The organization of digital platforms Architecture and interfaces in a partial organization perspective

Digital platforms quickly rose to be a pervasive feature of contemporary life as they impact various domains. While researchers agree that digital platforms come in diverse forms and perform various functions, the platform debate lacks a general concept capturing their common underlying features. Considering digital platforms from an organizational perspective, we extend on seminal insights and develop the concept of platform organization. We argue that platform organizations share a common basic architecture and rely on digital technology to create a social order. According to our argument, digital platforms consist of a formal organization at the core using digital interfaces to govern the activities of participants in the platform periphery. We posit that this periphery manifests a specific social order following from a particular configuration of the five elements of organization outlined by the partial organization approach. Our concept of platform organization allows researchers to abstract from domain-specific functions and investigate underlying commonalities and differences of digital platforms. We illustrate our argument by drawing on three salient examples from the current debate: Uber, Twitter and Wikipedia. Applying the partial organization approach reveals that all three platform organizations implement a decided order using digital interfaces, while the social order they each enable varies by a degree of organization.

The growing number of empirical studies, usually focused on single platforms, analyse specific features of their case but rarely generalize their findings beyond the platform's application domain (Bossetta, 2018;Kleis Nielsen & Ganter, 2018;Nielsen, 2018). Researchers have developed a variety of schemes that categorize platforms inside (Vallas & Schor, 2020) or across domains (Ansell & Miura, 2020;Dolata, 2019a), yet these schemes are usually tailored to highly specific research questions and derive their criteria for sorting from platform functionalities. They neither provide a coherent set of relevant dimensions for platforms in general nor allow researchers to analyse the diversity of platforms systematically. Other researchers generalize platforms as technical infrastructures that enable various forms of digitally mediated interactions (Helmond, 2015;Plantin et al., 2018). The focus of such contributions lies on the consequences of a typical set of technologies on interactions (Bucher & Helmond, 2016), organizations (Vergne, 2020), or society as a whole (van Dijck et al., 2018). In all these approaches, researchers underestimate the fact that each digital platform encompasses its own social order. This order is heavily reliant on, but not equal to its technological infrastructure. While researchers agree that digital platforms come in diverse forms and perform various functions, the platform debate lacks a general concept to capture their common underlying features.
Considering digital platforms from an organizational perspective, we see important commonalities: All major examples from the domain-specific discourses involve a formal organization that runs the platform. Also, digital platforms often put extant organizations under pressure to change established modes of operation (Davis, 2016). However, many scholars ignore such continuities between formal organizations and digital platforms, highlighting instead those features of platforms that relate them to markets or networks. Two prominent approaches currently advance discussions on the organizational properties of platforms: One approach investigates platforms as business models whereas the other conceptualizes platforms as governance mode.
Researchers regarding platforms as business models have developed elaborate conceptual tools to analyse how markets change through the technology-driven rise of this new form of organizing supply and demand (Sanchez-Cartas & León, 2021) or innovation processes (Gawer, 2020;Langley & Leyshon, 2017;Sturgeon, 2021). While this approach has emphasized the specific technological properties of 3 digital platforms and portrayed the relationships emerging from their use, it remains heavily focused on firms' strategies. As this research analyses how firms employ platform technology to compete and innovate in markets, it tends to disregard the rich societal underpinnings that are known to enable market relations (Granovetter, 1985). Researchers also struggles to integrate non-profit cases that lie beyond markets, such as Wikipedia (Benkler & Nissenbaum, 2006;Lakhani & Wolf, 2003). While many salient platforms clearly rise and thrive in market environments, digital platforms are not confined to the market domain. We can therefore assume that they are not solely instruments of firm strategies but represent a more general type of social order.
The idea that digital platforms constitute a general type of social order is the common denominator of a second research approach. Its adherents try to understand platforms as governance mode (Dolata, 2019a;Flyverbom et al., 2019;Grabher & König, 2020;Kornberger et al., 2017;Kretschmer et al., 2020;Stark & Pais, 2021). Aiming to fit platforms somewhere between, or on par, with established forms like network, markets and hierarchies, researchers in this stream emphasize the embeddedness of platforms in society with its manifold problems of coordination. Here, researchers treat digital platforms as novel modes of governance that help to coordinate actors in various societal domains.
The narrow focus on societal features, however, underestimates how much the specific technological properties of platforms matter for coordination (similar point made by Faraj & Pachidi, 2021).
This state of research leads to a situation where the two most prominent approaches often focus either on platform technology in terms of firm strategies and markets or on highly general forms of social order without a particular focus on platform technology. However, especially organization scholars need to better understand how technology becomes "constitutively entwined" (Faraj & Pachidi, 2021) with social order if they want to provide insights into contemporary and future changes of organizing brought on by digital platforms (Davis & Sinha, 2021). Currently, the debate lacks the analytical tools to systematically investigate the commonalities and differences of platforms.
