1 Introduction

Innovation is a central policy goal. In the data economy, the quest for more innovation is supported by a growing number of regulatory instruments. Widely varying in scope, such instruments do nonetheless paint a shared frame for the “innovation problem”: moving from the assumption that innovation and economic growth are deeply interlinked, the total welfare generated by the data economy increases with the rise of available data. Public policies, thus, are meant to solve the wicked problem of data access.Footnote 1 A case in point are the data-sharing provisions encompassed, for instance, in the General Data Protection Regulation (GDPR), the Open Banking Directive (PSD2), the Public Sector Information Directive (PSI), the Digital Markets Act (DMA), and the proposed Data Act. The hard truth inspiring policy-making seems to be that making more data available exponentially increases the chances of develo** novel, disruptive technologies. But policies pursuing sharing-led increments of innovation in the data economy are at great risk of fallouts. Widening data access requires a certain degree of cooperation among a broad range of economic players; at the same time, the combination of data entails risks for competition (Lundqvist, 2018). Regulatory provisions crucially impact the complex net of competitive and cooperative relations that characterize the digital economy. Cooperation and competition draw the lines of innovation trajectories. Hence, policies aimed at increasing innovation in the data economy may alter the qualities of such innovation.

Indeed, not all innovation is born equal. It is an empirically grounded assumption that private incentives to innovate have positive social repercussions; it is pairwise true that technological change appoints new winners and losers.Footnote 2 As the term itself suggests, in novare means to introduce novelty: new ideas, methods, products as well as new social roles, relations, and systems. To stimulate innovation means to feed such change. Redistributive policy instruments can adjust the payoffs of innovative processes ex-post. But change can rarely be reverted. For this reason, policymakers face a daunting challenge: not only should they ponder how to encourage innovation, but they also need to push change in the right direction. The regulation of the data economy comprehends both a quantitative and a qualitative dimension. Data-driven transformations need to be encouraged and channeled. How can policy foster the best possible innovative scenario?

This paper contributes to the debate on the regulation of data-driven innovation by framing the “innovation problem” in terms of innovation commons. The concept of innovation commons was retrieved from the institutional political studies of Ostrom (Hess & Ostrom, 2003; Ostrom, 1990, 2005) and became an established economic concept thanks to the work of Potts and Allen (Allen & Potts, 2015, 2016; Potts, 2018). In the dynamic context of the data economy, the political design of innovation commons sets the ground for future cooperation and competition. Ultimately, it determines the qualities of the resulting innovation. How should policy-makers design innovation commons? I suggest the answer to be context-dependent. Successful innovation commons channel existing incentives to cooperate and compete, and steer innovation in the legislator’s desired direction. Useful lenses for the scrutiny of digital contexts are offered by the study of business ecosystems. By focusing alternatively on the relations among players in different positions of the system it is possible to gain a multidimensional and functional understanding of the digital context. More specifically, the ecosystem metaphor supports the identification of two different economic functions for data: data as a resource and data as an infrastructure. These two data paradigms represent implicit assumptions driving policymakers in the design of innovation commons. The qualitative dimension of innovation depends on whether data is treated as a resource or as an infrastructure.

The attentive reader has already noted that the paper makes exhaustive use of metaphors. Metaphorical thinking may not appear as the first choice framework to analyze technologically-driven changes in society. As Ricœur elegantly observed, technical and poetic language are at two ends of a single scale (2003). While technical terminology is meant to be precise, clearly defining a univocal meaning to a specific word, analogies work by suggestion, evoking a similarity between two concepts without specifying the distinctive elements that make the comparison possible. Nonetheless, analogies and metaphors are often used to describe technological innovations. Computer desktops, artificial intelligence, and computer viruses are only some of the examples that relate the digital domain to the tangible world. By associating two seemingly unrelated concepts, the author transfers the set of implications commonly attributed to the subsidiary subject (the second term of the analogy) to the subject matter of the analysis. A desktop is organized and kept in order, intelligence allows inductive and deductive reasoning, and a virus spreads when two objects enter into contact. One single word substitutes a long list of attributes, presenting them cohesively and coherently.

Metaphors, in this way, allow us to “identify the similar in the dissimilar” (Black, 1955). Their role is not limited to embellishing an article by way of rhetoric figures: they have the power to redescribe reality, influencing how we approach a new concept. Analogies are not only a matter of language but also a matter of cognition (Vedder, 2002). They play a major role in the sha** of a narrative around the object of study. Narratives, in turn, have the power to deeply influence the behaviors of society, even more so if the unrealistic assumption of pure rationality which characterizes neoclassical economic models is abandoned (Shiller, 2019).

By influencing users’ and companies’ incentives to make their data available for re-use, narratives moving from different characterizations of data have a strong effect on innovation patterns. The direction of innovation in the digital sector is sensitive to the perception of players themselves, which often paints the frontiers of future possibilities. In the case of data, the heavy recourse to analogies highlights a common understanding: data are not simply representational resources (Gray, 2017). Metaphorical thinking draws “data imaginaries” and “data speak,” carving visions and rhetoric that try to encompass the molding effects data exercise on society. In this sense, metaphors and analogies contribute to sha** the data infrastructure. Metaphorical thinking (co)directs innovation.

Analogies also provide policymakers with mental models to rationalize the complexity of the environment. They determine the characteristics of the data economy that they deem relevant. Consequently, they drive the identification of legislative priorities. The characterization of data adopted by lawmakers influences the adoption of a protective or optimistic attitude toward data. It influences the balance that will be struck between protective measures aimed at safeguarding citizens from new harms and novel rules promising to unlock all the potentialities of data-driven innovation. This, in turn, determines the direction taken by investments in R&D. Eventually, how the technology is conceptualized decides the future developments of the technology itself.

The following sections build up a metaphorical framework for the design of innovation commons. Section 2 expands on the effects of innovation and society, and motivates the need for innovation commons. A first taxonomy of innovation is outlined in Section 3: cumulative, combinatorial, and generative innovation represent distinct goals and require ad hoc policies. Subsequently, the concept of business ecosystem is offered as a tool to read the context in which the regulator wishes to intervene (4). Section 5 identifies two (coexisting) economic functions for data in ecosystems and outlines the differences between the two perceptions of reality. An attempt to point at possible consequences stemming from the adoption of one perspective instead of another is presented in Section 6. Section 7 concludes.

2 Data-Driven Innovation Commons

Innovation benefits society.Footnote 3 The assumption is certainly simplistic: in reality, it is only true given several co-occurring conditions.Footnote 4 Notwithstanding the debates nuancing the statement, the EU pursues innovation as a social goal. The European “innovation agenda” was first compiled in the mid-1990s (Borrás, 2003). The term innovation entered the political arena substituting and expanding the previously prominent couple “science and technology,” to whom is dedicated Title XIX of the TFEU.Footnote 5 Since the articles contained in Title XIX replace their analogous in the TEC, it is safe to say that the pursuit of technological progress is a foundational goal of the EU.Footnote 6 Advances in technology are expected to promote competitiveness—and competitiveness fosters growth. However, reframing scientific and technological policy in terms of innovation policy marks a shift in the understanding of the problem. A new light is shed on the socio-organizational dynamics concomitant to the production of knowledge (Borrás, 2003). The pursuit of innovation is no longer an apanage of a single policy area: it is a transversal policy goal.

