Complexity And Information Systems Research In The Emerging Digital World

Complexity is all around us in this increasingly digital world. Global digital infrastructure, social media, Internet of Things, robotic process automation, digital business platforms, algorithmic decision making, and other digitally enabled networks and ecosystems fuel complexity by fostering hyper-connections and mutual dependencies among human actors, technical artifacts, processes, organizations, and institutions. Complexity affects human agencies and experiences in all dimensions. Individuals and organizations turn to digitally enabled solutions to cope with the wicked problems arising out of digitalization. In the digital world, com- plexity and digital solutions present new opportunities and challenges for information systems (IS) research. The purpose of this special issue is to foster the development of new IS theories on the causes, dynamics, and consequences of complexity in increasing digital sociotechnical systems. In this essay, we discuss the key theories and methods of complexity science, and illustrate emerging new IS research challenges and oppor- tunities in complex sociotechnical systems. We also provide an overview of the five articles included in the special issue. These articles illustrate how IS researchers build on theories and methods from complexity science to study wicked problems in the emerging digital world. They also illustrate how IS researchers lever- age the uniqueness of the IS context to generate new insights to contribute back to complexity science.

1 Keywords: Complexity, sociotechnical systems, emergence, coevolution, chaos, scalable dynamics digitalization

1 Hind Benbya and Ning Nan served as associate editors for the special issue. Huseyin Tanriverdi and Youngjin Yoo served as senior editors. William McKelvey

was a SE for the special issue but was unable to participate in the writing of the introductory essay.

DOI: 10.25300/MISQ/2020/13304 MIS Quarterly Vol. 44 No. 1, pp. 1-17/March 2020 1

Benbya et al./Introduction: Complexity & IS Research

Introduction

When we conduct a search on Google, it returns hundreds, of thousands, results instantaneously. The results not only reflect the interests of the one who is doing the search, but also the millions of internet users who created or clicked on hyperlinks of websites. As more users search, link, and click with similar keywords, the results will continue to change according to user location and search time. A search for “Korean restaurants” in Munich, Germany, for example, gives different results from a search in Cleveland, OH, USA. Con- ducting the same search a day or two later also produces different results. A simple Google search result is an emer- gent property, a complex web of interactions among users, websites, topics, advertisers, and many other social or tech- nical entities. In short, our daily experience of using mundane digital tools is a dynamic emergent outcome of complex sociotechnical systems.

As early as 2010, the world-wide production of transistors has exceeded that of rice, and is much cheaper (Lucas et al. 2012). Devices—large and small—powered by microprocessors and connected by the internet are filling every inhabited corner of the earth. Some of these devices are not just passively waiting for commands; equipped with a powerful artificial intelligence engine, they often act on their own. We already see autonomous vehicles on the streets interacting with traffic signals that respond to changing traffic patterns, in the midst of human-controlled vehicles and pedestrians. Sprinklers are connected to the weather service on the internet to control the amount of water on a lawn. The temperature of millions of houses is controlled by Nest connected to the Google Home Assist service. Connected speakers recommend different music playlists based on the time, location, and, of course, your preference. Social network services also enable every user as a potential content creator on the internet. Once created, user-generated content can be liked, shared, and mashed with other content by other users, often creating unpredictably complex forms of diffusions. Digital platform ecosystems such as Uber and AirBnB connect millions of users and providers globally. More than 80% of movies watched on Netflix are recommended by algorithms.2

These examples illustrate truly astonishing advances from the humble start of computers in organizations in the early 20th

century. After merely a few decades, what once seemed to be glorified calculators have evolved into digital technologies that permeate our lives and work. These digital technologies in turn foster new sociotechnical systems such as wikis, social

media, and platform ecosystems that are fundamentally changing the way people work and live.

Not every technological invention has such a transformational impact. What set apart digital technologies? At the heart of digital technologies is symbol-based computation. Bistrings (0s and 1s) provide a standard form of symbols to encode input, process, and output of a wide variety of tasks (Faulkner and Runde 2019). They reduce the design specificity of hard- ware for operationalizing the symbol-based computation. Furthermore, simplicity of bitstrings eases the effort to shrink the size, reduce the cost, and increase the processing power of hardware. Symbol-based computation provides a generali- zable and applicable mechanism to unite the operations of matter and the abstract mental processes (Lovelace 1842). It lays the foundation for digital technology to rapidly advance beyond the function of a calculator. More importantly, symbol-based computation sets in motion the emergence of complex sociotechnical systems.