We approach this problem from a partial organizational perspective and assume that digital platforms constitute a specific form of organization. We develop the concept of platform organization to capture the characteristic features of this organizational form. We argue that platform organizations share a common basic architecture and rely on digital technology to create a social order. According to our argument, digital platforms consist of a formal organization at the core using digital interfaces to coordinate the activities of participants in the platform periphery. We posit that this periphery manifests a specific social order following from a particular configuration of the five elements of organization outlined by the partial organization approach (Ahrne & Brunsson, 2011). Our concept of platform organization allows researchers to abstract from domain-specific functions and investigate underlying commonalities and differences of digital platforms. We illustrate our argument by drawing on three salient examples from the current debate: Uber, Twitter and Wikipedia. Applying the partial organization approach reveals that all three platform organizations implement a decided order trough digital interfaces, while the social order they each enable varies by a degree of organization. 4 2. The architecture and social order of digital platforms Taking an organizational perspective and extending on seminal research, we propose to capture the general properties of platform organizations with the design pattern of the platform architecture and combining this view with the insights of the partial organization approach on diverse social orders.
2.1 The platform architecture: core, interfaces and periphery In the technology and innovation management literature, the term "platform" denotes a specific form of modular technology that enables companies to connect to external actors in new ways (Gawer, 2009b;Tiwana et al., 2010). It also describes a business model where firms try to attract external actors using their platform technology to generate value for the firm (Langley & Leyshon, 2017;Thomas et al., 2014). Researchers in this field investigate how technology design or strategic decisions influence the success of the business model (Gawer, 2020) or how the success of this business model influences traditional markets (Sturgeon, 2021). In this stream of research, the study of platforms predates the current debate about the consequences of digitalisation (Gawer, 2009a).
Early on, the term "platform" has been used to define technologies as well as organizations (Thomas et al., 2014). According to Baldwin and Woodard (2009), both kinds of systems can exhibit the same architecture. A system has a platform architecture if it contains a core of tightly coupled elements, a periphery of components loosely coupled to the core, and interfaces to connect the loosely coupled parts of the platform. The interfaces define the rules of interaction between core and periphery as well as those between the components in the periphery.
The periphery of platforms is highly dynamic: components can enter and exit the system or change their internal structure with relative ease. The stability of platform systems results from the core, which provides basic functionality for the system, and from the interfaces, which integrate core and periphery. Interfaces codify a small set of obligatory features for all interactions in the system. All components must abide by the interface definitions to be able to connect reliably to the core. In this respect, interfaces constrain the variability of components where the specified aspects are concerned.
However, components are free to vary in respect to everything else. By setting transparent conditions for connectivity, interfaces thus facilitate the diversity of components beyond the small set of explicitly defined constraints.
Architecture, interfaces and components are established concepts in platform research. As yet, researchers almost exclusively use them to describe platform technology (Bossetta, 2018;Gawer, 2020;Tiwana et al., 2010). The distinction between the basic building blocks of the platforms' social order, and consequently the existence of the common pattern of organization, is often ignored. Instead, researchers run the risk of conflating the platform with (aspects of) the core organization (Dolata, 2019b;Gawer, 2009b;Langley & Leyshon, 2017;Rahman & Thelen, 2019) or with the activities at the periphery (Davis, 2016;Viégas et al., 2007;Weller et al., 2014). We propose to apply the terms of the architecture concept to describe the structural pattern that underlies the social order on digital platforms. To capture the details of this social order we resort to the partial organization perspective. 5 2.2 Partial organization: elements of decided social order A major approach that engages with the social orders brought about by digital platforms draws on governance concepts to understand problems of social coordination (Grabher & König, 2020;Stark & Pais, 2021;Vallas & Schor, 2020). Traditionally, this research tries to ascertain whether patterns of coordination conform to networks, markets or hierarchies (Powell, 1990). In the debate on digital platforms, several researchers have highlighted cases that show a high similarity with well-known forms like networks (Grabher & König, 2020), markets (Rochet & Tirole, 2006), organizations (Kornberger et al., 2017;Mair & Reischauer, 2017) or communities (Dolata & Schrape, 2016).
The established governance concepts have shed light on salient features of such empirical cases, yet only partially captured the whole phenomenon. Some researchers have started to extend the collection of established forms to include new governances that rely on digital technology (Dolata & Schrape, 2016;Kretschmer et al., 2020;Stark & Pais, 2021). However, although traditional categories fail to adequatly capture the charateristics of platforms, some researchers simply add a new form and conceptualize "Platforms as Platforms" (Stark & Pais, 2021, p. 53). In contrast, others argued (Schüßler et al., 2021) that platforms neither fit into a well-known governance category nor clearly represent a single new type. Rather, platforms could consitute novel configurations of traditional types or fluidly move between known categories.
In this ambiguous situtation we turn to a theoretical framework that can cope with equivocal forms of social order and delineates phenomena that lie between the neatly organized terminologies of traditional governance approaches. Ahrne and Brunsson (2011) develop the concept of partial organization to distinguish forms of social order based on decisions from others, where order emerges from informal relationships or institutions. In contrast to those emergent forms of order, decided orders result from conscious and deliberate choices of human actors. As such, they emphasize uncertainty, assign responsibility, and invite contestation.
Here, formal organizations present the most obvious, but not the only form of decided order. To identify forms of decided order outside of formal organizations, Ahrne and Brunsson define five elements of organization: membership, rules, monitoring, sanctions, and hierarchy. Membership identifies the insiders and outsiders of a social order. Rules define specific expectations which members are supposed to follow. They also establish ways to monitor how well members conform to these expectations, and positive or negative sanctions for (a lack in) conformity. A hierarchy exist if some members are singled out to hold others responsible to the rules. Since regular formal organization exhibit all five elements they can be considered "complete organization", in contrast to cases that exhibit only some of the five elements. Hence, the perspective of partial organization can also relate decided and emergent aspects of social order (Laamanen et al., 2020). Employing the elements as a general taxonomy (Apelt et al., 2017) allows researchers to transfer insights from organizational theory to other application domains (Ahrne et al., 2016). 6 2.3 Defining the platform organization: Architecture and social order To define our concept of a platform organization we now integrate the insights from the platform architecture approach and the partial organization literature.