The EU growing concern for innovation is supported by the literature pointing at innovation as a major source of economic growth (Gilbert, 2006). Ignoring the reasons driving firms to invest in research and development (R&D), the outcome of such activities entails a positive effect on society. This perspective is confirmed by empirical evidence showing that the social positive return from R&D investments exceeds the private one (Griliches, 1992). A pivotal study in this field was carried out by Mansfield in 1977. Measuring both the private and social returns of seventeen industrial innovations, he confirmed previous literature results in appreciating a higher median rate of social returns compared to private ones.Footnote 7 Within his sample, he estimated a 56% rate of social return from industrial innovation. Additionally, he observed that private returns are characterized by extreme variability. Investments in innovation are risky. Lastly, in 30% of the cases, the private returns were so low that no firms able to make an accurate prediction would have invested in the innovation. Nonetheless, social returns were consistent. With the complete information about the success of the technology, it would have still made sense from a societal perspective to innovate; but firms’ incentives would have been null.

More recent research on social returns of innovation flags the multiple spillovers that stem from innovating activities. Hence, it is unlikely for R&D investments alone to drive all of the productivity gains. Jones and Summers highlight that productivity gains from innovation are driven by a wider set of innovative efforts (Jones & Summers, 2020).Footnote 8 Incorporating more variables into the analysis, they found out that the magnitude of social gains from innovation might have been consistently overestimated by models purely based on R&D investments. To estimate more accurately the average social returns from innovations, they consider the case in which gains from R&D pay off slowly, delaying the achieved benefits and thus reducing their present value. Moreover, they account for the non-R&D-related costs of innovation: among them, investments in capital assets such as equipment, machinery, and software. The resulting analysis, however, confirms that social gains from innovation remain large. The result is strengthened by the observation that other factors can lead to an underestimation of social returns of innovation when only R&D investments are taken into account: inflation bias, gains in health and longevity, and international spillovers are usually not considered. The conclusion is that, even under less simplified assumptions than the ones Mansfield relied upon, investments in innovation promote large average social gains.

Incentives to innovate, however, depend on the expected returns for the single firm investing in innovation. Private expected returns might be lower than social benefits—in this case, missing investments harm society more than firms. Such risk of underinvestment is coherent with a characterization of innovation as a commons. The term commons can indeed be used with two meanings: first, it may refer to a resource that is both rival and non-excludable; second, it can indicate the institutions that “govern the appropriation and provisioning of the resources among the community” (Potts, 2018).Footnote 9 Innovation commons, thus, jointly define the resources that, if shared with the community, would enable innovation, and the conditions under which such resources are shared. In this paper, data-driven innovation is assumed to be a common. This entails that (1) more data-driven innovation is desirable, and (2) the innovation problem can be construed as a problem of combined knowledge (Potts, 2018).

Data-driven innovation undoubtedly yields transformative effects on society. The term data-driven innovation defines the use of data and analytics to improve or foster new products, processes, organizational methods, and markets (OECD, 2015).Footnote 10 Machine learning techniques, artificial intelligence, and always-online interconnected objects are just some of the technological innovations made possible by data exploitation. The data economy continues to increase in size (MIT Technology Review Insight & Infosys Cobalt, 2021). Undisputedly, economic growth does not automatically translate into an even increase in well-being across all the strata of society. The altered social landscape brought upon by the recent rapid surge in the use of pervasive technologies is fraught with risks; the wealth generated by this flourishing sector of the economy may be unevenly distributed. Nonetheless, technological progress in the collection, storage, and analysis of digital information allows for the creation of new value (Buchholtz et al., 2014; OECD, 2015, 2017). It permits to augment the size of the pie to be shared between the participants in the economy. This adds to the direct positive effect of innovations in sectors such as health, science, and education, and to the benefits consumers receive in terms of increased variety (OECD, 2015).

Data drives innovation, but the mere existence of data does not necessarily bring about more innovation. The effective exploitation of the value creation potential of data demands, to begin with, appropriate technologies for their analysis and adequate knowledge management systems to ensure that the information they carry is not lost (Bresciani et al., 2021). Data science, machine learning, artificial intelligence (AI), and computing technologies empower data-driven innovation (Luo, 2023). Changes in how information is stored and made available for use, and in the technology adopted to exploit it, determine the progress of innovation. Ultimately, data flows draw the frontiers of the data economy. Data generation makes innovation possible; rules for data access and use define the direction of the innovation trajectories. The implications across society stemming from the distribution and use of data are wide-ranging (Sadowski, 2019). They dictate which players participate in the economy, the distribution of power among them, and which actors are simply left out.Footnote 11 In other words, it marks the characteristics of data-driven innovation.

Hence, the design of innovation commons is a necessary but daring task. New spaces for the combination of data-embedded information are key to facilitating the transmission of knowledge and fostering innovation. At the same time, the architecture of said spaces has long-lasting effects on the society of the future. Posing that the sharing of informational resources will produce innovation, and thus economic growth, how does the resulting society look like? The next section provides an overview of the three main modes in which data drives innovation, clearing the field for a successive evaluation of data-driven innovation’s socio-relational dimension.

3 Which Innovation for the Common Good?

The title of this section hints at the book “Economics for the common good” (Tirole & Rendall, 2017). In this paper, innovation is conceptualized in a similar way to what Tirole and Rendall did with economics: a fundamentally positive force, whose transformational potential comes with challenges that cannot be ignored. The authors define the “common good” in terms of general interest, which may be opposed to the interests of individuals. To identify the common good, one needs to place themselves behind a veil of ignorance and, pretending not to know their position in society, point out what is desirable. The notion builds on an intellectual tradition originating with Hobbes and Locke, continued by Rousseau, and more recently refined by Rawls and Harsanyi (Rawls, 1999, p. 2).

Among the many definitions that have been offered for the term innovation, one is particularly convenient to investigate innovation commons: innovation is “a new pattern of bits of information” (Macdonald, 1998). In the digital sector, such bits take a binary form. Data is, indeed, machine-readable encoded information (Zech, 2016).Footnote 13 New patterns of data enable data-driven innovation; they—to phrase it better—are data-driven innovations. Variations in such patterns may occur based on the rules governing their creation. Which instructions can be followed in the quest for innovation? The strategy adopted to explore the potentially infinite space of new information (i.e., the search space) substantiates diverging results. Players’ limited and value-infused perception of the search space infuses the choice of one strategy over another.

Of course, other factors count in the selection of any innovation strategy. A prominent place among them is reserved for the availability of data (and information), the costs and benefits of sourcing additional one, and more generally the fit with players’ overarching goals and declared mission. The analysis of the socio-relational dimension of innovation commons offered in Section 4 is intended to facilitate a deeper comprehension of these factors. Before plunging into the depth of the network dynamics of digital complex systems, however, it is fundamental to offer a taxonomy of the possible data patterns constituting the output of the economic players—and the subject of policy-making. Bits of information can prompt cumulative, combinatorial, or generative innovation (3.1). Data, specifically, can drive all three kinds of patterns (3.2): luckily, it is possible to identify a baseline regulatory goal capable of fostering innovation in all these forms (3.3).

