draw insights from the “triple helix” and the sectoral innovation system (SIS) approach and differentiate three subsystems to examine how data shape the dynamics of actor’s roles and their relations in value creation, knowledge generation, and regulation formation. In the empirical analysis, we apply a case study of China’s AI innovation system. The current AI technology is mostly based on machine learning and relies on the feeding of data for better accuracy (Kaplan and Haenlein, 2019). As a general-purpose technology (Agrawa et al.,2019), AI can build on existing practices and has been pervasively adopted by incumbent sectors to upgrade their traditional businesses, forming an “AI+X” industry. To date, a vertical industry value chain has been formed around AI, ranging from infrastructure providers (e.g. computing chips and sensors) to technology developers (e.g. computer vision, voice recognition, and natural language processing) to application scenarios (e.g. manufacturing, finance, transportation, healthcare) (Figure 1). China has become one of the leading countries in AI industry development (Xue et al., 2018; Arenal et al., 2020). Underlying this fast growth is the mass production, flow, distribution, and control of data, which has resulted in a shift of relations among heterogeneous actors. Figure 1. Open in new tabDownload slide A sketch map of the vertically related industry chain around AI. Source: compiled by the authors Figure 1. Open in new tabDownload slide A sketch map of the vertically related industry chain around AI. Source: compiled by the authors The findings show that actor relations in China’s AI innovation system are deviating from China’s “statist” triple helix configuration, and the industry is gaining a more important role in value creation, knowledge generation, and regulation formation. Due to their strong capacity to aggregate and utilize data, digital platforms have become the prime mover in shaping the AI innovation systems, where the actors’ roles, networks, and institutions are still very much in flux. This research thus contributes to the innovation system research by revealing how data shape meso-level actor relations in the era where innovations are becoming increasingly virtualized and digitalized. This article is organized as follows. Section 2 provides a review of the innovation system approaches and current status of actor relations, based on which we build an analytical framework to illustrate the dynamics of the actors’ relations in a SIS from the following three aspects: the business subsystem, knowledge subsystem, and regulatory subsystem. Section 3 introduces the research methods and data sources. Section 4 presents the empirical results. Finally, we conclude our contributions and suggestions for future research in Section 5. 2 Theoretical basis and analytical framework 2.1 Innovation systems and actor relations Since Freeman’s (1987) seminal work on the national innovation system (NIS), various innovation system (IS) approaches have been developed, such as regional innovation systems (e.g. Cooke et al. 2004), SISs (e.g. Malerba, 2002), technological innovation systems (e.g. Bergek et al., 2008), and global innovation systems (e.g. Binz and Truffer, 2017). Albeit with different core interests and orientations, these IS approaches all highlight the interactions between institutions, actors, and networks (Weber and Truffer, 2017). From an IS perspective, an innovation involves systematic interactions among heterogeneous actors for knowledge generation, diffusion, and commercialization, and the linkages between governments, firms, universities, and other institutions play important roles (Malerba and Montobbio, 2004). In an innovation system, firms are viewed as central actors, and their learning and capability accumulation is of central importance for innovation (Malerba and Nelson, 2011). Within the business sphere, vertical linkages with upstream suppliers may provide new knowledge and information for production and innovation, while interactions with demanding users are also important channels for learning and new firm entry (Adams et al., 2019). In addition to these business actors, innovation needs to go beyond the technology focus and build a supporting and orienting system to include other types of factors, such as primary and secondary education, universities, the public research system, and governments. Universities are the key actors in conducting basic research, as well as in the formation of human capital. The government plays a major role in directing the development of specific sectors through plannings, policies, regulations, and standards. Also, the roles of other actors, such as financial organizations and intermediary organizations, in supporting research, technology diffusion, and production should not be downplayed (Watkins et al., 2015). The government–industry–university relation (“triple helix”) is viewed as one of the most important characteristics of an innovation system (Etzkowitz and Leydesdorff, 2000). In some countries, the national state dominates and directs the relations between academia and industry (a “statist” model), while in other countries, the three institutional spheres are separated (a “laissez-faire” model) (ibid.). In practice, most countries adopt a “balanced” model, in which the three institutional spheres overlap, with each taking some of the roles of the others. Hence, there are various relationships among these spheres, e.g. technology transfer, collaboration and conflict moderation, collaborative leadership, substitution, and networking (Ranga and Etzkowitz, 2013). However, the triple helix research has mainly focused on macro-level theorizing and has paid little attention to how actors’ practices affect the dynamics of actors’ relations at the meso-level (Sarpong et al., 2017). Though confined by national characteristics, the types of actor relations in a country may differ from sectoral system to sectoral system due to sectoral-specific knowledge base, learning process, demand characteristics, and institutions (Malerba and Montobbio, 2004). Malerba (2002: 250) defines a sectoral system of production and innovation as “a set of new and established products for specific uses and the set of agents carrying out market and non-market interactions for the creation, production, and sale of those products.” SIS highlights the role of sector-specific characteristics, such as the knowledge base and institutional regimes, in explaining why a country is successful in certain sectors but not in others (Malerba and Nelson, 2011). Knowledge plays a key role in innovation and affects firms’ learning capabilities. In some sectors (e.g. biotechnology), the knowledge base is highly dependent on scientific research; thus, universities and public research institutes play a vital role (Chen and Lin, 2016). However, in other sectors (e.g. software service), knowledge is largely gained from interacting with the demand side; thus, supplier-user relations are of more importance (Malerba and Nelson, 2011). Additionally, institutions such as laws, standards, norms, and cultural beliefs shape the interactions of the actors in a sectoral system. Some institutions (e.g. patent systems) are constituted at the national level, but some are specific to sectors, imposing different impacts on the behaviours of the actors in different sectors. 2.2 Research gaps and an analytical framework To date, the SIS literature has focused on the role of the knowledge base and institutional characteristics in shaping actor relations. However, from an actor-oriented perspective, the distribution of resource endowments and actors’ strategies can significantly influence the meso-level actor configurations. As Markard and Truffer (2008: 446) express, “the constraining effect of given resource endowments may even hold (at least temporarily) for a broader range of actors in the innovation system, thus leading to an entrenchment of prevailing structures, even to path dependencies or lock-in effects.” Actors with complementary critical resources tend to coordinate and collaborate more, while actors with similar or substitutable resources may compete more. In the early stage of development, those who control a wide range of critical resources could become prime movers and be able to direct the trajectory of an innovation system (Markard and Truffer, 2008). Digitalization has not only accelerated the pace of innovation activities but also promoted new models of research and innovation (e.g. open innovation), bringing new requirements for IS research (Weber and Truffer, 2017). Previous research has shed some light on the role of material resources (e.g. capital), immaterial resources (e.g. reputations, custom relations), and human resources (e.g. technical expertise) in influencing the relations between business actors (Markard and Truffer, 2008). However, it is less known whether and how the relations between different types of organizations would be transformed when data become a more foundational resource. As discussed above, data are becoming an increasingly critical asset for not only businesses but also public policy and academic research. Although data can be infinite and non-competitive, the actual control of data is very uneven among system actors, which could result in a transformation of the actors’ roles and their relations in an innovation system. Particularly, the network effect of the platform economy has enabled digital platforms to become the main aggregators of data. The integration of data, capital, and technology has driven the growth of giant digital platforms (Lee et al., 2018). Kenney and Zysman (2016: 62) even claim that “if the industrial revolution was organized around the factory, today’s changes are organized around these digital platforms.” It is reasonable to expect that the role of heterogeneous actors and their relations in SISs may experience substantial transformations under the context where data become a critical factor of production and innovation. Yet, whether and how exactly data affect actors’ relations should be investigated by the main functions they are involved in an innovation system. Alongside the process of innovation generation and commercialization, scholars have identified different types of subsystems or