Following Baldwin and Woodard (2009), we regard platform architecture as a general design pattern that can be applied to organizations as well as to technology. We propose that the organizational models of all digital platforms conform to a platform architecture (Ametowobla, 2020). From this organizational perspective, a digital platform organization is enabled by a core, i.e. a formal organization operating a computer system which forms the backbone of the platform. We call this core the platform organizer. In many salient cases the platform organizer is a company, like Twitter, Inc. or Uber Technologies. Sometimes, it is of another organizational type, e.g. a voluntary association like the Wikimedia Foundation. The digital platform organization also comprises a periphery made up by external components, e.g. users or organizations, that engage with the platform to interact with each other. These components are merely loosely coupled with the platform core (Kirchner & Beyer, 2016): They can easily join or leave the platform periphery and do not act as employees or directly contracted agents of the core.
The relationships between the core and the periphery of a platform organization are governed by interfaces, e.g. interactive features of websites or smartphone apps. Interfaces regulate the inputs and outputs of the platform system and thereby essentially specify how components of the periphery can relate to the core and to other components, which forms of interactions are possible and what data is collected and returned. The digital interfaces enable and inhibit certain activities on the platform, e.g.
because inputs need to conform to defined formats or because certain outputs are simply not possible.
With this regulatory capacity interfaces can also define multiple roles in the periphery by assigning different labels, restrictions or capabilities to different user groups. In this way, interfaces define the standards that govern the technical capabilities as well as the social behavior at the platform periphery (general point made by Brunsson & Jacobsson, 2000).
A digital platform organization is thus formed by the amalgamation of two different social orders, arranged in a platform architecture: The core is made up by a complete organization, comprising e.g. employees, formal rules and hierarchical structures. This complete organization at the core directly controls the interfaces. The periphery lacks the fully fletched features of a complete organization as e.g. its members do not sign employment contracts and remain only loosely coupled to the core.
Empirical studies of digital platforms in various domains have shown that the social order of the platform periphery can be captured with the partial organization approach (Kirchner & Schüßler, 2019;Nielsen, 2018). These studies have highlighted that the peripheries of digital platforms can resemble one established type of social order, e.g. a market (Kirchner & Schüßler, 2019) or exhibit features of several social orders at the same time, e.g. networks and organizations (Nielsen, 2018).
Activities at the platform periphery are not merely appendix to the core but form an arena of social relationships in its own right. Through interface design, the platform organizer decides on central features of the periphery. Interfaces define or at least influence strongly who can take part in interactions at the periphery (membership), what interactions are possible or legitimate (rules), how rule-followers and rule-breakers can be distinguished (monitoring) and treated accordingly (sanctions), and which participants can hold the others accountable to the platform rules (hierarchy). The main functions of the platform are generated by software, and all participants connect to the platform through websites or apps. The platform organizer develops and operates the digital platform technology, embeds the interfaces in it, and thereby enables activities and governs interactions at the platform periphery.
While the interfaces embedded in the platform technology are designed by the platform organizer and as such are the result of decisions, other aspects of the social order of the periphery emerge through user activities and interactions. In this respect platform organizations always build on an infrastructure of decided elements that allows for the unfolding of emergent elements in the course of technology use. This interplay between decided and emergent elements relates to research on gradations and ambiguous states of organizational patterns, that some described with terms like "degrees of organization" or "organizationality" (Ahrne et al., 2016;Kirchner & Schüßler, 2019;Grothe-Hammer et al., forthcoming;Dobusch & Schoeneborn, 2015). The social order at the platform periphery is thus enabled by the digital platform technology, but remains distinct from it. Platform organizers decide which aspects of the platform periphery they design through interfaces and which aspects they leave up to emergence, but they do not fully control the results and thus the emergent aspects of the social order at the periphery.
Against the backdrop of this concept of platform organization, we focus our complex conceptual argument using four core insights: (1) Digital platforms in various domains conform to the general architecture of a platform organization, where a core and a periphery interrelate via interfaces. (2) By designing interfaces, the core governs activities and establishes a social order in the platform periphery.
(3) This social order in the periphery follows from a configuration of the five organizational elements as defined by the partial organization approach. (4) Some platform organizers rely more on decided elements of organization to govern the periphery while others leave more up to emergence, meaning that digital platforms can be more or less organized.
In the following sections we illustrate our argument by presenting three seminal cases of platform research as platform organizations.

Considering Uber, Twitter and Wikipedia as platform organizations
To indicate how our concept can help to uncover commonalities and differences of diverse digital platforms, we use it to describe three prominent digital platforms: Uber, Twitter, and Wikipedia. These three exemplary cases stand for the diversity of functions digital platforms fulfil in different domains.