3.1 Cumulative, Combinatorial, and Generative Innovation

Innovation takes multiple shapes, and is thus described in multiple ways. Setting aside its socio-relational dimension for the time being, the focus is here kept on technological innovation. A widely adopted definition of technological innovation is that of “a new or improved product or process whose technological characteristics are significantly different from before” (OECD, 1992). Focusing on the word “improved,” one can get the idea that innovation is a cumulative process: incremental changes follow one another, to the point that the resulting product/process is so different from the original one to be defined as new. Indeed, the accumulation of knowledge is a possible driver of innovation. The basic principle behind the functioning of carts and cars is the same and has been known to humanity since the dawn of time: a round object in movement dissipates less energy than an angled one. Over thousands of years, new knowledge was accumulated and applied to the cart. The development of the combustion engine determines the invention of cars; nonetheless, the functioning of cars capitalizes on all the knowledge cumulated by human experience with carts. The collection of information enables cumulative innovation. The more information is retained in a system, the higher the potential for growth (Winters, 2014, p. 22). Technologies provide solutions to specific problems, they define products or processes; domains are a collection of mutually supportive technologies. Each domain possesses a unique set of accumulated knowledge, practices, and mindset. They own their proper grammar (Arthur, 2014, p. 47).

Changes in the domain are the main way in which technology progresses. Sometimes, such changes originate from the illuminate minds of gifted individuals. Other times, it is chance that provides unique opportunities. More often, however, it is the spill out of information among different domains that enables the change. When unprompted change is driven by large, varied, and uncoordinated participants, innovation is said to be generative (Zittrain, 2005). Generative innovation is easier when the function of technological components is not well-defined (Arthur, 2014). The reason is simple: a kid playing with a red brick will come up with infinite roles for it; should the brick be given the shape of a car, the possibilities become limited. Similarly, multi-purpose technologies foster creative recombination (Murmann & Frenken, 2006).

Multi-purpose technologies are, indeed, technologies that possess a generative capacity. The concept of generative capacity was debuted in the context of social policy research by Schön and referred to the ability of metaphors to continuously generate new perspectives on the world by carrying familiar meanings in new domains (Schön, 1993).Footnote 15 Imported in linguistics, it is traditionally associated with the ability of alphabets to constantly generate new meanings through the recombination of sounds (Chomsky et al., 2006). The meaning of the term “generative” was retained when translated into innovation studies: here, generative capability came to define an overarching capability that enables continuous innovation (Guo et al., 2022). Generative innovation is unending and self-sustaining.

In some instances, the availability of information constrains the trajectory of innovation: the ebbs and flows of human technological progress are affected by the physical movement of individuals carrying expertise, ideas, and knowledge. In cyberspace, physical limits can be relaxed. It remains to be ascertained whether other limitations are in place in the digital world. Does the nature of data itself nudge to the promotion of cumulative, combinatorial, or generative innovation specifically? The question is bound to draw in respondents’ perceptions about the essence of data. Pending further assessments, however, the following Section 3.2 attempts to offer a first irrefragable answer on the potentialities of data for innovation.

3.2 Data Drives Just Anything

The effective management of innovation commons is contingent on the identification of the target type of innovation. Indeed, slightly divergent institutions encourage respectively cumulative, combinatorial, and generative innovation (West, 2009). An appropriate starting point for any policy evaluation stands in the recognition that data potentially enable all three of the above-mentioned innovation paradigms.

Data-driven innovation can be cumulative. At a basic level, the availability of large quantities of data continuously enables the discovery of novel insights. This appears to be the assumption behind the release by the Novartis Institute for Biomedical Research, in 2007, of an incredibly wide amount of data retrieved from the analysis of the genome of more than 3000 type 2 patients (West, 2009). Besides, the technologies adopted in data analyses improve cumulatively as well. The accurate targeting of advertisement services offered by Google or Meta, for instance, builds over decades of systematic data collection. The algorithm adopted by Netflix to provide personalized movie recommendations constantly improved over time as the company gained access to a wider audience. A quantitative study conducted in 2021 on innovation in AI used to mitigate and adapt to climate change showed that new AI patents in mitigation and adaptation technologies are associated with an exponential number of subsequent innovations (Verendel, 2023).

Combinatorial innovation can be data-driven, too. It is common practice among medical scientists, for instance, to mine literature and open data to facilitate diagnostic decision-making in cancer treatment (Ding & Stirling, 2016). Data-driven technologies are often combined: again in the medical sector, blockchain technology can be combined with machine learning to protect personal, highly sensible data collected by medical devices (Snow, 2021). The resort to combinatorial strategies to explore the space of new possibilities bears the considerable advantage of reducing the uncertainty intrinsic to the innovation process. By combining known components, inventors can sensibly reduce the variation in the expected success of their efforts (Fleming, 2001). Completely new components can lead to spectacular failures or triumphant breakthroughs; a combination of old components brings upon more modest but less uncertain results. The concept is exemplified plainly by innovative digital products resulting from the application of data-driven technologies to previously analogical domains (Hylving & Schultze, 2013). The knowledge accumulated in the domain of app development spills onto wearables as much as smart TVs, autonomous vehicles, or IoT appliances.Footnote 16 Full digital combinatorial innovation is possible too. In this sense, Application Programming Interfaces (APIs) are a keystone. Thanks to APIs and agile development methods, multiple services can quickly be integrated into a single-user application (Yildiz, 2022). Interoperability facilitates combinatorial innovation. The fungibility of data investments promotes inter-sectoral jumps. As a result, few big players controlling common APIs can expand in multiple sectors: the markets of competition blur, and new risks materialize (Sharon, 2021).Footnote 17

Last, the characteristics of the data economy facilitate generative innovation. Data-enabled technologies such as data analytics, data mining, Artificial Intelligence (AI), and the Internet of Things (IoT) can be easily transferred from one domain to another. Their decision problem is mostly defined in wide terms and can quickly adapt to the context. Machine learning, for instance, is used to tailor Netflix’s recommendation as well as to identify unknown influences among historical painters—and for uncountable other applications. Coherently, the data economy is a dynamic environment. Big data have a generative capacity (Scholz, 2017, p. 70). Generative algorithms are eyed as the actual more promising development in the data economy (World Economic Forum, 2023; Minevich, 2023). Generative innovation involves a potentially infinite amount of economic players, although strong variations can occur in their degree of awareness and the share of value they capture.Footnote 18 Petabytes of data provide information that answers unposed questions (Anderson, 2008). Whether or not this signifies “the end of theory,” as algorithms generate more insightful, useful, accurate, or true results than specialists crafting targeted hypotheses and strategies (Graham, 2012), it is undeniable that innovation is increasingly the result of inductive, rather than deductive, reasoning (Mazzocchi, 2015).Footnote 19

Consider, for instance, the evolution of watches. The knowledge accumulated in the domain of application development integrates the know-how of producers of watches and their physical components. In 1972, Hamilton released the first digital watch under the name Pulsar Time Computer. The product was a success, contributing to sha** social imaginaries and expectations about the future (Kent, 2021). It represents, mostly, a leading example of combinatorial innovation. The subsequent and frequent releases of new versions, updated only in the graphical interface, can undoubtedly be classified as cumulative improvements. Generative innovation only happens when wearables are integrated with software and operating systems. Technologies maturated in the context of smartphone development, applied to the hardware of a digital watch, gave life to an entirely new product with use not comparable with watches’. Wearables permit reading the time; a large share of fitness fiends among consumers suggest that measuring exercise and monitoring slee** time are more appealing than simply checking the time (The Economist, 2015). Mostly, the wide set of applications available to integrate the smartwatch offers novel and ever-changing uses.