functions. For instance, Valkokari (2015) differentiates three ecosystems: a business ecosystem, a knowledge ecosystem, and an innovation ecosystem based on their outcomes, interactions, actor roles, and logic of action. Business ecosystems focus on customer value creation, knowledge ecosystems focus on knowledge generation, and innovation ecosystems are the integrating mechanisms of the former two ecosystems. However, while most IS research has highlighted the activities of knowledge generation and value creations, few studies have paid attention to the process of institution formation, which could significantly affect the outcomes of the business subsystem and knowledge subsystem. In IS studies, an institution is usually viewed as an unmoveable explanatory variable (Weber and Truffer, 2017). In fact, the institution is not always “out there” as an exogenous factor; rather, it is often in the process of “making”. Formal institutions (e.g. regulations and standards) are often viewed as outcomes of the “top-down” process, and the government is the primary actor involved, but the recent literature has challenged this tradition and has emphasized the role of industry actors in shaping the institutional environment, especially in emerging industries (e.g. Battilana et al., 2009; Yap and Truffer, 2019; Yu and Gibbs, 2020). As the “triple helix” literature suggests, the industry is mainly involved in wealth generation, academia primarily focuses on knowledge exploration, and the government creates a normative system that regulates the innovation process (Vaivode, 2015). Here, we take this rather generalized division as a frame and propose to analyse the dynamics of actors’ relations in an SIS from the following three subsystems: a business subsystem, a knowledge subsystem, and a regulatory subsystem (Table 1). The basic function of the business subsystem is exploiting knowledge and other resources for customer value creation, which involves a vertical chain of related industries from upstream suppliers to downstream users. Consumers also play a role in the development of new products mainly by providing information and feedback to producers (Lundvall, 2007; Fontana and Malerba, 2010). The main function of the knowledge subsystem is to develop new scientific knowledge, and universities and research institutes usually play the central role. Besides, universities are the main source of human capital. For the regulatory subsystem, its main function is to form formal institutions (e.g. intellectual property protection and industry standards) to regulate innovation activities. In this subsystem, the government has the main authority and legitimacy. Nonetheless, the same actor can simultaneously be involved and play different roles in each subsystem (Valkokari, 2015). Also, there could be diverse feedback loops between these subsystems. With the rising role of data in production, innovation, and policymaking, how will actors’ roles and relations in SISs be transformed? We explore this question by examining the actors’ relations in China’s AI innovation system. Table 1. Main function and actors of general SIS subsystems Business subsystem Value creation Firms and users are the core actors, other actors loosely involved (Valkokari, 2015) Knowledge subsystem Knowledge development Universities are the primary actors in the generation of scientific knowledge and the formation of human capital; firms are mainly involved in applied knowledge development. Regulatory subsystem Regulation formation Government has the main authority and legitimacy; universities, firms and other actors play a complimentary role. Business subsystem Value creation Firms and users are the core actors, other actors loosely involved (Valkokari, 2015) Knowledge subsystem Knowledge development Universities are the primary actors in the generation of scientific knowledge and the formation of human capital; firms are mainly involved in applied knowledge development. Regulatory subsystem Regulation formation Government has the main authority and legitimacy; universities, firms and other actors play a complimentary role. Open in new tab Table 1. Main function and actors of general SIS subsystems Business subsystem Value creation Firms and users are the core actors, other actors loosely involved (Valkokari, 2015) Knowledge subsystem Knowledge development Universities are the primary actors in the generation of scientific knowledge and the formation of human capital; firms are mainly involved in applied knowledge development. Regulatory subsystem Regulation formation Government has the main authority and legitimacy; universities, firms and other actors play a complimentary role. Business subsystem Value creation Firms and users are the core actors, other actors loosely involved (Valkokari, 2015) Knowledge subsystem Knowledge development Universities are the primary actors in the generation of scientific knowledge and the formation of human capital; firms are mainly involved in applied knowledge development. Regulatory subsystem Regulation formation Government has the main authority and legitimacy; universities, firms and other actors play a complimentary role. Open in new tab No. . Position(s) . Organization type . Location . 1 Director of public relation AI start-up in news and video Beijing 2 Director of Technology and Information Government department Suzhou 3 Vice executive secretary AI industry association Suzhou 4 Director of AI lab Research Institute in nanotech Suzhou 5 Co-founder/Chief Technology Officer (CTO) AI start-up in healthcare Suzhou 6 Director of R&D management Transnational robot company Suzhou 7 Chief executive officer (CEO) Start-up in robots Suzhou 8 Co-founder/CEO AI start-up in voice recognition Suzhou 9 CTO AI start-up in face recognition Suzhou 10 CEO AI start-up in healthcare Suzhou 11 Director of Design Architecture design institute Suzhou 12 Director Research institute in automobile Suzhou 13 Deputy director of technology & information Government department Shanghai 14 AI lab researcher/Professor University Hangzhou 15 Director of AI lab Internet company Hangzhou 16 Chief strategy officer ICT company Hangzhou 17 Director of R&D AI research institute Hangzhou 18 Director of public relation Company in security technology Hangzhou 19 Dean of AI Research Institute University Shanghai 20 Director of R&D management Transnational robot company Suzhou 21 Director of AI lab University Hefei 22 Deputy director of strategic planning ICT company Shenzhen 23 Director of game Internet platform Shenzhen 24 Director of AI lab Internet platform Shenzhen 25 Co-founder/CEO AI start-up in image recognition Shenzhen 26 Chief strategy officer; Director of R&D ICT company Shenzhen 27 Researcher of public policy research Internet platform Shenzhen 28 CEO Start-up in robots Shenzhen 29 CTO Start-up in robots Shenzhen 30 Deputy director of public relation Start-up in 3D sensor Shenzhen 31 CEO Start-up in robots Shenzhen 32 CEO Start-up in robots Shenzhen 33 Director of public relations Transnational IT company Beijing 34 Chief AI Scientist; Director of public relation Internet platform Beijing 35 Deputy director of research cooperation AI start-up in face recognition Beijing 36 Professor in semiconductor University Beijing 37 Vice president AI start-up in face recognition Beijing 38 Director of AI ethics and safety AI research institute Beijing 39 Researcher of AI chips ICT company Hangzhou 40 Professor in semiconductor University Hangzhou 41 Director of AI lab Internet platform Hangzhou 42 Director of cloud research Internet platform Hangzhou 43 Director of technology Database company Shanghai 44 Vice president Start-up in AI chips Shanghai 45 Co-founder/Vice president AI start-up in voice recognition Beijing 46 Deputy director of public policy research Online car-hailing platform Beijing 47 Vice president Internet platform Beijing 48 Vice president Internet platform Beijing 49 Secretary-general Internet industry association Beijing 50 Director of data authority Government department Hangzhou 51 Director of big data centre Government department Shanghai 52 Director of public relation AI start-up in news and video Beijing 53 Director of autonomous driving Internet platform Beijing 54 President/CEO AI start-up in healthcare Harbin 55 CEO AI start-up in healthcare Harbin 56 CEO Start-up in big data Harbin 57 President AI start-up in agriculture Harbin 58 President AI start-up in oil mining Daqing 59 President/CEO Start-up in public service Daqing 60 Deputy general manager of UK branch AI start-up in face recognition London 61 Director of public policy research Start-up in education Beijing 62 President ICT company Shenzhen 63 President/CEO AI start-up in face recognition Shenzhen 64 Chief scientist Financial company Shenzhen 65 Chief scientist Chips manufacture Shanghai 66 Co-founder/CEO AI start-up in healthcare Tai’an No. . Position(s) . Organization type . Location . 