Each of our examples represents a different variety of the platform organization. We show that our concept helps to systematize the differences between these varieties and recognize as-yet undiscovered commonalities. 8

Uber
The mobility platform Uber connects people in search of transport (for themselves, their food, or other stuff) with people who can provide it, and arranges the whole process from pickup to payment. The platform is most successful in the field of personal transport services and offers ride-hailing in more than 60 countries (Mazareanu, 2020). By providing a cheap and convenient alternative to traditional taxis, Uber has disrupted the sector of inner-city transport profoundly, albeit more in the United States than in other countries (Thelen, 2018). It is one of the most controversial providers of on-demand labour (De Stefano, 2015;Vallas & Schor, 2020), and a paradigmatic example in predictions about the future of work (Rosenblat, 2018) and organizing (Davis, 2016).
Core organization. Uber Technologies is a US corporation founded in 2009 (Hall & Krueger, 2018).
According to its own definition, the company offers private car-owners a means to provide paid rides and thus a highly flexible way to earn money (Hall & Krueger, 2018). It thereby acts as an intermediary between transport service providers and consumers. Uber Technologies does not own the cars used for Uber rides, nor does it pay for any of the costs incurred.
Interfaces. Uber is accessible via mobile apps which track their users via GPS and display their position on a map, along with the current fares in different zones of the city. Drivers and riders can rate members of the other group and see each other's rating. Riders can use their app to order a drive, watch it approach, and pay. Drivers can use their app to accept ride requests and see the position of a rider they have accepted. Many of the core functions of the platform are automated, e.g. the matching of ride requests to drivers, the calculation of the overall rating, or the deduction of fares after rides and their transfer (minus commission) to drivers' accounts. Beside these straightforward functions, the apps build on elaborate algorithms processing user data automatically to determine details of the apps' functionality. Most prominent among them is surge pricing, which determines fares depending on current demand. All of these algorithms influence the interactions on the platform strongly (Heiland, 2020;Rosenblat & Stark, 2016;Schmidt, 2017;Schor et al., 2020).
Periphery. The periphery of Uber resembles a market, with clear distinctions between participants who supply and those who demand services and elaborate mechanisms to match both sides. There are only few ways to differentiate among drivers or riders. All interfaces are designed to enable and facilitate standardized transactions between the groups and to increase their amount. They do not support direct interactions inside groups or means to help participants single themselves out.
Membership. Uber has two complementary types of members: Riders need a mobile device to install the Uber App and a means to pay online (Uber, 2021d). Prospective drivers need a car meeting predefined quality criteria, a driving licence, and have to pass a background check (Rosenblat, 2018).
Even though the threshold for Uber membership is thus higher for drivers than for riders, these prerequisites are still low compared to alternatives (Gavin, 2020;Gawer & Srnicek, 2021). The platform is only open to members who are free to use it as much -or as little -as they want.
Rules. Most of the platform rules are explicitly defined, either directly in the terms and conditions each user has to accept to become a member or in the myriad of user information pages referenced by this 9 document (Uber, 2021c). They are also almost completely embedded into the platform technology.
Uber Technologies can change these rules unilaterally without consulting its members (De Stefano, 2015) and members are forced to accept the changed version after login before they can use the app (Rosenblat & Stark, 2016).
Monitoring. The platform combines three forms of monitoring: automated tracking, user rating, and self-monitoring. Members are tracked via GPS, so the platform stores their current position, the route and speed of rides, and generally their pattern of movement. All other interactions on the platform, e.g. each drivers' earnings or hours online, are also stored. After each ride, drivers and riders are expected to rate each other's performance on a scale of one to five, and the platform calculates an average score for each participant. Both forms of monitoring data are input for algorithms that create elaborate statistics for drivers. These statistics act as tools for self-monitoring (Rosenblat & Stark, 2016).
Sanctions. Drivers and riders use ratings to sanction each other's behaviour during the ride. The average rating score calculated from these individual ratings is the basis for centralized sanctions: If participants' score falls below the minimum, they are excluded from the platform. These sanctions apply to both groups of members (Uber, 2021b), but while drivers are sanctioned automatically, consequences for riders are rare. The core organization changes the thresholds for these sanctions in accordance with its business strategy (Hall & Krueger, 2018). Uber also uses tracking data to sanction drivers, e.g. if drivers accept too few rides while online, or decline too many accepted rides once they have seen the details, they can be excluded from the platform (De Stefano, 2015).
Hierarchy. Uber's interfaces are designed to create a buyer's market, but there is no formal hierarchy between riders and drivers. Both groups of members also have no internal hierarchy. The only hierarchy on the platform is that between core and periphery.

Twitter
The social media platform Twitter enables users to communicate via short messages directed to the public, specific groups of members or individual users. It is the prototype of its own genre, the "microblogging" platform (van Dijck, 2013), and one of the most frequented social media platforms worldwide (Bruns & Stieglitz, 2014). Twitter is used extensively by private individuals and organizations of all kinds. The platform provides one of the standard communication channels for political campaigning, news outlets and public communication in general (Weller et al., 2014).
Core organization. Twitter is operated by Twitter, Inc, a private company listed at the New York Stock Exchange. Reading and writing on the platform is free of charge, and Twitter, Inc has been experimenting with various business models to generate profits (van Dijck, 2013). As of 2021, the company's main sources of revenue are targeted advertising and fees for privileged direct access to portions of its platform data (Twitter, 2021b).