Emanating from symbol-based computation are a few complexity-inducing characteristics of digital technologies.

• Embedded: as described by the vision for symbol-based computation (Lovelace 1842; Shannon 1993, Turning 1950), digital capabilities are increasingly embedded in objects that previously have pure material composition (Yoo et al. 2012). Digital capabilities can encode and automate abstract cognitive processes for converting new information into adaptive changes of objects. They also enable objects to provide decision support to adaptive cognitive processes of social actors.

• Connected: objects embedded with digital capabilities and users of such objects can be connected into webs of sociotechnical relations (Sarker et al. 2019) because symbol-based computation homogenizes data (Yoo 2010). When information is shared in the webs of socio- technical relations, abstract cognitive processes encoded in objects or possessed by social actors become mutually dependent.

• Editable: digital technologies are editable (Kallinikos et al. 2013; Yoo 2012) due to symbol-based computation. This editability allows increasingly diverse cognitive processes to be introduced into the webs of socio- technical relations. Recurrent adaptation of diverse, connected, and mutually dependent objects and social actors can amplify or diminish an initial change in a sociotechnical system, producing outcomes that defy simple extrapolation from the initial change (Arthur 2015; Holland 1995; Page 2010). Complexity, therefore, becomes a salient attribute of sociotechnical systems.

2 See https://mobilesyrup.com/2017/08/22/80-percent-netflix-shows-

discovered-recommendation/.

2 MIS Quarterly Vol. 44 No. 1/March 2020

Benbya et al./Introduction: Complexity & IS Research

• Reprogrammable: through the separation of hardware and software of symbol-based computation, digital tech- nology is reprogrammable (Yoo et al. 2010). The same hardware can perform different functions depending on the software that runs on the device.

• Communicable: digital technologies are communicable by following a set of agreed-upon protocols (Lyytinen and King, 2006; Yoo 2010). With the pervasive diffu- sion of digital technologies, they now form a global digital infrastructure (Tilson et al. 2010).

• Identifiable: each and every device connected to the digital infrastructure is uniquely identifiable through its own unique address (Yoo 2010). The increasing digital penetration leads to a higher degree of identifiability, allowing for more granular manipulation levels of digital objects.

• Associable: digital objects are associable through shared traits. The associability of distributed heterogeneous devices and data allows one to identify emerging patterns across different realms and geographies in a way that was simply not possible in the past.

Digital technologies not only give rise to complex sociotech- nical systems; they also distinguish sociotechnical systems from other complex physical or social systems. While com- plexity in physical or social system is predominantly driven by either material operations or human agency, complexity in sociotechnical systems arises from the continuing and evolving entanglement of the social (human agency), the symbolic (symbol-based computation in digital technologies), and the material (physical artifacts that house or interact with computing machines). The functions of digital technologies and the roles of social actors are perpetually defined and redefined by each other (Faulkner and Runde 2019; Zittrain 2006). This sociotechnical entanglement limits the generali- zability of complexity insights obtained from nondigital systems to complex digital systems. Furthermore, while material operations or human agency either increase or dampen complexity in physical or social systems, digital tech- nologies can both mitigate and intensify complexity. This is because individuals and organizations engaged with complex sociotechnical systems often turn to digital technologies (e.g., data analytics) for solutions to complex problems. Yet, the application of a solution can instigate a new round of digitally enabled interactions that diminish the intended effect of the solution. This dual effect of digital technologies on com- plexity can produce dynamic interaction patterns and out- comes that are qualitatively different from those in other complex systems.

The distinct effects of digital technologies on complex socio- technical systems present an important opportunity for infor- mation systems (IS) researchers to extract novel insights regarding the nature and relevance of digital technologies. IS researchers can apply theories and methods from complexity science to model observations that defy simple extrapolation from initial changes in a sociotechnical system. In this essay, we introduce key complexity theories such as emergence, coevolution, chaos, and scalable dynamics as the most likely foundation for IS researchers to rethink predictability, caus- ality, boundary, and durability of observations in the digital world. Subsequently, we explain how the centrality of symbol-based computation in IS research paves the way for IS-specific research themes to extend complexity science. The articles in this special issue are briefly described to illus- trate a few prominent themes such as IS development for rapidly changing requirements and using digital technologies to steer or tame complexity.