In conclusion, data are little red bricks.Footnote 20 They can be piled up one over another to improve existing constructions; combined with a set of wheels they will turn into cars; they can generate several new exciting games whose limit only lies in the fantasy of the kid playing with them. Posit, however, that multiple children decide to join and make their own bricks available to create a more intriguing construction. They will surely need rules. In designing those rules, what kind of construction should be selected as a goal? More overtly, if data can drive cumulative, combinatorial, and generative innovation alike, which one should be the objective of policies for innovation commons? In this paper, I prudently approach this interrogative: before advancing with the analysis, the next Section 3.3 identifies a safe baseline for political and regulatory action.

3.3 Data Access, a Common Objective

Policymakers engaged in the regulation of the data economy face a fundamental question: would the pursuit of one kind of innovation hinder the evolution of another? Would initiatives supporting cumulative innovation, for instance, affect the evolution of combinatorial or generative innovation? In providing an answer to such a quest, the bottom line is that any institutional response to data-driven innovation commons should have an overall positive effect on society as a whole. In other terms, a policy should not provoke more damage than benefit. The uncertainty inherent in the regulation of new technologies risks being harmful in the long run (Anderlini et al., 2013). Analytical tools are needed to reduce uncertainty.

At a basic level, there is one policy goal encouraging all three kinds of innovation. Fostering data access and sharing is a fundamental enabler of data-driven innovation. Innovation builds on existent information. While single data points, taken alone, carry little information, data sharing is undermined by the risk of communicating information.Footnote 21 Sharing can be hindered as a consequence of what Kenneth Arrow described as the “information paradox”: a potential buyer of information cannot assess the value of the transaction before they receive the information itself, but if the seller were to reveal the content of the information to the buyer before concluding the contract, there would be no incentives for the buyer to proceed with the transaction, as he would already possess the information (Arrow, 1962, p. 19). Sometimes, however, information is spontaneously shared by the players in the economy. That is possible if they share a common goal. Allen and Potts studied information commons in the early process of collective pooling of information (Allen & Potts, 2015; Potts, 2018). When uncertainty is higher, and the possible innovation trajectories are almost boundless, information is extremely valuable.Footnote 22 According to their framework, as soon as innovation becomes established the need for cooperation is bound to decrease. Uncertainty over the innovation trajectory lessens. Competition begins to operate. Incentives to solve the innovation paradox shrink, eventually leading to a reduction in the amount of information exchanged.

Competitive dynamics carve innovation into one of the three above-mentioned shapes. As such, they also affect cooperation. Incentives for data sharing are linked to the competitive and cooperative relations existent in the digital economy: a possible framework for the analysis of such relations is presented in the next Section 4. By introducing the concept of the business ecosystem, the ambition of this paper is to offer the reader a pair of glasses to more clearly distinguish the need for institutional intervention in this extremely dynamic sector of the economy. Mostly, these lenses permit us to discern which innovation do data enable, and when. Commons could thus be managed to favor, respectively, cumulative, combinatorial, or generative innovation. A context-dependent intervention, I argue, necessitates tools to understand the context. The concept of ecosystem is instrumental to root up the relational dynamics of the data economy.

4 Ecosystems, Loci of Innovation

The development of business ecosystems is the organizational backbone of digital-based innovation. Ecosystems enable what Benkler defines as commons-based peer production (Benkler, 2002), a third mode of production alternative to markets and firms especially frequent in the digital economy.Footnote 23 Peer production can be described as “a process by which many individuals, whose actions are coordinated neither by managers nor by price signals in the market, contribute to a joint effort that effectively produces a unit of information or culture” (Benkler, 2003, p. 1256). Commons-based peer production, in the context of the digital economy, depends on the aggregation of independent firms which autonomously “scour their information environment in search of opportunities to be creative in small or large increments” (Benkler, 2002, p. 376). Such exploration is based on data sharing: by accessing new data firms retrieve new information, reduce their level of uncertainty, and undertake innovation ventures. Note that aggregation implies a certain degree of cooperation. Cooperative relations are nurtured to better respond to continual competitive threats. In the digital sector, innovation is the engine of competition (OECD, 2022).Footnote 24 Cooperation, in this challenging arena, becomes a competitive instrument (Teece, 1992). Cooperative data-driven innovation is faster and more apt to respond to ever-evolving threats (Petit & Teece, 2020). The complex net of cooperative relations that rises around major players’ technologies constitutes the structure of business ecosystems.

The picture of technology-intensive machines voraciously analyzing exceptionally wide datasets recalls sci-fi imaginaries rather than organic biological ecosystems. Still, the term ecosystem is part of the academic jargon of multiple fields interested in the data economy: from business, management, and innovation studies to computer sciences, it is rapidly spreading to the legal and economic literature. In particular, competition law scholars are advancing the idea that ecosystems allow to grasp and systematize the multi-dimensional nature of competition in the digital sector (Jacobides & Lianos, 2021; Petit & Teece, 2020; Robertson, 2021). More than that, the term made it to legal texts (Digital Markets Act, Digital Services Act) and Court decisions (Google LLC and Alphabet, Inc v European Commission, 2022). In this paper, business ecosystems are defined as comprising firms that collectively offer value to customers, independently setting their business strategy but strongly connected one with another. Independence and interdependence are both necessary but insufficient conditions. Economic players independently designing their products are part of marketsFootnote 25; full interdependence between outputs is achieved in hierarchical organizations such as firms or conglomerates. Ecosystems, like commons-based peer production systems, stand in between “hierarchies” and “markets” (Benkler, 2003; Gawer, 2014; Jacobides et al., 2018).

Although it is possible to identify examples of business ecosystems that originated as early as the 1920s, the managed business ecosystem as an organizational form is connected to the computer industry of the 1960s. James Moore identifies two major shifts that played a pivotal role in its affirmation (Moore, 2006). The first was the development of the family of computers IBM System/360.Footnote 26 Thanks to a modularized architecture, IBM was able to offer several variations of the same product able to accommodate the needs of different market segments, without the need to develop and maintain multiple product lines (Liu, 2016). The modularized architecture allowed for the development of complementary markets for specific parts of the computer (Moore, 2006). Modules made it possible to launch an extremely complex product on the market, at the same time assuring that it could easily evolve to accommodate changes in demand. The second paradigmatic shift identified by Moore as generative of the business ecosystem as a managed form of business organization was operated in the same period by HP. While IBM established a new technical paradigm, based on modularized interoperable architecture, HP laid the foundation for a new cultural paradigm. The company’s internal organization was grounded on collaboration. Small groups of engineers would cooperate on specific projects and flexibly re-arrange themselves at their conclusion. The organization was based on open and loose groups, among which information flew relatively freely (Burgelman et al., 2017). The collaboration of autonomous individuals promotes creativity (Benkler, 2003).