Interfaces. Twitter enables its users to publish short messages (tweet) and republish (retweet), answer to (reply) and show support for (favour) the messages of others. Messages are displayed on the Twitter website or dedicated app. Users can address the accounts of other users in their messages (mention) or align their message with the conversation about a specific topic by adding designated keywords (hashtags). Users can choose to receive all messages posted by specific accounts automatically (follow). By default, all messages are public, but authors can decide to make some messages available only to individual users or only to followers. Through these eight interfaces, Twitter users can communicate with each other and direct their conversations. The platform groups related messages to so-called threads (consisting of a message and its replies) or feeds (consisting of messages grouped by various other criteria). All groups influence how users perceive and react to the conversations on Twitter, and all are the direct result of users' interaction with the aforementioned interfaces. In addition to these user-generated groups, the platform displays a constantly changing set of promoted messages (trends). These are chosen by algorithms that evaluate platform data according to rules that are not transparent to platform participants (Leavitt, 2014).
Periphery. The social order of the Twitter periphery is usually regarded as a network (Bossetta, 2018;Halavais, 2014). More specifically, the platform contains two interrelated network layers: User accounts are connected to other accounts through the directed relations generated with the followinterface. Messages are connected to other messages through retweets, replies, or hashtags.
Connections between those two layers are generated through the manual interactions of users and the algorithmic proposals of the platform technology.
Membership. Everybody with access to the internet can use the Twitter website to search, read, favour, report misuse or create links from other websites to Twitter messages. All other forms of interaction, especially writing messages and following, are reserved for members. Membership is available to all visitors of the website without any preconditions.
Rules. Platform rules define acceptable message formats, acceptable interactions, and acceptable message content. While rules about format and interactions are highly specific, rules about content are extremely generalized (Twitter, 2021a). Although Twitter has a long list of guidelines on acceptable message content, they effectively allow most of the people to write what they want most of the time.
As of 2021, most rules about format and interaction are embedded into the platform technology, but none of the rules about content is. This difference has consequences when it comes to monitoring. Sanctions. Negative sanctions on Twitter range from warnings assigned to single messages through temporary account suspensions up to complete revocation of membership. Reviewers' interpretation of the rules is highly important for the choice of sanctions (Twitter, 2021a). However, direct sanctions by official reviewers are comparatively rare. Instead, the platform provides its users with the means to dole out "local sanctions" according to their taste: Participants favour or retweet messages they like, ignore those they dislike, and follow, unfollow or block authors according to the content they produce.
These individual-level sanctions are highly relevant for the social order of Twitter: They accumulate and increase a message's audience and thereby the author's influence on current conversations exponentially (Bossetta, 2018). The propositional algorithms amplify such dynamics by presenting popular messages to more users and thus inciting more interaction. They thus act like a positive sanction rewarding users who publish popular messages.
Hierarchy. Twitter members produce the content that members and non-members alike can read.
There is no formal hierarchy: Participants can only report infractions, but not enforce the platform rules. There is however a highly skewed distribution of influence since some members' messages reach an exponentially larger audience than others' (Bossetta, 2018). On the one hand, this reach emerges from users' individual decisions about who to follow and which messages to favour. On the other hand, it is affected by the platform technology: The algorithms of Twitter treat accounts unequally when devising trends or other propositions (Leavitt, 2014). Platform participants with higher reach have no direct authority over others, but their decisions (about what to publish) determine which forms of content other members get to read. Twitter thus exhibits a form of stratification, an asymmetric order of influence among participants which is emergent, but nudged along by the opaque algorithms devised by Twitter, Inc.

Wikipedia
Wikipedia is a free online encyclopaedia based on wiki software, i.e. software enabling users to read and edit websites inside a web browser. The platform is a showcase for peer production: Volunteers create, edit and monitor all articles, develop the software, and are involved in almost all aspects of platform operation (Niederer & van Dijck, 2010;Viégas et al., 2007). Versions of Wikipedia exist in over 300 languages, and the platform is one of the globally most visited sites on the internet (Wikipedia, 2021b).
Core organization. Wikipedia belongs to the Wikimedia Foundation, a non-profit organization founded in 2003, two years after the establishment of Wikipedia (Wikipedia, 2021a). The foundation raises donations for Wikipedia and related projects (van Dijck, 2013). It provides the technical infrastructure, basic administrative support and coordination for the platform, but does not operate it. Instead, each language version is governed by an autonomous organization, the country chapter (Wikipedia, 2021a).
The Wikimedia foundation connects the chapters and organizes collaborative decisions about shared rules and strategy (Dobusch & Kapeller, 2018). The purpose of both the foundation and the chapters is the safeguarding of the platform goals: to produce an encyclopaedia that is open to all (Weltevrede & Borra, 2016).

Interfaces. The Wikipedia website consists of encyclopaedic articles about topics of general interest,
articles with background information about the platform or individual members, and talk and history pages for each article (Weltevrede & Borra, 2016). Encyclopaedic articles hold the platform's content; they form the encyclopaedia proper and are the source of Wikipedia's popularity. Background articles, talk pages and article histories enable interaction on the platform; they provide the means and common reference points for discussions and are the source of Wikipedia's continual processes of organization (Viégas et al., 2007). Platform participants can search and read articles, create new and edit existing ones, review and comment on article changes and discuss these changes or any other aspect of the platform with each other (Wikipedia, 2021b). Most of the platform chores are done by bots, small programs created to perform standardized tasks without user interaction (Niederer & van Dijck, 2010).