1 Director of public relation AI start-up in news and video Beijing 2 Director of Technology and Information Government department Suzhou 3 Vice executive secretary AI industry association Suzhou 4 Director of AI lab Research Institute in nanotech Suzhou 5 Co-founder/Chief Technology Officer (CTO) AI start-up in healthcare Suzhou 6 Director of R&D management Transnational robot company Suzhou 7 Chief executive officer (CEO) Start-up in robots Suzhou 8 Co-founder/CEO AI start-up in voice recognition Suzhou 9 CTO AI start-up in face recognition Suzhou 10 CEO AI start-up in healthcare Suzhou 11 Director of Design Architecture design institute Suzhou 12 Director Research institute in automobile Suzhou 13 Deputy director of technology & information Government department Shanghai 14 AI lab researcher/Professor University Hangzhou 15 Director of AI lab Internet company Hangzhou 16 Chief strategy officer ICT company Hangzhou 17 Director of R&D AI research institute Hangzhou 18 Director of public relation Company in security technology Hangzhou 19 Dean of AI Research Institute University Shanghai 20 Director of R&D management Transnational robot company Suzhou 21 Director of AI lab University Hefei 22 Deputy director of strategic planning ICT company Shenzhen 23 Director of game Internet platform Shenzhen 24 Director of AI lab Internet platform Shenzhen 25 Co-founder/CEO AI start-up in image recognition Shenzhen 26 Chief strategy officer; Director of R&D ICT company Shenzhen 27 Researcher of public policy research Internet platform Shenzhen 28 CEO Start-up in robots Shenzhen 29 CTO Start-up in robots Shenzhen 30 Deputy director of public relation Start-up in 3D sensor Shenzhen 31 CEO Start-up in robots Shenzhen 32 CEO Start-up in robots Shenzhen 33 Director of public relations Transnational IT company Beijing 34 Chief AI Scientist; Director of public relation Internet platform Beijing 35 Deputy director of research cooperation AI start-up in face recognition Beijing 36 Professor in semiconductor University Beijing 37 Vice president AI start-up in face recognition Beijing 38 Director of AI ethics and safety AI research institute Beijing 39 Researcher of AI chips ICT company Hangzhou 40 Professor in semiconductor University Hangzhou 41 Director of AI lab Internet platform Hangzhou 42 Director of cloud research Internet platform Hangzhou 43 Director of technology Database company Shanghai 44 Vice president Start-up in AI chips Shanghai 45 Co-founder/Vice president AI start-up in voice recognition Beijing 46 Deputy director of public policy research Online car-hailing platform Beijing 47 Vice president Internet platform Beijing 48 Vice president Internet platform Beijing 49 Secretary-general Internet industry association Beijing 50 Director of data authority Government department Hangzhou 51 Director of big data centre Government department Shanghai 52 Director of public relation AI start-up in news and video Beijing 53 Director of autonomous driving Internet platform Beijing 54 President/CEO AI start-up in healthcare Harbin 55 CEO AI start-up in healthcare Harbin 56 CEO Start-up in big data Harbin 57 President AI start-up in agriculture Harbin 58 President AI start-up in oil mining Daqing 59 President/CEO Start-up in public service Daqing 60 Deputy general manager of UK branch AI start-up in face recognition London 61 Director of public policy research Start-up in education Beijing 62 President ICT company Shenzhen 63 President/CEO AI start-up in face recognition Shenzhen 64 Chief scientist Financial company Shenzhen 65 Chief scientist Chips manufacture Shanghai 66 Co-founder/CEO AI start-up in healthcare Tai’an Open in new tab No. . Position(s) . Organization type . Location . 1 Director of public relation AI start-up in news and video Beijing 2 Director of Technology and Information Government department Suzhou 3 Vice executive secretary AI industry association Suzhou 4 Director of AI lab Research Institute in nanotech Suzhou 5 Co-founder/Chief Technology Officer (CTO) AI start-up in healthcare Suzhou 6 Director of R&D management Transnational robot company Suzhou 7 Chief executive officer (CEO) Start-up in robots Suzhou 8 Co-founder/CEO AI start-up in voice recognition Suzhou 9 CTO AI start-up in face recognition Suzhou 10 CEO AI start-up in healthcare Suzhou 11 Director of Design Architecture design institute Suzhou 12 Director Research institute in automobile Suzhou 13 Deputy director of technology & information Government department Shanghai 14 AI lab researcher/Professor University Hangzhou 15 Director of AI lab Internet company Hangzhou 16 Chief strategy officer ICT company Hangzhou 17 Director of R&D AI research institute Hangzhou 18 Director of public relation Company in security technology Hangzhou 19 Dean of AI Research Institute University Shanghai 20 Director of R&D management Transnational robot company Suzhou 21 Director of AI lab University Hefei 22 Deputy director of strategic planning ICT company Shenzhen 23 Director of game Internet platform Shenzhen 24 Director of AI lab Internet platform Shenzhen 25 Co-founder/CEO AI start-up in image recognition Shenzhen 26 Chief strategy officer; Director of R&D ICT company Shenzhen 27 Researcher of public policy research Internet platform Shenzhen 28 CEO Start-up in robots Shenzhen 29 CTO Start-up in robots Shenzhen 30 Deputy director of public relation Start-up in 3D sensor Shenzhen 31 CEO Start-up in robots Shenzhen 32 CEO Start-up in robots Shenzhen 33 Director of public relations Transnational IT company Beijing 34 Chief AI Scientist; Director of public relation Internet platform Beijing 35 Deputy director of research cooperation AI start-up in face recognition Beijing 36 Professor in semiconductor University Beijing 37 Vice president AI start-up in face recognition Beijing 38 Director of AI ethics and safety AI research institute Beijing 39 Researcher of AI chips ICT company Hangzhou 40 Professor in semiconductor University Hangzhou 41 Director of AI lab Internet platform Hangzhou 42 Director of cloud research Internet platform Hangzhou 43 Director of technology Database company Shanghai 44 Vice president Start-up in AI chips Shanghai 45 Co-founder/Vice president AI start-up in voice recognition Beijing 46 Deputy director of public policy research Online car-hailing platform Beijing 47 Vice president Internet platform Beijing 48 Vice president Internet platform Beijing 49 Secretary-general Internet industry association Beijing 50 Director of data authority Government department Hangzhou 51 Director of big data centre Government department Shanghai 52 Director of public relation AI start-up in news and video Beijing 53 Director of autonomous driving Internet platform Beijing 54 President/CEO AI start-up in healthcare Harbin 55 CEO AI start-up in healthcare Harbin 56 CEO Start-up in big data Harbin 57 President AI start-up in agriculture Harbin 58 President AI start-up in oil mining Daqing 59 President/CEO Start-up in public service Daqing 60 Deputy general manager of UK branch AI start-up in face recognition London 61 Director of public policy research Start-up in education Beijing 62 President ICT company Shenzhen 63 President/CEO AI start-up in face recognition Shenzhen 64 Chief scientist Financial company Shenzhen 65 Chief scientist Chips manufacture Shanghai 66 Co-founder/CEO AI start-up in healthcare Tai’an No. . Position(s) . Organization type . Location . 1 Director of public relation AI start-up in news and video Beijing 2 Director of Technology and Information Government department Suzhou 3 Vice executive secretary AI industry association Suzhou 4 Director of AI lab Research Institute in nanotech Suzhou 5 Co-founder/Chief Technology Officer (CTO) AI start-up in healthcare Suzhou 6 Director of R&D management Transnational robot company Suzhou 7 Chief executive officer (CEO) Start-up in robots Suzhou 8 Co-founder/CEO AI start-up in voice recognition Suzhou 9 CTO AI start-up in face recognition Suzhou 10 CEO AI start-up in healthcare Suzhou 11 Director of Design Architecture design institute Suzhou 12 Director Research institute in automobile Suzhou 13 Deputy director of technology & information Government department Shanghai 14 AI lab researcher/Professor University Hangzhou 15 Director of AI lab Internet company Hangzhou 16 Chief strategy officer ICT company Hangzhou 17 Director of R&D AI research institute Hangzhou 18 Director of public relation Company in security technology Hangzhou 19 Dean of AI Research Institute University Shanghai 20 Director of R&D management Transnational robot company Suzhou 21 Director of AI lab University Hefei 22 Deputy director of strategic planning ICT company Shenzhen 23 Director of game Internet platform Shenzhen 24 Director of AI lab Internet platform Shenzhen 25 Co-founder/CEO AI start-up in image recognition Shenzhen 26 Chief strategy officer; Director of R&D ICT company Shenzhen 27 Researcher of public policy research Internet platform Shenzhen 28 CEO Start-up in robots Shenzhen 29 CTO Start-up in robots Shenzhen 30 Deputy director of public relation Start-up in 3D sensor Shenzhen 31 CEO Start-up in robots Shenzhen 32 CEO Start-up in robots Shenzhen 33 Director of public relations Transnational IT company Beijing 34 Chief AI Scientist; Director of public relation Internet platform Beijing 35 Deputy director of research cooperation AI start-up in face recognition Beijing 36 Professor in semiconductor University Beijing 37 Vice president AI start-up in face recognition Beijing 38 Director of AI ethics and safety AI research institute Beijing 39 Researcher of AI chips ICT company Hangzhou 40 Professor in semiconductor University Hangzhou 41 Director of AI lab Internet platform Hangzhou 42 Director of cloud research Internet platform Hangzhou 43 Director of technology Database company Shanghai 44 Vice president Start-up in AI chips Shanghai 45 Co-founder/Vice president AI start-up in voice recognition Beijing 46 Deputy director of public policy research Online car-hailing platform Beijing 47 Vice president Internet platform Beijing 48 Vice president Internet platform Beijing 49 Secretary-general Internet industry association Beijing 50 Director of data authority Government department Hangzhou 51 Director of big data centre Government department Shanghai 52 Director of public relation AI start-up in news and video Beijing 53 Director of autonomous driving Internet platform Beijing 54 President/CEO AI start-up in healthcare Harbin 55 CEO AI start-up in healthcare Harbin 56 CEO Start-up in big data Harbin 57 President AI start-up in agriculture Harbin 58 President AI start-up in oil mining Daqing 59 President/CEO Start-up in public service Daqing 60 Deputy general manager of UK branch AI start-up in face recognition London 61 Director of public policy research Start-up in education Beijing 62 President ICT company Shenzhen 63 President/CEO AI start-up in face recognition Shenzhen 64 Chief scientist Financial company Shenzhen 65 Chief scientist Chips manufacture Shanghai 66 Co-founder/CEO AI start-up in healthcare Tai’an Open in new tab 3 Methods and data 3.1 Methods This article adopts a case study of the AI industry in China, which is among the few forerunners in AI development (Xue et al., 2018; Arenal et al., 2020). A case study can provide an in-depth and detailed analysis of a particular case (e.g. an event, a project, and a system) through one or more methods (Yin, 2010). It allows researchers to inspect a phenomenon with more concrete and practical information, and to develop theories through testing, falsifying, or expanding existing theoretical explanations (Baxter, 2010). Theoretically, a sectoral system of innovation is not bounded to any geographical scale, but it is usually practised within territorial spaces with distinct institutions, actors, and networks. In other words, a country’s NIS characteristics confine the features of the SISs within its territory (Malerba and Montobbio, 2004). It is, therefore, necessary to examine the NIS characteristics when examining a specific SIS in a particular country. For a long period, China’s NIS has shown a “statist” characteristic, in which the state orients industry development and academic research (Ranga and Etzkowitz, 2013). When it comes to the AI innovation system, the government-industry-university relation is changing. In one of the latest contributions, Arenal et al. (2020) adopt an asymmetric triple helix model to summarize China’s AI innovation system, in which top-down government supports and bottom-up industry practices drive China’s AI development. However, existing research only provides a descriptive analysis of the status of actor actions but lack insights about what contributes to the dynamics of the triple helix configuration. This article explores how data shape the dynamics of government-industry-university relations in China’s AI innovation system. 