Modularized architecture and collaborative culture are, indeed, distinct features of digital ecosystems. Chiefly, digital ecosystems permit the extension of those paradigms beyond the borders of the firm. Complex innovations are developed as combinations of distinct modules. Specific product design choices help firms to generate product families and lead to systematic, quick innovation through the use of common assets (Gawer, 2014). Innovation is boosted by specialization. In each module, knowledge is accumulated. As such, new firms can access the market by providing new complementary solutions that can be integrated into the main product. At the same time, module recombination allows for multiple possibilities for combinatorial innovation. It is by recombining the components that Apple can accommodate the needs of all the market segments covered by its personal computers offer. The collaborative culture that made HP successful in a highly technological sector is frequently adopted by firms belonging to the same ecosystem. Unless they are fully complementary, companies in the ecosystem alternate competitive and cooperative relations based on time, market, and functions.Footnote 27 They are said to “coopete” (Brandenburger & Nalebuff, 1998).

Autonomous and asynchronous innovation can be conducted by multiple firms using the same resource. At the same time, access to resources may provide a competitive advantage. Data pooling and sharing are influenced by inter- and intra-ecosystem coopetitive dynamics. On the bones of business ecosystems, data ecosystems take shape. An interesting strand of the literature on the regulation of data is devoted to the identification of technical solutions capable of fostering open data ecosystems or explicitly facilitating data reuse by supporting collaborative networks.Footnote 28 The object of these studies is constituted by spontaneous or privately managed data-sharing platforms. It is, however, important to note that data and business ecosystems are not overlap**.Data ecosystems enable data commons; business ecosystems to innovation commons. By focusing on the latter, the article intends to flag the potential rather than the actual data commons: the ones that could happen, if regulation is successful. Moreover, the focus is kept exclusively on the data commons that foster innovation. Indeed, the structure of business ecosystems determines the architecture of innovation.Footnote 29 Regulation on data access affects the information available to ecosystem members. Thus, they influence the trajectory of innovation. How it happens are hard to predict due to the complexity of ecosystems’ structures. The next part (4.1) attempts to analyze the distribution of information in digital ecosystems.

4.1 Information-Bound Innovation Trajectories

Digital ecosystems are complex systems (Briscoe, 2010). The number of publications dedicated to the understanding of the digital economy might make this assertion sound trivial. But puzzled researchers of the digital era might feel reassured recalling the definition of complexity provided by complexity science: the complexity of a system is related to the amount of information necessary to describe it (New England Complex Systems Institute, n.d.). The description of digital ecosystems requires more information than the description of traditional markets. The concept can be better understood through an example. Consider a hypothetical “grandparent test”Footnote 30 on the environment of operations of a hotel and the environment of operations of its closer digital equivalent, Airbnb. A traditional hotel purchases toiletry sets, breakfast products, and cleaning services from its suppliers, and provides a room for the night to its clients. The operations of Airbnb involve a significantly wider number of players (Fig. 1). Additionally, note that the relationship among them is more varied compared to the traditional market. Contracts and purchases are no longer the main way of interaction. Players that are not in direct communication can nonetheless be highly interdependent (Shaughnessy, 2019). The amount of information needed to describe Airbnb activities is significant.

Fig. 1
figure 1

Source: own elaboration based on Shaughnessy, 2019

The Airbnb Ecosystem.

Full size image

The main characteristic of complex systems is that small changes in one of the parameters can produce large changes in the aggregated behavior of the system (Petit & Schrepel, 2023). This can be easily understood by referring to a peculiar example of complex systems: the atmosphere. In 1962, the meteorologist Edward Lorenz observed that a butterfly’s flap in Brazil could cause the formation of a tornado in Texas.Footnote 31 In the context of business ecosystems, the butterfly effect explains the unpredictability of innovation trajectories. The course of technology is influenced by what Arthur referred to as “small historical events” (Arthur, 1983).Footnote 32 Apparently, random choices in the early stages of development of a new technology become cemented in the technological structure of the economy. Arthur states that “micro-events become magnified by positive feedbacks; their cumulation decides the outcome and forms the causality” (Arthur, 1983). The anticlockwise hands in the 1433 clock displayed in Florence’s cathedral testify that casualty ultimately is entrenched in conventions.Footnote 33 “History becomes destiny” (Arthur, 1983, p. 16).

Strong path dependencies undermine the applicability of the neoclassical “rational agent” assumption.Footnote 34 Participants in the system are extremely bounded agents: their decisions are heavily dependent on their starting point, and the effects of such decisions are determined by a net of interdependencies of which they are likely not fully aware. Ecosystem complexity is made manageable through modularity (Baldwin, 2007; Moore, 2006). An organizational structure is modular when it is composed of elements, i.e., modules, that independently perform distinctive functions (JK Gershenson et al., 2003; Simon, 1962). It tends to emerge in large systems characterized by a high number of interdependencies (Simon, 1962; DL Parnas, 1972; Ethiraj & Levinthal, 2004). The separation into modules allows for the creation of sub-systems. Participants in each sub-system are closely connected to one another and loosely related to participants in other sub-systems. Within modules, the unit of analysis is limited to a reduced amount of interactions: hence, complexity is reduced. Conventions and standards dominate exchanges among the different modules.

Modularity is the key to agile innovation. If technological paradigms continuously shift, how can firms build resources and capabilities that sustain competitive advantage? Modularity allows parallel work to proceed independently. Chiefly, multiple components of a complex product can be innovated at the same time. The case of IBM/360 illustrates the advantages of a modular product design for innovation. First, autonomous innovation takes place within components. New, updated data entry units could be released anytime, as long as they respected the measures of the console. Research and development are independently carried on by each component. More generally, players producing alternative modules compete with each other. In dynamic environments, competition is based on innovation (OECD, 2022). A modular structure, by simplifying complexity, reduces the information necessary to adduce incremental improvements. Moreover, reverse engineering and imitation are made easier. Cumulative knowledge and joint problem-solving within modules are incentivized (Pil & Cohen, 2006). Thus, we could expect higher rates of incremental innovation.