Periphery. While passive readers are important for the purpose of the Wikipedia, only the editors interact on the platform and thus form the periphery. By changing articles, editors interact indirectly with all others who engage with the same article at some other time. Most participants never move beyond this kind of fringe participation (Niederer & van Dijck, 2010), but some become more involved over time. These most active participants create encyclopaedic articles, discuss their quality and other matters with each other, and resolve conflicts between their fellow editors (Viégas et al., 2007). They also develop the platform's rules and its technology further (Bryant et al., 2005). The highly engaged participants form the Wikipedia community, a network of volunteers committed to the purpose of the platform. These "Wikipedians" follow and defend the formal platform rules, even though their interactions and the stability of the network as a whole are mostly based on shared norms of collaboration (Benkler & Nissenbaum, 2006;Bryant et al., 2005).

Membership. Wikipedia articles are visible to everybody with internet access. This openness is central
to the platform purpose: "to benefit readers by acting as a free widely accessible encyclopaedia" (Wikipedia, 2021d, emphasis in the original). Non-members can not only read, but also edit articles and take part in discussions. In fact, almost all of Wikipedia's features are available to members and non-members alike (Bryant et al., 2005). Membership is easy to obtain: website users can join Wikipedia by choosing a username and a password and creating an account. Membership comes with a few additional rights (e.g. the right to create articles and personal user pages), but no specific duties.
Rules. Wikipedia has a large and ever increasing set of rules regulating every aspect of the platform (Butler et al., 2008). These rules are binding, but not stable; just like the articles of the encyclopaedia, they are continuously open for discussion and change by collective decisions made in the periphery (Sanger, 2005;Tkacz, 2014). They are thus formalized and explicit, but cannot be attributed to one responsible decision-maker. In consequence they can be regarded as emergent. These emergent rules are based on a few fundamental principles which are not up for change (Tkacz, 2014;Weltevrede & Borra, 2016). Dynamic and stable rules alike are collected in policies and guidelines; only very few are embedded in Wikipedia's technology or automated in one of the bots crawling the platform (Niederer & van Dijck, 2010).
Monitoring. The elaborate rule system of Wikipedia creates a great need for monitoring activity.
Members and non-members alike monitor articles, and active members often use bots, e.g. to be notified when articles change (Niederer & van Dijck, 2010). All monitoring duties are voluntary, but members of the community often "adopt" sets of articles they feel responsible for (Bryant et al., 2005).
Sanctions. Wikipedia has a set of formal positive and negative sanctions. The platform principles postulate freedom and consensus, so formal sanctions are legitimatized only as tools to resolve conflict after discussions have failed (Wikipedia, 2021c). This means that users are only sanctioned when they repeatedly violate the rules (and thus provably impede consensual conflict-resolution). It also means that there are ways to appeal and that sanctioning is highly transparent (van Dijck, 2013, p. 145).
Negative sanctions remove a user's right to edit specific articles or topics, for a specific amount of time, or permanently (van Dijck, 2013). Positive sanctions confer respected members of the community with the rights to sanction others (Bryant et al., 2005;Niederer & van Dijck, 2010).
Hierarchy. The periphery has a strong formal hierarchy. While unregistered users and novice members can edit articles, their only way to influence fellow participants is the convincing argument in a discussion. This changes when they start ascending the hierarchy: On every level, members gain additional permissions (embedded in technology) allowing them to edit more or stop others from interacting with the platform (van Dijck, 2013). None of the positions has formal duties, and none lets one member define such duties for another. However, positions in the hierarchy come with the strong informal expectation to use the granted permissions actively and in accordance with the rules. The resulting "system of normative control" (van Dijck, 2013, p. 147) blends aspects of decided (permissions) and emergent order (expectations).

Summary of results and case comparison
The concept of platform organizations enables us to identify commonalities and differences of three exemplary platforms which provide different functions and applications across domains and are usually investigated in different scholarly discourses. Table 1 depicts the salient features of each platform that we drew on in the previous section.
As we have shown, all three can be regarded as platform organizations according to our definition of the concept: Uber belongs to a listed corporation operating the platform as a means to generate profit for its stakeholders. To this end, the core organization Uber Technologies organizes the platform periphery as a market providing income to the core. The interfaces are designed to create and stabilize this marketplace, foster its growth and sustain a steady stream of commissions. The social order enabling this gainful arrangement combines relatively high restrictions for membership, closely monitored rules for most interactions and more rigid sanctions for drivers than for riders. It has no hierarchy among elements of the periphery, but a simple hierarchy based on the platform architecture: the core organization takes all decisions about the platform, and members of the periphery abide by them or leave.
Twitter also belongs to a listed corporation and is also this corporation's sole basis for income. Twitter, Inc., does not organize a market, but a network of microbloggers engaged in lively conversations. By writing, reading and commenting, the participants in the periphery produce data which can be monetized by the core. The interfaces are designed to ensure that the conversations never stop and that Twitter, Inc. is supplied with a continuous flow of new user data. The social order of the platform is highly permissive, with limited access to conversations to all, a low threshold for membership, and rules, monitoring, and sanctions designed to primarily restrict the form of conversations but not their content. This content is secondary to the core, but most relevant to participants. Spurred on by Twitter's algorithms, their individual decisions to prefer some members' content before others accumulate and produce the exponential differences in reach that are characteristic for the social order of Twitter. However, these differences do not constitute a hierarchy, since all rights to sanction participants remain in the core. Twitter, like Uber, only exhibits the basic architectural hierarchy between core and periphery.