3.2 Data and analysis Semi-structured interviews are the main data source of this research. Compared to questionnaires, semi-structured interviews allow flexibilities in the spoken exchange of information to better fit each interviewee’s context and expertise. From May 2018 to May 2020, we conducted 66 face-to-face interviews with AI-related companies, government agencies, universities, industry associations, and research institutions in China (one in the UK) (see Appendix 1). These interviews were mainly guided by the following questions: (a) What are the factors facilitating or obstructing China’s AI industry? (b) How are critical resources (e.g. data and talent) distributed and how do they flow within the innovation system? (c) Whether and how the relations between heterogeneous actors have been transformed? (d) How do different actors respond to the emergence of AI and its consequent changes? In most of these interviews, we managed to talk to respondents from the top management level (e.g. CEOs, founders, and department managers) and gathered in-depth observations from industry experts. In each interview, we asked the interviewees to share their working background, the development status and practices of their organizations, and observations on China’ AI innovation system. These interview materials are supplemented by many secondary industry reports, government documents, and news reports. Besides, the authors have participated in many conferences, workshops, and meetings regarding AI development and governance with many experts from the industry, academia, and government, which also offers a great platform to directly observe the relations between these actors. The data collected from these sources were transcribed, analysed and triangulated. We adopted a standard qualitative data analysis approach through data description, classification, and connection (Dey, 2003). The key process is breaking up the transcribed data, organizing them into different themes, and identifying relationships between these themes to answer our research questions. Though the data sources are mainly acquired from various individual interviewees (e.g. entrepreneurs, government officials, researchers), the analytical focus of this research is the relations between different types of organizations at a higher level of aggregation. In the empirical presentation, we aggregated relevant themes to the three system activities: knowledge generation, value creation, and regulation formation, and summarized how data shaped actor relations within them. Typical expressions from the interviewees were cited or quoted to illustrate our case description. For ethical concern, these interviewees were anonymized but numbered according to our interview sequence in the following case presentation. 4 Dynamics of actor relations in china’s AI SIS 4.1 The development of AI industry in China China’s academic research on AI began in the early 1980s but progressed slowly. Since the late 1990s, data, which are one of the pillars of current AI development, grew drastically in China. China was among the few countries that seized the windows of opportunity created by the advent of the internet and became one of the leading countries in the internet economy (UNCTAD, 2019). In addition to the digital giants, such as Baidu, Alibaba, and Tencent (BAT) that emerged from China’s internet industry, a dynamic and innovative ecosystem with many indigenous innovations has also emerged. The pervasive penetration of the internet in China provides a massive scale of data to fuel AI development. During this process, the computing capacity to deal with large-scale data has also improved considerably. In the early 2010s, the breakthrough in deep learning algorithms from the global academic sphere completed the last pillar of AI development. Chinese digital giants began to actively attract a large number of deep learning scientists from research labs to the industry. Many deep learning scientists from universities also started their own AI businesses. Sense Time is a typical example. Sense Time was founded in 2014 by Professor Tang Xiao’ou, who had led a world-leading research team in computer vision at the Chinese University of Hong Kong. The research team achieved many advanced algorithms in computer vision but was trapped by the limited data volume and computing capacity (Interviewee 35). The situation soon changed when venture capital helped Tang Xiao’ou establish Sense Time and invested heavily in recruiting AI talents and building computing platforms with one of the largest GPU (Graphics Processing Unit) clusters in Asia. Sense Time’s technology was soon adopted by many sectors, which in turn enabled Sense Time to build a huge image database. With this virtuous circle, venture capitalists continued to invest in this company, making it one of the world’s most valued AI unicorns. The victory of AlphaGo over human Go players in 2016 noticeably enhanced investors’ faith in AI (Interviewee 5, 9, 29, 45). Traditional sectors began using AI for upgrading at a faster rate, and new firms sprang up to exploit the huge markets in various scenarios. In 2017, China’s central government implemented the Development Plan for Next-Generation Artificial Intelligence, aiming to build China as a global AI superpower by 2030 (Roberts et al., 2020). For local governments, AI is mostly viewed as another lane for economic development and regional competition, and they are enthusiastically promoting AI industries with various policy supports (Interviewee 2, 13). By 2018, China had the second-largest number of AI firms and received the largest scale of AI investment globally (Xue et al., 2018). If quality is considered, however, China’s AI innovation system is way behind that of the United States in terms of top talent, basic knowledge, and hardware (Ding, 2018). Nevertheless, China has accumulated a dominant advantage in data (ibid.), which are not evenly controlled by industry, state, and academia. Consequently, we observe a pattern change in the actors’ roles and relations in the business, knowledge, and regulatory subsystems. 4.2 Actor relations in the business subsystem 4.2.1 Vertical relations between upstream producers and downstream users China’s AI value chain consists of infrastructure providers, technology developers, and application users (Figure 2). Many industrial interviewees stated that there is no such an “AI industry” but only an “AI+ industry” (e.g. Interviewee 3, 16). That is, AI alone cannot create any value without integrating it into traditional sectors. Recognizing the strategic importance of data and application markets, as well as facilitated by the lower threshold of AI algorithm development, “now almost everyone is planning to expand their business to the whole industry chain” (Interviewee 58). Figure 2. Open in new tabDownload slide Trends of business expansion in China’s AI industry chain. Figure 2. Open in new tabDownload slide Trends of business expansion in China’s AI industry chain. For upstream infrastructure suppliers, to gain access to data and sector-specific knowledge, as well as to make fast profits from the huge application market, many have penetrated the application scenarios by providing various types of end products. As the application markets are in the process of being defined and reshaped, it is relatively easier for infrastructure providers to enter the market because of the connections they have established. Take Huawei, the dominant telecommunication equipment provider in China, as an example. With the rise of AI, Huawei has made great investments in many types of AI, including AI chips, algorithms, industrial applications such as car networking and smart manufacturing, and end products such as smartphones, scrutiny cameras, personal commuters, and wearable devices. By 2018, although still positioning itself as an infrastructure provider (Interviewee 62), Huawei’s consumer business had become its largest source of sales revenue. On the other hand, several entrepreneurs from the lower parts of the value chain have been concerned that these infrastructure providers’ extensions to the downstream industry have somewhat endangered their vertical collaborations (Interviewee 60). Similarly, technology developers, such as Sense Time, Magvii, Yitu, and iFly, and digital platforms, such as BAT, have expanded their businesses to application scenarios through, e.g. collaborating with downstream users, acquisitions, investing in new application companies, and selling end products directly. Many of these companies are striving to build their business ecosystems because they need to get into application scenarios to combine their technologies with sector-specific data and know-how so that the performance can be improved (Interviewee 8, 45). End products such as smartphones, intelligent speakers, and automobiles have been highly valued because they can provide channels to obtain user data and marketing opportunities. For instance, Tencent is a digital platform focusing on social networks and entertainment, but it has also established a department of autonomous driving to gain a new channel to