Second, combinatory innovation takes place by exchanging and replacing lower-performance modules with higher-performance modules. In the IBM/360, exploitation of combinatory innovation gave rise to multiple models, each of whom was better suited for different typologies of users. The relative ease with whom it was possible to recombine the essential elements of the computer gave birth to a flourishing market of non-original substitutes. Peripheral products could be attached to the System/360 processor thanks to its standard interface. Third-party suppliers and manufacturers quickly entered the new market. Consequently, competition increased. A modular architecture facilitates the development of new markets, encouraging value creation.Footnote 35

Lastly, the modularized architecture of digital ecosystems enables generative innovation. The early production of the IBM/360 can be considered a case of a closed ecosystem. When ecosystems are open, that is to say when the membership can be acquired by any third party capable of annexing their product to the system’s complex offering, the boundaries become more malleable (Um et al., 2013). Innovation can advance by harnessing the distributed creativity of heterogeneous players (Yoo et al., 2012). The generativity of open ecosystems “comes from the variety of plug-ins of different kinds,” whereas closed ecosystems can only rely on a variety of modules of the same kind (Um et al., 2013; Yoo et al., 2012). The generative capability of the Android ecosystem, for instance, is given by the virtually never-ending diversity of the third-party applications that can run on it. An Android smartphone can turn into a training device, a music player, or even a metal detector. The openness of the system benefits from product-agnostic modules, such as Google Maps APIs, which can be integrated into a multitude of different products. Open systems count more members; more members translate into increased complexity. For this reason, open ecosystems often proliferate around the figure of a leader: hierarchy is a powerful way to manage complexity (Simon, 1962).Footnote 36 Undoubtedly, leaders detain a substantial advantage over the other members: they can increase variations in their products and raise the overall flexibility of the system without incurring high transaction costs (Um et al., 2013). By designing the architecture, they can easily steer innovation trajectories. Their control over innovation is, however, not absolute. Generative innovation relies on exchanges and unforeseeable contacts, and increases with the flow of information within and outside the ecosystem.

The relative position of economic players, the openness of connection between modules, and the overall architecture of the system affect the characteristics of innovation in the data economy. The level of competition ecosystem’s members are most sensible to, together with the perceived rivalry of data, determines the incentives to share data. Mostly, remember that the economic players aggregated in an ecosystem are independent. Ultimately, their competitive strategy determines what kind of innovation they want to pursue. As their competitive strategy depends on which information they have, and considering that data embed such information, access to data determines the resulting innovation. For this reason, the next Section 5 concentrates on the possible roles that data can assume in ecosystems. Different (perceived) data functions determine the willingness to access data commons and the strategic choice of the participants in the economy. The resulting innovation will be qualitatively different. Finally, when policies for innovation commons endorse a specific data function, they implicitly support specific qualitative features of innovation.

5 The Dual Role of Data

The context in which innovation commons take shape comprises both the external forces fostering or hindering collaboration and an individual calculation of the costs and benefits of sharing. Incentives to cooperate, thus engaging in data commons, and compete, i.e., excluding opponents from the commons, depend on the perceived value of data. The latter is, in turn, linked to players’ perceived economic function of data. The modular architecture of digital ecosystems permits the identification of two different economic functions for data. Each function facilitates a specific kind of innovation. The next paragraphs focus on the perceptions that data points are ecosystem resources shared according to players’ incentives to cooperate or compete.

Section 5.1 examines two analogies that describe different declinations given by the literature to data as resources. In particular, Sub-section 5.1.1 focuses on the analogy of data as commodities; Sub-section 5.1.2 presents data as common pool resources. Access to resources enables cumulative and combinatorial innovation. But data is also an infrastructure for digital ecosystems: this is the subject of Section 5.2. The relational nature of data, and the embedded generative potential, serve the economic players to organize the world. The infrastructural role of data enables generative innovation, which has a greater chance of having disruptive effects on society. As illustrated in Section 4, the ecosystem-mediated relationship among the firms participating in an innovation common determines the characteristics of the resulting innovation. As a consequence, different functions for data can prevail at different levels of the ecosystem. Policies aiming at fostering innovation commons are expected to appropriately select the narrative they adopt towards data and match it with the role of the firms that are expected to partake in the commons and the characteristics of the desired innovation.

5.1 Data as a Resource

Data points are the immaterial but fundamental inputs of the data economy. Two analogies represent the relational implications of data as an economic resource: when data is equated to commodities (Sub-section 5.1.1), a light is shed on its competitive consumption; when data is liked to common pool resources (Sub-section 5.1.2), attention is brought to the importance of cooperation.

5.1.1 Commodities

The metaphor that had the most powerful grip on public opinion is certainly one that associates data with a commodity. “Data is the new oil” is a sentence that rapidly surged to the position of workplace litany, when not a company mantra, to the point that it is considered by many a tired cliché (Gilbert, 2021). In highlighting the fundamental role of data in fueling companies’ growth, the metaphor pairs a descriptive claim with a normative one. Neglecting data management becomes the equivalent, for a company, of forgetting to fill the tank of the car. Data is freed by its technical aura and normalized, becoming part of firms’ daily operations as raw materials, with their supply taking a central role in the development of the business strategy. Oil is seldom substituted by gold, highlighting the value that the resource has for businesses. The other facet of the same medal is the analogy equating data to a currency that consumers spend inadvertently. This metaphor provides a tentative explanation for the rapid rise of Big Techs, firms whose business model does not usually involve direct payment by the consumer. The mystery of how those firms could produce value while offering a free service is thus quickly solved: users do not pay with money but with data. Data is liked to a currency.Footnote 37 It provide companies with a resource whose value they are unaware of, and Big Tech companies able to extract data from users can resell it by making an immense profit. An illustration published on the cover of The Economist in 2017 perfectly represents the metaphor. Big Techs are drawn in the guise of oil platforms, a wordplay based on their status as online platforms. The implicit admonition contained in the analogy warns against the free and unconscious transfer of valuable goods that accompanies many digital actions.

The success of the data is the new oil, gold, or currency metaphors can be due to their ability to provide a characterization of data that reflects and explains their behavior as observed by individual users and professionals in their daily experiences. However, its deconstruction reveals that it moves from a series of implicit assumptions, transferring to data a set of non-trivial economic properties. Data is a scarce resource that has to be extracted, possesses commercial value and is fungible. Moreover, the amount of data at their disposal determines businesses’ competitive strength: no market player will be willing to make its data available to other firms (Graef, 2016). As a matter of fact, due to the high fungibility of data, companies operating in the data economy can quickly expand their activities in different markets. This represents a considerable obstacle to data sharing. Firms considering its data a commodity are not likely to make it accessible for re-use. The risk they incur is a loss of competitive advantage.

The economic function of data is (or is considered to be)Footnote 38 the one of commodities when firms are subject to competitive threats. In this regard, it is important to observe the multidimensionality of competition in the digital ecosystems. Static, price-based competition can take place horizontally among complementors offering substitutable products; vertical intra-ecosystem competition refers to value captured through joint collaboration; innovation-based competition takes place between different ecosystems that offer comparable value added to customers through the provision of multiple products. The ecosystem theory reaffirms that the assessment of firms’ competitive advantage cannot overlook the analysis of the aggregate level. In any case, the extraction and accumulation of data having the function of a commodity support cumulative evolution within the boundaries of the module—be it the single firm or the ecosystem vis-à-vis competing ecosystems.