Table 1: Architecture and elements of organization across the three cases (own depiction)
Note: Own depiction, * all three cases exhibit a basic hierarchy between core and periphery.
In contrast to our two commercial examples, Wikipedia belongs by a non-profit organization founded by members of the Wikipedia community as a means to safeguard the community and its creations.
The periphery comprises a network of contributors dedicated to the creation of an open encyclopaedia as well as to the idea of self-organization. The interfaces are designed to simplify participation, support collaboration and foster transparency. The line between non-members and members is fluid, reflecting

Degree of organization
High -Extensive rules -Extensive monitoring -Extensive sanctions

Medium -Extensive rules -Extensive monitoring -Sporadic sanctions
Low -Limited rules -Sporadic monitoring -Sporadic sanctions the ideal of openness. The social order of the platform is backed by a large number of rules, emergent responsibilities for monitoring, precisely defined sanctions, and a steep formal hierarchy. Surprisingly, the strong norms of openness and participation guiding interactions on the platform produce a fairly bureaucratized form of social order, albeit one based on continuous discussions and regular votes. The Wikimedia foundation owns the technical infrastructure of the platform and ensures that the dynamic rules conform to the fundamental principles of Wikipedia. The platform thus also exhibits the basic architectural hierarchy between core and periphery, even though the platform organizer delegates much authority to the participants at the periphery.
The periphery of all three platforms has a characteristic social order which is the result of partial organization. Some of the elements of organization are the result of decisions, some are emergent, and these differ from platform to platform. On Twitter, platform-wide decisions define only a loose frame for interactions. Rules, monitoring and sanctions mostly emerge when members decide who to follow, favour, or block. The platform has a stratified order of influence, but that, too, is emergent, as it results from the accumulated preferences of individual members. All in all, Twitter's degree of organization is low.
Wikipedia is much more organized than Twitter, with many formal rules, a partly automated system for monitoring, and a legalistic apparatus for sanctions and appeals backed by an elaborate hierarchy. its members-only access to the platform, the highly regulated processes and the predefined role set for the periphery. It is exactly this very high degree of organization that exposes Uber Technologies to lawsuits where Uber drivers claim that they are in fact employees suffering a wrongful classification of their membership status with the company (Vallas & Schor, 2020). According to the core organization, the platform does not regulate participants' activities to such an extent that drivers' membership would constitute an employment relationship. Instead, the algorithms supposedly merely offer recommendations to support participants' autonomous decisions. These arguments are fundamental for Uber Technology's business model (Uber, 2021a) which is currently disputed in science (Prassl & Risak, 2016) as well as in courts around the world (De Stefano et al., 2021). If we reformulate the argument from our perspective, the platform organizer claims that relevant aspects of Uber's social order are emergent, and not the result of decisions by the core.

Discussion and conclusion
In this paper we approached digital platforms from a partial organization perspective and assumed that they constitute a specific form of organization -the platform organization. We argued that all major digital platforms share a common basic architecture where a formal organization employs digital technology to create a social order in its periphery. We developed the concept of platform organization to reveal the commonalities and differences of digital platforms. The concept shows how a platform organizer uses digital interfaces to govern the interactions among participants loosely coupled to the core. Here, the platform organizer configures organizational elements and enables a social order at the platform periphery. We highlighted that digital platforms do not merely differ by functions or practical applications across domains but exhibit a common architecture with varying degrees of organization.
We focused our argument along four core insights, claiming (1) that all digital platforms share a common architecture, (2) that the digital interfaces deployed by the platform organizer give form to the social order of the periphery, (3) that the five elements of organization defined in the partial organization perspective capture the fundamental features of this social order, and (4) that digital platforms can be more or less organized. We applied our concept to Uber, Twitter, and Wikipedia to illustrate the argument. Drawing on these three seminal cases as platform organizations allowed us to elaborate the four core insights, which we now discuss.
All cases clearly exhibit the architecture of stable core, dynamic periphery and interfaces mediating between the platform parts. In two of our cases the core is formed by listed corporations, in one the platform organizer is a non-profit foundation. On Uber, the core organization creates a market for transport services, on Twitter it makes conversations possible, and on Wikipedia it supports a collaborative environment in which participants can write an encyclopaedia. While the individual platforms thus facilitate highly diverse activities and types of exchange between platform users, their underlying principles are highly similar.
The organizers of all three platforms employ digital technologies to enable a social order in the platform periphery. Although the digital technologies used in each case seem similar when regarded superficially, closer inspection shows that each platform has different interfaces. These interfaces enable and impede a set of activities and regulate some aspects of these activities more thoroughly than others. In consequence, we find a distinct social order manifesting in the periphery of each platform organization. Researchers focusing only on this periphery will see that in their case, digital technology enables a market (Uber), a network (Twitter), or a community (Wikipedia).
The fundamental features of our three examples are captured through the five elements of organization proposed by the partial organization approach. Platform organizers substantially differ in their approaches towards governing the social orders they have created. These differences mostly arise from the ways in which platform organizers embed rules, monitoring and sanctions into the digital interfaces of the platform. The configuration of rules, monitoring, and sanctions defines the boundary between members and non-members. This boundary can be solid, as in the case of Uber, or fluid, as on Twitter and Wikipedia. Only the latter platform has a formal hierarchy among its participants. In the other two cases, the platform organizers do not delegate the right to sanction to the periphery.