obtain data and users. A technology director from the department illustrated the rationale as follows: “On the one hand, autonomous driving could bring an industrial transformation where cars are at the centre for future mobility and entertainment. On the other hand, it is about the volume of data related to autonomous driving, which could be very important to our cloud business.” (Interviewee 53) Application users in traditional sectors have also realized that they are holding a goldmine of data. As the threshold of AI algorithm development has been greatly lowered by global open-source platforms, many application users actively go upstream through the “+AI” strategy to take full advantage of their data assets and even become AI technology developers. For example, Ping An Group, one of the largest integrated finance companies in China, attaches strategic importance to the role of AI in business development and has recruited a research team of more than 1000 talented individuals in AI, blockchain, and cloud computing. The primary rationale is to exploit Ping An’s rich data and application scenarios in investment, insurance, and banking and build technology platforms for external partners. These efforts have enabled Ping An to become a leading company in Robo-advisor and financial risk control. In August 2019, the Ministry of Science and Technology named Ping An Group one of the National Open Innovation Platforms for Next-Generation Artificial Intelligence. Nonetheless, it is still very difficult for technology developers and application users to become infrastructure providers, particularly in the areas of AI chips, which entails a high accumulation of knowledge and capital. Even so, as the demand for specialized AI chips rises drastically, the U.S. semiconductor giants are unable to meet such a diversified demand and leave a growing niche market to Chinese domestic AI chip firms, such as Hisilicon, Cambricon, and Horizon (Interviewee 39, 44). Meanwhile, several technology developers and digital platforms have also developed their own AI chips for different AI technologies. Baidu, for instance, has announced its “Kunlun” series of AI chips and an open-source platform, PaddlePaddle, for autonomous driving (Interviewee 34). Alibaba has also developed its own AI chip and cloud computing capacities. Similar to Kenney and Zysman’s (2020) observations in the Western context, Chinese digital platforms are increasingly inserting themselves into most parts of the AI industry value chain, leading to a competitive landscape of platform ecosystems. As an entrepreneur observed, “This industry, if it does not belong to A, then it must belong to T or B; almost every new AI firm has some connections with BAT” (Interviewee 8). However, China’s application market is very segmented and is large enough to avoid disordered competitions among these players so far, and collaborations are still essential in many application scenarios, such as smart cities and autonomous driving (Interviewee 16, 42). 4.2.2 Relations between digital platforms and end-users Involving consumers in the production process is not new in the internet economy. R&D, manufacturing, and marketing are increasingly intertwined, and consumers are more involved in these processes and are even a source for new ideas, promoting the idea of the “prosumer” (Ritzer and Jurgenson, 2010). In regard to AI, the critical input of users has been changed from ideas and feedback to data. AI collects and analyses each end user’s data and provides personalized services. Every search, recommendation, and comment online, will become the analytical material for customized services, such as advertisements and content delivery. These data are co-created by the users and the producers, and users are automatically embedded in the production process. In this process, however, individual users do not have much say about the service they can receive, which instead is usually decided by the algorithms. The “black box” nature of AI algorithms further endangers the trust relations between users and technology owners (Interviewee 49). On the one hand, an increasing number of users are concerned about what data are being collected by AI companies and how it is being collected. On the other hand, they can hardly detect whether they are suffering from algorithmic discrimination. Some discriminations are caused by data bias, some are caused by machine self-learning, while human decisions cause other cases of discrimination. This opaque internal process leaves space for intentional interventions by the platforms. As a senior manager of an e-commerce platform observed in the following quote: “Indeed, consumers can barely do anything for the algorithm bias. You are helpless because it is a black box; you cannot see inside. If you want to verify, but the platform simply denies, who knows what is truly going on?……once you rely on the technology, what you get is not necessarily what you want”. (Interviewee 48) Digital platforms were supposed to break the information opacity to connect producers and users. With the application of AI, however, digital platforms are able to divide the whole market into more individual markets, separating the search results of each consumer, and the platforms become the only “insider”. There have been many news reports about digital platforms using AI to charge people differently, but investigation authorities have found it very challenging to collect sufficient evidence for these accuses. 4.3 Actor relations in the knowledge subsystem According to China’s AI Development Report 2018 (Xue et al., 2018), China has outperformed the United States in terms of the number of both AI paper publications and patents. Globally, universities are the main drivers in AI paper publications. Among the top 100 institutions in AI paper publication, 87 are universities. Nevertheless, we observe a trend in China that the industry, particularly large digital platforms, is becoming more important in AI knowledge development, while the role of academia is decreasing. This trend is also a global phenomenon, as many AI scientists are moving to the industry (Benaich and Hogarth, 2020). Typical examples include Google’s acquisition of DNNResearch, which was founded by the father of deep learning, Geoffrey Hinton, and the outflow of CMU’s driverless car research team to Uber. This trend is closely related to the nature of deep learning research, which survives on data and computing capacity, but both of them are mainly controlled by the industry. In China, a large number of top deep learning scientists are moving from academia to industry through either starting new businesses or joining AI companies. On the one hand, universities severely lack data and are not able to provide adequate computing capacity for AI research. Deep learning research is too costly for most universities. Not only does the purchasing of computing chips necessitate a huge amount of funding but the operation and maintenance are also great expenses (Interviewee 35). In addition, many university researchers are complaining that the academic performance evaluation system is excessively publication oriented, which could reduce their incentives for long-term innovation (Interviewee 21). On the other hand, China’s AI companies are actively recruiting academic talent to the industry with very attractive incentives. In China, PhD students majoring in AI are in short supply. AI companies have resorted to recruiting overseas talent, particularly Chinese AI researchers in Silicon Valley. As a typical example, Baidu hired the leading deep learning scientist Andrew Ng from Stanford University as its chief scientist in 2014. Many deep learning researchers were attracted to Chinese AI companies because of China’s rich data and application scenarios (e.g. Interviewee 5, 34, 55). Meanwhile, Chinese firms are learning from Microsoft and Facebook and are building many ivory-tower-style research labs, where researchers can conduct research and write papers with high levels of freedom. The aggregation of talent, data, computing capacity, and application scenarios makes the industry an increasingly important player in AI knowledge development. Many companies are encouraging researchers to publish scientific papers. Tencent’s and Netease’s AI labs, for example, have placed publishing research in top conferences as an important indicator of a researcher’s performance, and these companies reward them with high financial incentives (Interviewee 15, 21). As illustrated in Figure 3, in the top two computer vision conferences, the top AI companies are catching up with academics in terms of publication performance. Figure 3. Open in new tabDownload slide Chinese papers accepted by CPVR and ICCV in 2019. Note: CVPR, IEEE Conference on Computer Vision and Pattern Recognition; ICCV, IEEE International Conference on Computer Vision. Figure 3. Open in new tabDownload slide Chinese papers accepted by CPVR and ICCV in 2019. Note: CVPR, IEEE Conference on Computer Vision and Pattern Recognition; ICCV, IEEE International Conference on Computer Vision. Consequently, the mode of university–industry collaboration is changing. Academia is now more eager to seek collaborations with the industry because of data. In the past, it was relatively easier for university researchers to obtain access to industry data through collaboration, but this situation changed as firms became increasingly sensitive to data protection. As a university professor exemplified his collaboration with a large digital platform: “In the first year, the company had an initiative inviting scholars to use their data for research. In the second year, we could no longer have access to their data unless we conduct the research in its company. In the third year, the project just terminated. Because of the privacy issue of data, it is too sensitive.” (Interviewee 14) Nonetheless, collaborating with universities is still very important for the industry to undertake long-term basic research. As a top manager from Huawei explained, “The Moore’s law has reached a limit, and future computing frames need new breakthroughs in theories, so we need collaboration with universities” (Interviewee 62). The joint research lab is one of the most popular forms of collaboration. For instance, the Hangzhou