5.1.2 Common Pool Resources

Multiple scholarly contributions on the data economy equated data to a common pool resource.Footnote 39 Data is non-rival: consumption by one actor does not prevent re-use by others. In addition, it has limited excludability: it is common practice to limit data access through the adoption of technical barriers. Data can be kept a secret. Technical solutions meant to control third-party access are commonly adopted by companies operating in the data economy. And, if it is undeniable that cybersecurity attacks can undermine the efficacy of such solutions, it is pairwise true that so far the large data holders such as Google, Meta, and others “do not seem to suffer from a vast copying and leaking of their huge amount of collected data” (Kerber & Schweitzer, 2017). Goods characterized by non-rivalry and limited excludability are considered to be impure public goods (Leach, 2003). Thus, the most appropriate framework for the analysis would thus be the one of common pool resource (CPR) elaborated by the Nobel laureate academic Elinor Ostrom, who in 2003 co-authored a seminal paper studying information as a CPR (Hess & Ostrom, 2003). Innovation is possible when data governance can overcome collective action problems associated with data access.Footnote 40

The CPR perspective focuses on the social need for firms’ cooperation. Data that holds little to no value for the firm that held their access are often left abandoned in data lakes or data swamps. If such data was to be shared, new firms could use the same data points in a different domain. Thus, combinatorial evolution could surge.

5.2 Data as an Infrastructure

The last analogy equates data to an infrastructure. In the same way of roads and electric systems, data comprises “a backbone for much of modern social and economic activity” (Ruhaak, 2020). Considering the multitude of downstream innovations made possible by the use (and re-use) of data, its wide availability will foster technological progress and increase social welfare. Data can be considered a multi-purpose resource (OECD, 2015). The value of data in contexts different than the one in which it is originated may be difficult to assess, given that the value of information is context-dependent, but it is most likely positive (OECD, 2015). The infrastructural angle raises attention to the transformative effects that data-driven innovation entails on the economy and society. Relatedly, the emergence of such characterization of data is more recent: it moves from the recognition that the disruptive changes brought by the quick technological advances in the data economy affected a multiplicity of actors, many of whom in unforeseeable ways. As such, the analogy sheds light on the multiplicity of technological trajectories typical of the digital economy. The value of an infrastructure varies depending on the activity that it enables. When data is liked to infrastructures, then, the focus is not on their use value nor their commercial value: the policy discourse is centered around their potential value (Ducuing, 2020). Brett Frischmann proposes a three-step test to define an intangible as an infrastructural resource: it should be non-rivalrous in consumption to some appreciable extent; social demand for it shall be driven primarily by downstream product activities that require the resource as an input; it shall serve as input for a wide range of goods, be they commercial or non-commercial. Infrastructures are “used by many different users, with the usage evolving over time, as may the type of users” (Frischmann, 2012). The combination of the three criteria results in the definition of intellectual infrastructure as “non-rival inputs for a wide variety of outputs” (Frischmann, 2012). The OECD underlines that such a definition perfectly describes the nature of data (OECD, 2015). Indeed, the value from data is created subsequently to their transfer and in relation to reuse.

The economic function of data as an infrastructure potentially enables the flourishing of generative innovation. Infrastructures are a way in which interconnected systems can be conceptualized (Henfridsson et al., 2013). In digital ecosystems, data connect the modules with the rest of the ecosystem. But data also transmits elements of the social context in which they have been co-created. As Star and Ruhleder write, “an infrastructure occurs when the tension between local and global is resolved” (1996). The data to which players in the ecosystem have access determines the technological landscape that they can explore; at the same time, the distribution of data in the ecosystem defines the connections with the other economic players. Gray renames the data infrastructure “data worlds”: his invite is to not consider data exclusively as resources, but to investigate how political, social, and cultural values emerge from data infrastructures (Gray, 2017).Footnote 41 Data worlds provide the horizon of intelligibility; the information to which each participant in the ecosystem has access defines its ability to move in the world. Ultimately, they decide the direction of innovation of the ecosystem and, as such, the well-being of society.

Data worlds, according to Gray, offer transnational coordination (Gray, 2017). Data as infrastructures, indeed, assist the governance of the ecosystem. They are an instrument for the ecosystem leader in its quest to maintain alignment of complementors’ interests. Ecosystem leaders influence the architecture of the ecosystem through the design of their products. Usually, this includes a certain influence on the definition of standards and rules for interoperability. This directly affects the allocation of data, sha** the modules of the ecosystem; in turn, size and relations between modules strongly affect innovation trajectories. The starker the influence of the leader over the ecosystem’s infrastructure, the more power will it have over the orchestration of data resources. The higher chances of successful generative innovation materialize when data infrastructures are open, participative, and dynamic.

The same data serves the function of resource and infrastructure. The perceptions of the players, together with contextual use-related factors, determine the economic players’ attitude towards cooperation through data commons. The same metaphors guide regulators’ interpretation of the complex and dynamic digital economy. It helps them understand which function does data hold. If data is perceived as a commodity, a CPR, or an infrastructure, the legislative focus will be drawn to different ecosystem levels. The kind of innovation encouraged by the regulatory intervention will thus differ. Metaphors are abstract guides for the interpretation of reality but entail material effects. The three above-mentioned metaphors for data address a need from a public policy perspective. Paraphrasing the OECD, they “provide a framework that can guide policymakers in identifying when data warrant their attention” (OECD, 2015, p. 178). The next Section 6 attempts to sketch the consequences stemming from the adoption of one metaphor other than another; further research is needed to disentangle the implicit assumptions driving policymakers in the design of innovation commons.

6 Policy Implications

Data enables innovation. How it happens is mediated by business relationships in ecosystems. The analysis presented in this paper invites us to adopt a more granular view of the data economy. Cooperation through data commons has different costs and foreseen benefits depending on whether data are treated by ecosystems’ members as resources or if they represent an infrastructure. Taking an ecosystem perspective assists in identifying areas and modalities of intervention able to leverage existing incentives. Different options of data governance emerge as possible. Policies aiming to foster innovation commons shall take into account the structure of the business ecosystem they intend to address to (1) be effective, and (2) prevent unexpected long-time effects on the ecosystem governance, affecting multiple actors (Cennamo, 2021).

Different regulatory solutions may foster innovation commons in digital ecosystems. When data has the role of a commodity, the starting point of the legislative discussion lies in the recognition that data hold commercial value. Therefore, the regulator adopting this point of view is likely to increase transparency in the market, so that consumers are fully aware of the economic value of the data they are creating. Relatedly, the legal framework is charged with the task of facilitating the emergence of a healthy and well-functioning market for data. Although a certain degree of concentration among data providers appears as unavoidable, directly stemming from the characteristics of the commodity extracted and supplied, the role of legal institutions driven by the willingness to foster commons for commodity data is to ensure access to data to all the companies operating in downstream markets. Those companies will be enabled to use them to fuel the provision of new and better services, generating growth and innovation. If the grip that the analogy holds among managers and the public is reflected in the policymaking, we can expect the legislation to focus on the commercial value of data, providing incentives for their exchange in huge quantities, considering the fungibility they have across different markets and ultimately ensuring that all the companies transforming them in goods and services have access to this essential commodity.