With our concept of platform organization, we can distinguish the three exemplary cases by degree of organization and see that our examples range between higher organized and lower organized platforms. This degree of organization of each platform depends on the extent to which platform organizers try to intervene in the social order of the periphery by either deciding on the implementation of organizational elements via interface design or leaving their development up to emergence. Uber presents a highly organized case with extensive rules, extensive monitoring and extensive sanctions. Wikipedia ranges in the middle as it has extensive rules and extensive monitoring, yet only deploys sanctions sporadically. Twitter has only a very limited set of rules and conducts sporadic monitoring and sanctioning, which makes it a case of a low organized platform.
Concluding from our argument, we can identify the platform organization in general as a form of decided social order. The platform organizer's decisions about the interfaces are fundamental for the social order manifesting at the periphery. These decisions shape the digital technology on which all interactions and relationships on the platform depend. By providing this technology, platform organizers retain the capability to interfere strategically in the periphery. While organizers may leave aspects of the social order up to emergence, they can always reconfigure the interfaces to trim and adjust emerging patterns in accordance with their organizational goals. The architectural pattern of the platform organization thus creates a decided order with only a basic hierarchy between core and periphery.
Our concept of platform organization goes beyond established research as we show that platforms are much more than technologies combined with business models that facilitate innovations or market transactions. We argue that any given platform establishes a particular type of social order at its periphery. Our general perspective allows us to integrate platforms like Wikipedia, that do neither pursue profits nor participate in market competition, into platform research. Conversely, we can extend our perspective to the ongoing debates about "innovation" and "transaction platforms" (Gawer, 2020). We encourage researchers in these debates to consider the many societal preconditions of the development, success or failure of these platforms. Extending our perspective in this vein, we also criticise the popular term "platform ecosystem" (Cusumano & Gawer, 2002) because it elicits a notion of emergent order and naturally growing relationships. By accentuating the social order of digital platforms, we highlight the fact that all relations that can develop on these platforms derive from decisions made by the platform organizer. Metaphors emphasizing emergence and suggesting quasi-natural processes effectively conceal this position of the platform organizer and the fact that each digital platform presupposes a decided order.
We also extend on research in which platforms are considered as governance mode by showing that platforms always require a formal organization at the core, but exhibit various forms of coordination among their peripheral elements. Here, we highlight the importance of digital interfaces for establishing and maintaining social order in the platform periphery. Advancing the partial organization perspective, we also show that digital platforms are based on a decided order in which a complete organization, the platform organizer, implements a partial organization via digital interfaces. Digital platforms can be more or less organized, as the magnitude of decided and emergent aspects of the partial organization in the periphery varies considerably between cases. Depending on their degree of organization, digital platforms might seem more like regular formal organizations, as in the case of Uber, more like networks, as in the case of Twitter or more like communities, as in the case of Wikipedia. Our concept of platform organization enables researchers to identify two distinct layers of governance: an architectural layer where the core governs the periphery, and a component layer where coordination at the core and at the periphery follow different modes. We show that digital platforms always amalgamate two types of social order, a formal organization as platform organizer and some form of partial organization in the periphery.
We also contribute to the ongoing debate on "degrees of organization" or "organizationality" (Ahrne et al., 2016;Dobusch & Schoeneborn, 2015;Kirchner & Schüßler, 2019;Grothe-Hammer et al., forthcoming). Here, we posit that the diffusion of digital technology into various spheres of the society spurs an abundance of patterns of organization. However, while many of these organizational patterns involve formal organizations, like digital platforms, many organized activities no longer take part within formal organizations. Rather, organizations establish decided social orders that serve to coordinate activities of many other actors, be they individuals or other organizations. An explicitly organizational perspective can provide valuable insights into the ongoing relevance of formal organizations for a society where digital technology advances an abundance of forms of "organization outside organizations" (Brunsson & Ahrne, 2019).
Finally, we would like to emphasise the fundamental relevance of digital technology for platform organizations. Our concept highlights that digital platforms share the common architecture of a core and a periphery that relate through interfaces. However, core-periphery patterns constitute a common issue in organizational research and related research fields. For example, seminal research on work systems (Kalleberg, 2001) or industry networks (Piore & Sabel, 1986) already highlighted the interrelations between core and peripheral elements. From a historical perspective the emergence of digital platforms (Davis, 2016) can be understood as a steady erosion organizational structures transforming into core and periphery patterns, e.g. through outsourcing or franchising. Against this background of gradual transformations, we posit that the platform organization constitutes a qualitative leap: The unique and essential property distinguishing digital platforms from other forms of organizing core-periphery structures is their reliance on digital interfaces. Digital interfaces regulate the relationships between core and periphery as well as the relationships between peripheral elements through the platform core. Essentially, digital technology amplifies and focuses the core's ability to regulate, monitor and sanction states and activities of peripheral elements. The interfaces used in platform organizations thus raise core-periphery-relations to a new level of centralized governance. Our concept thus points towards interfaces as the central components and distinguishing feature of the platform organization as organizational form.
Overall, our paper offers a theoretical concept that defines the platform organization as a specific form of social order and combines the perspective of a general platform architecture with the approach of partial organization. From this combination, we derive dimensions for analysing how the social order 19 of a platform organization is configured. With our concept we highlight commonalities and differences of platforms providing diverse functions and applications across societal domains. Our argument shows that an integration of approaches of platform architecture and social order on platforms can inform a deeper understanding of what digital platforms are and what they do.