government, Alibaba, and Zhejiang University established a public research lab together, which aims to advance the frontier research in the AI area (Interviewee 17). However, for many AI firms, collaborating with universities also conveys a more practical purpose. As a public relation manager from an AI firm said: “First, it is a requirement to cooperate with universities if we want to undertake a national research project. Second, through collaboration, we can build tight relationships with the university researchers, who might endorse our projects when needed” (Interviewee 18). Meanwhile, there has been concern among academia that the function of educating talent would also be weakened, as many top scholars have moved to the industry (Benaich and Hogarth, 2020). In contrast, the industry is increasingly taking a role in the formation of human capital. Large firms, such as Microsoft and Baidu, have tried to combine the function of both enterprise and college, educating AI talents in-house (Interviewee 34). In particular, Microsoft Research Asia (MSRA) has played a very pivotal role in cultivating top AI talents for China’s AI industry. Established in 1998, MSRA has focused on AI technologies such as voice recognition and image recognition and has accumulated world-leading research strength in these areas. MSRA has an open and free research culture and offers many internship opportunities for young researchers. A large number of China’s IT and AI entrepreneurs have working experience in MSRA (Interviewee 25, 33). Among them are many big names such as Kaifu Lee, Yaqing Zhang, Xiao’ou Tang, and Hongjiang Zhang. These talent spillovers played a very important role in the rise of China’s IT and AI companies, such as BAT, Xiaomi, JD, Byte Dance, Sense Time, and Magvii. BAT attracted most of the AI talents, but they also cultivated many talents for the industry. 4.4 Actor relations in the regulatory subsystem The regulations around data play a singularly important role in shaping the AI industry. Since the beginning of the internet economy, China has had a rather lax regime on data use. In recent years, privacy issues have risen with AI (Roberts et al., 2020). While the EU has implemented strict regulations on data ownership and uses, e.g. the General Data Protection Regulation (GDPR), China’s loose regulation continues into the era of AI, though it is currently under heated debates. Although AI is also associated with negative concerns regarding ethics, privacy, and employment, the general public in China has a very optimistic attitude towards technology and is more willing to accept AI technologies (Arenal et al., 2020). For example, Xue et al. (2018) found in their survey that only 2.4% of the respondents expressed objections to the development of AI in China. This lax selection environment is believed to have facilitated China’s prosperity in the AI industry. As many formal institutions associated with the burgeoning AI industry have not been fully established, the industry plays an important role in shaping them. On the one hand, unlike past catching-up pathways where Chinese governments can learn from the experiences of developed countries in industrial policies and governing related social or environmental risks, China has been one of the few forerunners in AI development, and there are few lessons that China can use for reference (Interviewee 46, 47). China has to cross the river by touching the stones itself in AI governance. Furthermore, the high complexity level of AI algorithms makes it difficult for experts, let alone politicians and ordinary people, to learn how pubic benefits will be affected. Taking online car-hailing as an example, the government lacks the data and technological tools to detect whether platforms use price discrimination to maximize their profits. Under such an uncertain context, the government faces substantial governing challenges, and it is evident that the traditional top-down approach in policy-making is no longer effective. As a secretary-general from China Internet Industry Association observed: “In China, the government owns the most power, and it is very often the most effective way to do things, but now the problem is the policy-making process is long, and the government cannot fully understand the frontier of industry development” (Interviewee 49). On the other hand, the industry is at the front of the innovations and has large real-time data and industry-specific expertise. Consequently, the government often needs to resort to the industry’s statistics, technology, and expertise for effective policy-making, and the industry is more deeply involved in policy-making processes and develops the capability to shape/direct policy-making. Indeed, many AI companies are well represented in the formation of national AI policies and planning. For example, at the National People’s Congress in 2017, Tencent’s and Baidu’s CEOs proposed enhancing AI research and using AI to find missing children, to reduce urban traffic congestion, and to promote AI integration with other industries. As another example, some industry players, such as Magvii’s founder and Kai-fu Lee, were deeply involved in the formation of China’s AI development principles, published by the Ministry of Science and Technology in June 2019. At the operational and technological level, the industry has more say in establishing standards. In many cases, there is a lack of existing standards in this fast-moving industry, and AI firms have to build their own standards and scale them up to the industry level (Interviewee 63). Meanwhile, several digital giants and unicorns have established various form of specialized public policy research institutes, aiming to study the emerging issues of the industry and to lobby for favourable policies and regulations. These institutes have hired many researchers and practitioners, such as lawyers, economists, and former government officials. Taking advantage of their rich data and technological expertise, and usually together with public research institutes or universities to gain more legitimacy, these institutes publish many research reports regularly on emerging issues, such as data governance, privacy, ethics, and employment. These initiatives further enhance the industry’s role in influencing policy-making and public opinion. As a public relation manager from an AI company illustrated: “On the frontier social issues, we collaborate with universities to correct some wrong public ideas and to guarantee our products can be accepted in the market…we need borrowing strength from government agencies and universities, to make more friends” (Interviewee 52). The industry has also developed several indirect and subtle ways to influence government regulations, especially through collaboration with academia. The data controlled by the industry not only attract researchers in science and engineering but also researchers in social sciences. Alibaba, for example, initiated a “Living Water Initiative” to invite social science researchers to study issues such as the e-commerce ecosystem and its social impacts. The platform assists the researchers by providing funding, opening access to certain data, and organizing research fieldwork into its business scenarios. The company also gathers influential scholars and awards excellent research from the initiative annually. In this way, the company builds a tight relationship with academia and delivers its impact subtly. For many, it is reasonable for digital platforms to have more say in public governance, as they are undertaking increasing public duties that used to belong to the government. Digital platforms can use their own data to analyse emerging social phenomena and make suggestions for government regulations. For example, DiDi, the largest player in China’s online car-hailing sector, initiated an online public council in 2018 to discuss emerging but complex governance issues, such as “could DiDi drivers refuse to take drunk passengers” (Interviewee 46). In 1 week, 269,000 people joined the debate on Didi’s platform, and 86% of them responded with a positive answer. In this way, the platform can keep following the latest governing issues and invite the public to offer proposals, based on which platform-level rules can be formed, and then ask for other rounds of feedback until the rules are widely accepted. As these platforms serve a large proportion of China’s population, their platform-level rules to some extent have a constraining force on the whole society. 5 Discussion and conclusions As data become an increasingly critical input for production, innovation, and regulation, we observe a shifting pattern of actor relations in China’s AI innovation system, with the digital platforms playing a more important role in each subsystem. In the business subsystem, firms along the value chain are vertically expanding into each other’s fields to gain data and application markets. This situation has resulted in a more competitive landscape of several large platform ecosystems. Although end-users co-produce with producers by generating personalized data, they have lost much ability to bargain with those platforms who control and utilize their data to create business value. Digital platforms can divide users into separated markets, and algorithms decide the content of the service provided. The scale of the economy of data has enabled a few digital platforms to become the dominant players in business value creation. In addition to these changes in the internal relations within the industry, its external relations with other actors in the AI SIS also shift. In the knowledge subsystem, AI firms are taking more of a role in knowledge development and the formation of human capital, while academia is left in a more dependent position. Although university–industry collaboration is still essential to the industry, the rationales begin to convey more intent to impact policy rather than simply to generate knowledge. In the regulatory subsystem, the traditional top-down approach in policymaking is seriously challenged by radical technological innovations, as well as the mismatch of responsibility and capacity in public governance. Controlling data and industry-specific expertise enables the industry to gain a more important role in shaping new regulatory