The GDPR contains a curious case of the data-as-commodity metaphor. Art.20 of the Regulation establishes the right to “data portability,” giving the data subject the right to receive the personal data concerning her “in a structured, commonly used and machine-readable format” and to transmit those data to another controller. This particular article appears to pursue a different goal to the rest of the GDPR: while the remainder of the regulation builds the foundation of a fundamental right to data protection, art. 20 seems to be guided by the desire to push growth and competitiveness. One may go as far as arguing that, by facilitating the transfer of data from and to competing controllers, data portability could assist incremental innovation. Each player could rely on the same resources and independently pursue their innovation strategy. However, the GDPR does not contain any indication that supports the trade of such a precious commodity.Footnote 42 The result is an almost forgotten, never enforced, and arguably ineffective provision.

The data-as-CPR analogy designs a clear priority for the regulator that adopts it: to foster data sharing by increasing control over data. The attention is drawn to issues of data undersupply due to collective action problems determining an inefficient use (and re-use) of data. The regulator will aim at facilitating data production and reuse by correcting economic agents’ failure to spontaneously negotiate the optimal level of the good. The assignment of clear and defined control over data may be considered an unavoidable step. A specific declination of this view is offered by Birch when he refers to the assetization of data (Birch et al., 2020). The author problematizes innovation as being increasingly driven by the pursuit of rents, and proposes laws that protect (personal) data subjects with time-limited property rights.Footnote 43

The data-as-CPR analogy appears to have guided the European Commission in the drafting of the Data Governance Act, adopted in May 2022. The objective of the Regulation is the stimulation of innovation through the establishment of a clear framework for data reuse and sharing across specific sectors. Among the many provisions, several regard the so-called “data-altruism”—which is, the voluntary disclosure of personal data by data subjects. For example, Art. 17 regulates public registers of recognized data altruism organizations. Data is considered to be a necessary input, and the regulator has the duty to overcome the organizational and technical obstacles that impede innovation commons. However, access to data is provided only to certain pre-determined categories of economic players. As such, the resulting innovation can only be cumulative or, at best, combinatorial.

Lastly, when regulators recognize data as the infrastructure of the ecosystem, they will design institutions to make it available for use in a non-discriminatory way. They will have to address a problem connected with the general-purpose nature of infrastructures: as the value that will be produced through infrastructures cannot be known ex-ante, public policies should ensure that they are sufficiently produced (OECD, 2015). All the interested third parties could then profit from the resource, unlocking a wide and unpredictable range of downstream activities. Regulation towards infrastructure should be particularly mindful: data, in this declination, govern and enable the relationships among members and design the space in which they operate. The data infrastructure contributes to sha** ecosystems’ incentives, facilitating cooperation, and making generative innovation possible. Ecosystem leaders influence such infrastructure by governing the ecosystem and designing its architecture. Standards-setting and interoperability multiply the relations among the ecosystem members and govern their contacts with external (competing) ecosystems. Regulation underpinned by the infrastructure metaphor is likely to promote new standards and mandate interoperability. In the context of ex-post enforcement, considering data’s role as an infrastructure leads to measuring Big Tech’s impact (also) against the relational impact that it holds.

The recognition of the infrastructural function of data is quite recent. However, it is already possible to uncover instances in which the European regulator, more or less consciously, has looked at the market through the lenses of this metaphor. It is the case of the access-to-account rule (XS2A) contained in the Open Banking Directive (PSD2). The rule mandates incumbents (usually traditional banks) to disclose information on users’ accounts to third-party providers (prior authorization of the users themselves). The rationale behind it is explained by the legislator’s intention of facilitating the entrance of new agents.Footnote 44 By improving the level playing field for payment service providers, the XS2A rule does not only improve conditions for the entrance of banks’ direct competitors, but it promotes the flourishing of payment initiation services and account information services. Consumers could thus benefit from an infinite range of new products: new apps for managing expenses complement the banks’ product, new banks can more easily secure consumers or completely new services can arise. Ex-post, it appears that the latter case has been the most favored by the Directive. The PSD2 enabled, in particular, the blooming of “PayTech” companies (Polasik et al., 2020): a plurality of niche-targeted and varied services that offer customers previously inexistent tools. In this sense, the XS2A rule seems to have favored the emergence of generative, intra-module, innovation.

Through the XS2A, the European regulator intends to foster innovation by making data available not only to operators in downstream markets for which it represents a necessary raw material but also to newcomers who can use such a resource in ways that the directive is unable to foresee or restrict. This is consistent with the narrative that equates data to infrastructure and advocates that institutions should make sure that it is made available for use in a non-discriminatory manner. This way, all the interested third parties could profit from the resource, unlocking a wide and unpredictable range of downstream activities. However, it needs to be noted that no unique standard was defined for payment service providers’ APIs. The European Banking Authority was designated to draft the regulatory technical standards indicated in the legislative act, which were subsequently approved by the Commission. The definition of standards is a fundamental step for the success of any legislative intervention which aims to foster data infrastructures: the inappropriate design of APIs would have undermined the favorable outcome of the directive (Borgogno, 2019).

Standards play, indeed, a major role in defining the architecture of modularized systems. They influence the thickness of the transaction points, hence incentives to entry. The introduction of an alternative standard, determining a different organization of modules, is a major push for intra-ecosystem competition. The introduction of disruptive innovation is made possible by architectural shifts. As such, standard-setting intended to foster innovation and competition shall necessarily move from the definition of which kind of innovation and competition represents their goal. Incremental, combinatorial, or generative innovation? Horizontal intra-ecosystem competition, vertical intra-ecosystem competition, or inter-ecosystem competition?

Further research is needed to better delineate the policy implications stemming from the adoption of a metaphor other than another. Table 1 summarizes the findings enucleated in this section.

Table 1 Characterizations of data as commodities, public goods, and infrastructure
Full size table

But the examples presented above merely scratch the surface of the complex legal analysis needed to establish which function of data is recognized behind a legislative act. Additionally, although the regulation of the digital economy is a vibrant and rapidly evolving branch of the law, the majority of the legislative interventions have been enacted too recently to draw more than initial considerations on their effects. In this article, I offered a theoretical framework that awaits further testing. The next Section 7 offers some preliminary conclusions.

7 Conclusions

Data connects digital firms in a complex net. Its understanding is challenging: it requires taking into account the multiple horizons of intelligibility that coexist in the production (and fruition) of a single product. However, such disentanglement represents a necessary step to fostering cooperative innovation. The exchange of (data) resources, together with the smooth coordination of independently achieved technological progress, promises to advance society’s well-being in the form of generative innovation. The promotion of data-driven innovation is with good reason a goal of European policies for the digital sector. But the regulation of digital markets shall include, as a preliminary step of the legislative intervention, the outline of the kind of innovation that is intended to achieve. By altering the distribution of data in the ecosystem, regulation may modify the incentives to compete and cooperate. As such, a context-dependent approach shall be favored. Different dimensions of competition (horizontal inter-ecosystem, vertical inter-ecosystem, intra-ecosystem) may be dependent on different kinds of innovation (cumulative, combinatorial, generative). Further research is needed to examine the inevitable trade-offs among them, and the balancing actions available for government intervention. Ultimately, the regulation of the data economy should rest on the acknowledgment that the future trajectory of innovation depends upon the lenses through which the complexity of data worlds is cognized. Setting the rules means having a voice in the narrative guiding the development of technology, a narrative that, in conclusion, will be chorally co-created by the many participants of digital ecosystems.