frameworks rather than only being a consultant. Digital platforms also develop many direct and indirect channels to influence the changes of formal institutions, which in turn influence the value creation and knowledge generation in the innovation system. This article thus contributes to the IS literature in two aspects. The first contribution is to propose an actor-oriented analytical framework to examine the dynamics of actor relations in three SIS subsystems. The IS research has been criticized for being too structure-oriented at the expense of overlooking resources and actor strategies in shaping meso-level actor configurations (Markard and Truffer, 2008). In this article, we argue that actor relations should be examined in major system functions. Previous attempts have highlighted the role of actors in knowledge development and value creation, but few studies have examined the roles of different actors in the formation of formal institutions, which can significantly influence the former two activities. In particular, in emerging fields such as the AI industry, where the institutions are still in flux, actors can have much room to shape institution formation (Battilana et al., 2009). Therefore, we divide an SIS into three subsystems, namely, a business subsystem, a knowledge subsystem, and a regulatory subsystem, each with its main function and dominant actors. Such framing is helpful to examine more clearly the dynamics of the actors’ roles and relations in an SIS. Moreover, there is an interactive relationship between these subsystems, as the rise of an actor’s role in one subsystem could reinforce its competence in another subsystem. For instance, more value creation in the business subsystem enables the industry to have more financial capacity to invest in knowledge development and lobby for favourable regulations, which in turn facilitates the industry’s value creation. Such an actor configuration could lead to a realignment of technologies, user practices, norms, and institutions, resulting in a system-level transformation (Geels, 2004). This pattern of actor relations may go beyond the specific sectoral level and upscale to the national level, leading to a new triple helix configuration. The second contribution is that we reveal the role of data in transforming actor relations in data-driven innovation systems. It is certain that in different industries, the actor relations will appear in different forms depending on the knowledge base and institutional features. This analysis goes to the more fundamental level and investigates how the changes in the basic input of the innovation system—from physical materials to virtual data—affect the changing division of labour in the digital economy. In data-driven innovation systems, data are not only a critical factor of value creation but also a pivotal input for knowledge generation and better policymaking, which is far beyond the function of conventional resources. Consequently, we observe a more competitive relationship within the industry as firms compete to acquire more data and application markets. In the industry-university relation, the AI industry is substituting some roles of university in generating basic knowledge and cultivating talents, while university needs more collaboration with the industry to advance scientific research. Similarly, the government–industry relation is changing from a commanding one to a more collaborative one, in which the digital platforms become more active and influential in regulation formation due to their control of valuable data and frontier expertise. It is important to note that data alone cannot bring these transformations. Rather, it is the rise of data processing technologies such as IoT, big data, and AI that enables data to become a goldmine (Amoore and Piotukh, 2015). Though the government also has a large volume of public data, it lacks technological capacities and economic incentives to profit from them. The adoption of AI by digital platforms and traditional industries allows the value of data to be fully exploited, empowering digital platforms to become prime movers and game-changers in the emerging AI innovation system. The reach of these digital platforms is thus comparable to the giant petroleum firms in the oil era, but they can integrate into any businesses as long as telecommunication access is provided (Kenney and Zysman, 2020). The co-evolution between data and AI changes the existing cost structure, provides new space for innovation, and breeds new organizational modes, leading to a shift in techno-economic paradigms (Perez, 2010). The last issue relates to the extension of this observation to other contexts. It is acknowledged that an SIS transcends geographical boundaries and can have either a regional, national, or global reach. Therefore, an SIS in one country is not only confined by its national context but is also influenced by the global knowledge flow, actor networks, and institutions. China’s AI SIS is a part of the global AI innovation system and is subjected to the influence of the international institutional environment. The EU’s GDPR, for example, has caused a heated debate in China about how to use data, pushing China to implement stricter data regulations. In the knowledge subsystem, global open-source platforms and academic communities are still important sources of knowledge learning for China’s AI industry. At the national level, China’s NIS characteristics and institutions play a significant role in shaping the actors’ relations in its AI innovation system. Though AI development in China is largely market-driven and led by giant digital platforms, China’s central government and local governments play a pivotal role in scale-up AI experiments (Arenal et al., 2020). Admittedly, the findings of this article are conditioned by China’s context, but the general feature, causal power, and systematic impact of data may cross national boundaries. Indeed, similar findings that data and AI strengthen the capacity of giant digital platforms in value creation and knowledge development have also been observed in the Western context, especially in the United States (Benaich and Hogarth, 2020; Kenney and Zysman, 2020). While the rising role of data and digital platforms is pushing China towards a more “balanced” triple helix configuration, its impacts on other NISs may have different forms and extents, which is worth more attention in future research. To conclude, this research contributes to the innovation system literature by revealing how the change of the fundamental input of an SIS from physical resources to virtual data affects the changes of the actors’ roles and their relations. By examining three specific subsystems, we find that data are fundamentally different from conventional resources and controlling data will mean more influence in business value creation, knowledge development, and regulation formation. These changes will significantly reshape the evolution of an SIS and, to some extent, a country’s NIS. Surrounding the collection, processing, distribution, and utilization of data, new technology portfolios, business models, organization structures, industry ecosystems, and actor connections, may be built, resulting in new techno-economic paradigms. However, exploiting data is a trade-off between value capture and risk control (Foster et al., 2018). It is not clear for now how such changes could influence human society as digital platforms become more central in many parts of human society. After all, firms value profits more than public interests. Thus far, AI development is just in its very early stages. There are already many calls for regulation on the governance issues of AI at the micro-level (e.g. algorithm bias), meso-level (e.g. employment), and macro-level (e.g. democracy) (Haenelein and Kaplan, 2019). For future research, we need to develop better innovation system approaches to capture the digitalized, virtualized, and general-purpose features of data-driven innovations. It would be of much value to investigate the systematic impact of data in other data-driven industries and other geographical contexts. Future research could take one more step to examine how the interaction between data and digital technologies affect different types of power relations between actors, e.g. who has more or less power, who has power over whom, and who has what kind of power (Avelino and Wittmayer, 2016). It is also suggested to study how heterogeneous actors can develop a reasonable governance structure to address the social, economic, and ethical challenges brought by data-driven innovations. Funding This research was supported by National Key R&D Program of China [grant no. 2020AAA0105300], National Natural Science Foundation of China [grant no. 71810107004 and 71774097] and Tsinghua University Independent Research Program, NO. 2019THZW. References Adams P. , Fontana R., Malerba F. ( 2019 ), ‘ Linking vertically related industries: entry by employee spinouts across industry boundaries ,’ Industrial and Corporate Change , 28 ( 3 ), 529 – 550 . Google Scholar OpenURL Placeholder Text WorldCat Agrawal A. Gans J., & Goldfarb A. ( 2019 ). 'Economic policy for artificial intelligence', Innovation Policy and the Economy, 19(1), 139 – 159 . Amoore L. , Piotukh V. ( 2015 ), ‘ Life beyond big data: governing with little analytics ,’ Economy and Society , 44 ( 3 ), 341 – 366 . Google Scholar OpenURL Placeholder Text WorldCat Arenal A. , Armuñ a C. Feijoo C. Ramos S. Xu Z., & Moreno A. ( 2020 ). 'Innovation ecosystems theory revisited: The case of artificial intelligence in China' Telecommunications Policy , 44 ( 6 ), 101960. Avelino F. , Wittmayer J. M. ( 2016 ), ‘ Shifting power relations in sustainability transitions: a multi-actor perspective ,’ Journal of Environmental Policy & Planning , 18 ( 5 ), 628 – 649 . Google Scholar OpenURL Placeholder Text WorldCat Battilana J. , Leca B., Boxenbaum E. ( 2009 ), ‘ How actors change institutions: towards a theory of institutional entrepreneurship ,’ The Academy of Management Annals , 3 ( 1 ), 65 – 107 . Google Scholar OpenURL Placeholder Text WorldCat Baxter J. ( 2010 ), ‘Case studies in qualitative research’, in Winchester H. P. M., Rofe M. W. (eds), Qualitative Research Methods in Human Geography , 81–98. Oxford University Press: Oxford. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Benaich N. , Hogarth I. ( 2020 ), State of AI Report 2020, https://www.stateof.ai/. Bergek A. , Jacobsson S., Carlsson B., Lindmark S., Rickne A. ( 2008 ), ‘ Analyzing the functional dynamics of technological innovation systems: a scheme of analysis ,’ Research Policy , 37 ( 3 ), 407 – 429 . Google Scholar OpenURL Placeholder Text WorldCat Binz C. , Truffer B. ( 2017 ), ‘ Global innovation systems—a conceptual framework for innovation dynamics in transnational contexts ,’ Research Policy , 46 ( 7 ), 1284 – 1298 . Google Scholar OpenURL Placeholder Text WorldCat Chen S. H. , Lin W. T. ( 2016 ), ‘ The dynamic role of universities in developing an emerging sector: a case study of the biotechnology sector ,’ Technological Forecasting & Social Change , 123 , 283 – 297 . Google Scholar OpenURL Placeholder Text WorldCat Cooke P. , Uranga M. G., Etxebarria G. ( 2004 ), ‘ Regional innovation systems: institutional and organisational dimensions ,’ Research Policy , 26 ( 4–5 ), 475 – 491 . Google Scholar OpenURL Placeholder Text WorldCat Dey I. ( 2003 ), Qualitative Data Analysis: A User Friendly Guide for Social Scientists . Routledge : London . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Ding J. ( 2018 ), Deciphering China’s AI Dream: The Context, Components, Capabilities, and Consequences of China’s Strategy to Lead the World in AI . Faculty of Humanity Institute , University of Oxford :Oxford. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Etzkowitz H. , Leydesdorff L. ( 2000 ), ‘ The dynamics of innovation: from national systems and “mode 2” to a triple helix of university–industry–government relations ,’ Research Policy , 29 ( 2 ), 109 – 123 . Google Scholar OpenURL Placeholder Text WorldCat Fontana R. , Malerba F. ( 2010 ), ‘ Demand as a source of entry and the survival of new semiconductor firms ,’ Industrial and Corporate Change , 19 ( 5 ), 1629 – 1654 . Google Scholar OpenURL Placeholder Text WorldCat Foster J. , McLeod J., Nolin J., Greifeneder E. ( 2018 ), ‘ Data work in context: value, risks, and governance ,’ Journal of the Association for Information Science and Technology , 69 ( 12 ), 1414 – 1427 . Google Scholar OpenURL Placeholder Text WorldCat Freeman C. ( 1987 ), Technology and Economic Performance: Lessons from Japan . Pinter : London . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Geels F. W. ( 2004 ), ‘ From sectoral systems of innovation to socio-technical systems: insights about dynamics and change from sociology and institutional theory ,’ Research Policy , 33 ( 6–7 ), 897 – 920 . Google Scholar OpenURL Placeholder Text WorldCat Haenelein M. , Kaplan A. ( 2019 ), ‘ A brief history of artificial intelligence: on the past, present, and future of artificial intelligence ,’ California Management Review , 61 ( 4 ), 5 – 14 . Google Scholar OpenURL Placeholder Text WorldCat Kaplan A. , Haenlein M. ( 2019 ), ‘ Siri, siri, in my hand: who’s the fairest in the land? on the interpretations, illustrations, and implications of artificial intelligence ,’ Business Horizons , 62 ( 1 ), 15 – 25 . Google Scholar OpenURL Placeholder Text WorldCat Kenney M. , Zysman J. ( 2016 ), ‘ The Rise of the Platform Economy ,’ Issues in Science and Technology , 3 , 61 – 69 . Google Scholar OpenURL Placeholder Text WorldCat Kenney M. , Zysman J. ( 2020 ), ‘ The platform economy: restructuring the space of capitalist accumulation ,’ Cambridge Journal of Regions, Economy and Society , 13 ( 1 ), 55 – 76 . Google Scholar OpenURL Placeholder Text WorldCat Lane J. ( 2016 ), ‘ Big data for public policy: the quadruple helix ,’ Journal of Policy Analysis and Management , 35 ( 3 ), 708 – 715 . Google Scholar OpenURL Placeholder Text WorldCat Lee S. U. , Zhu L., Jeffery R., et al. ( 2018 ), ‘A data governance framework for platform ecosystem process management.’ in Weske M.ed), Business Process Management Forum . Springer : Cham . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lundvall B. ( 2007 ), ‘ National innovation systems—analytical concept and development tool ,’ Industry & Innovation , 14 ( 1 ), 95 – 119 . Google Scholar OpenURL Placeholder Text WorldCat Maciejewski M. ( 2017 ), ‘ To do more, better, faster and more cheaply: using big data in public administration ,’ International Review of Administrative Sciences , 83 ( 1_suppl ), 120 – 135 . Google Scholar OpenURL Placeholder Text WorldCat Malerba F. ( 2002 ), ‘ Sectoral systems of innovation and production ,’ Research Policy , 31 ( 2 ), 247 – 264 . Google Scholar OpenURL Placeholder Text WorldCat Malerba F. ( 2005 ), ‘ Sectoral systems of innovation: a framework for linking innovation to the knowledge base, structure and dynamics of sectors ,’ Economics of Innovation and New Technology , 14 ( 1–2 ), 63 – 82 . Google Scholar OpenURL Placeholder Text WorldCat Malerba F. , Montobbio F. ( 2004 ), ‘ Structural change in innovative activities in four leading sectors: an interpretation of the stylized facts ,’ Revue Économique , 55 ( 6 ), 1051 – 1070 . Google Scholar OpenURL Placeholder Text WorldCat Malerba F. , Nelson R. ( 2011 ), ‘ Learning and catching up in different sectoral systems: evidence from six industries ,’ Industrial & Corporate Change , 20 ( 6 ), 1645 – 1675 . Google Scholar OpenURL Placeholder Text WorldCat Markard J. , Truffer B. ( 2008 ), ‘ Actor-oriented analysis of innovation systems: exploring micro–meso level linkages in the case of stationary fuel cells ,’ Technology Analysis & Strategic Management , 20 ( 4 ), 443 – 464 . Google Scholar OpenURL Placeholder Text WorldCat Parkins D. ( 2017 ), ‘The World’s most valuable resource is no longer oil, but data,’ The Economist. Perez C. ( 2010 ), ‘ Technological revolutions and techno-economic paradigms ,’ Cambridge Journal of Economics , 34 ( 1 ), 185 – 202 . Google Scholar OpenURL Placeholder Text WorldCat Ranga M. , Etzkowitz H. ( 2013 ), ‘ Triple helix systems: an analytical framework for innovation policy and practice in the knowledge society ,’ Industry & Higher Education , 27 ( 4 ), 237 – 262 . Google Scholar OpenURL Placeholder Text WorldCat Ritzer G. , Jurgenson N. ( 2010 ), ‘ Production, consumption, prosumption the nature of capitalism in the age of the digital ‘prosumer ,’ Journal of Consumer Culture , 10 ( 1 ), 13 – 36 . Google Scholar OpenURL Placeholder Text WorldCat Roberts H. , Cowls J., Morley J., Taddeo M., Wang V., Floridi L. ( 2020 ), The Chinese approach to artificial intelligence: an analysis of policy, ethics, and regulation. AI & Society. 10.1007/s00146-020-00992-2 Rochet J. C. , Tirole J. ( 2006 ), ‘ Two‐sided markets: a progress report ,’ The RAND Journal of Economics , 37 ( 3 ), 645 – 667 . Google Scholar OpenURL Placeholder Text WorldCat Rumbold J. M. , Pierscionek B. K. ( 2018 ), ‘ What are data? A categorization of the data sensitivity spectrum ,’ Big Data Research , 12 , 49 – 59 . Google Scholar OpenURL Placeholder Text WorldCat Sarpong D. , Abdrazak A., Alexander E. A., Meissner D. ( 2017 ), ‘ Organizing practices of university, industry and government that facilitate (or impede) the transition to a hybrid triple helix model of innovation ,’ Technological Forecasting and Social Change , 123 , 142 – 152 . Google Scholar OpenURL Placeholder Text WorldCat Troisi O. , D’Arco M., Loia F., Maione G. ( 2018 ), ‘ Big data management: the case of Mulino Bianco’s engagement platform for value co-creation ,’ International Journal of Engineering Business Management , 10 , 1 – 8 . Google Scholar OpenURL Placeholder Text WorldCat UNCTAD ( 2019 ), Digital Economy Report 2019 Value Creation and Capture: Implications for Developing Countries . United Nations: Geneva . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Vaivode I. ( 2015 ), ‘ Triple helix model of university–industry–government cooperation in the context of nowadays social uncertainty ,’ Procedia Social & Behavioral Sciences , 213 , 1063 – 1067 . Google Scholar OpenURL Placeholder Text WorldCat Valkokari K. ( 2015 ), ‘ Business, innovation, and knowledge ecosystems: how they differ and how to survive and thrive within them ,’ Technology Innovation Management Review , 5 ( 8 ), 17 – 24 . Google Scholar OpenURL Placeholder Text WorldCat Watkins A. , Papaioannou T., Mugwagwa J., Kale D. ( 2015 ), ‘ National innovation systems and the intermediary role of industry associations in building institutional capacities for innovation in developing countries: a critical review of the literature ,’ Research Policy , 44 ( 8 ), 1407 – 1418 . Google Scholar OpenURL Placeholder Text WorldCat Weber K. , Truffer B. ( 2017 ), ‘ Moving innovation systems research to the next level: towards an integrative agenda ,’ Oxford Review of Economic Policy , 33 ( 1 ), 101 – 121 . Google Scholar OpenURL Placeholder Text WorldCat Weber S. ( 2017 ), ‘ Data, development, and growth ,’ Business and Politics , 19 ( 3 ), 397 – 423 . Google Scholar OpenURL Placeholder Text WorldCat Xue L. , Liang Z., Yu Z., et al. ( 2018 ), China Artificial Intelligence Development Report 2018 . China Institute of Science and Technology Policy at Tsinghua University :Beijing. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Yap X. , Truffer B. ( 2019 ), ‘ Shaping selection environments for industrial catch-up and sustainability transitions: a systemic perspective on endogenizing windows of opportunity ,’ Research Policy , 48 ( 4 ), 1030 – 1047 . Google Scholar OpenURL Placeholder Text WorldCat Yin R. ( 2010 ), ‘ Case study research. Design and methods ,’ Journal of Advanced Nursing , 44 ( 1 ), 108 – 108 . Google Scholar OpenURL Placeholder Text WorldCat Yu Z. , Gibbs D. ( 2020 ), ‘ Unraveling the role of green entrepreneurs in urban sustainability transitions: a case study of China’s Solar City ,’ Urban Studies , 57 ( 14 ), 2901 – 2917 . Google Scholar OpenURL Placeholder Text WorldCat Appendix 1. Profile of Interviewees © The Author(s) 2021. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - How data shape actor relations in artificial intelligence innovation systems: an empirical observation from China
JF - Industrial and Corporate Change
DO - 10.1093/icc/dtaa063
DA - 2021-02-24
UR - https://www.deepdyve.com/lp/oxford-university-press/how-data-shape-actor-relations-in-artificial-intelligence-innovation-UhWY4zFnr6
SP - 1
EP - 1
VL - Advance Article
IS - 
DP - DeepDyve
ER -