Complexity Science: Key Theories and Methods

Complexity science’s origins lie in 50 years of research into nonlinear dynamics in natural sciences and spans a variety of scholarly disciplines including biology (Kauffman 1993), chemistry (Prigogine and Stengers 1984), computer science (Holland 1995; Simon 1962), physics (Gell-Mann 1995), and economics (Arthur 1989). Developments across disciplines over time resulted in a meta-theoretical framework within which several theoretically consistent approaches and methods can be integrated.

Complexity science theories and methods combine different epistemologies (i.e., positivism, interpretivism, and realism) to provide novel opportunities to question assumptions (e.g., equilibrium, stability, etc.), manage tensions and paradoxes, and rethink the way we view many sociotechnical phenomena at the center of our field. Their value is particularly promi- nent when the research community faces new phenomena and questions that do not lend themselves well to the traditional, reductionist approaches.

Complexity Drivers and States

Complexity is an attribute of systems made up of large num- bers of diverse and interdependent agents3 that influence each

3 These could range from molecules to individual human beings to organized

collectives.

MIS Quarterly Vol. 44 No. 1/March 2020 3

Benbya et al./Introduction: Complexity & IS Research

Figure 1. States of Complex Systems (Benbya and McKelvey 2011)

other in a nonlinear way and are constantly adapting to inter- nal or external tensions (Holland 1995). Because such systems are constantly evolving, they have a large degree of unpredictability. They cannot, therefore, be understood by simply examining the properties of a system’s components.

Four key characteristics influence the level of complexity in a system: (1) diversity, (2) adaptiveness, (3) connectedness, and (4) mutual dependency among agents in the system (e.g., Cilliers 1998; Holland 1995). The nonlinear interplay of the above four characteristics coupled with increased tension in the form of external or internal challenges and/or oppor- tunities drive the system from one state to another.

A system can exist or fluctuate between three states or regions: stable at one extreme, chaos at the other, with an in- between state called the edge of chaos (Kauffman 1995; Lewin 1992). Figure 1 provides an illustration of the three states.

Specifically, in the stable state, the diversity, adaptiveness, connectedness,and mutual dependency of agents in the system are all at low levels. Consequently, adaptive tensions are low (Page 2010) and complexity is benign (Tanriverdi and Lim 2017). The system rapidly settles into a predictable and repetitive cycle of behavior. In such stable systems, novelty is rare. There is a tendency for stable systems to ossify.

As the diversity, adaptiveness, connectedness, and mutual dependency levels of systems reach moderate levels, the com-

plexity level increases (Page 2010). Systems with increased levels of complexity enter the so-called “edge of chaos” state or a region of emergent complexity (Boisot and McKelvey 2010). By staying in this intermediate state, these systems never quite settle into a stable equilibrium but never quite fall apart. They exhibit continuous change, adaptation, coevolu- tion and emergence (Kauffman 1993; Lewin 1992).

Increasing levels of tensions, beyond a certain threshold, might result in chaos or extreme outcomes (e.g., catastrophes, crises, etc.) which exhibit fractals, power laws, and scalable dynamics. Chaotic systems never really settle down into any observable patterns. Since they are sensitive to initial condi- tions, they can amplify exponentially and have monumental consequences (Gleick 1987).

Complexity Theories

As outlined above, many living systems (e.g., organisms, neural networks, ecosystems) on the edge of chaos appear to constantly adapt and self-organize to create configurations that ensure compatibility with an ever-changing environment. This perpetual fluidity is regarded as the norm in systems on the edge of chaos; it can lead to processes and outcomes as diverse as phase transitions, catastrophic failures, and unpre- dictable outcomes (see Table 1). Complexity theories such as emergence, coevolution, chaos, and extremes, as well as scalable dynamics, offer an explanation of such processes and outcomes.

4 MIS Quarterly Vol. 44 No. 1/March 2020

Benbya et al./Introduction: Complexity & IS Research

Table 1. Processes and Outcomes of Complex Systems

Complexity Theories Processes Outcomes

Emergence • Disequilibrium situations: tensions, triggers and small events outside the norm

• Positive feedback and bursts of amplification

• Phase-transitions • Self-organization

• Unpredictable outcomes: new structures, patterns, and properties within a system (e.g., distributed leadership emergence), a new level of analysis (e.g., a network), or a collective phenomenon (e.g., collective action)

• Emergence can take two forms: composition or compilation

Coevolution • Interdependency and boundary-crossing interrelationships

• Multilevel dynamics • Bidirectional or two-way causality

• Mutual influences • Reciprocal adaptations and changes over time

Chaos • Sensitivity to initial conditions • Constrained trajectory (e.g., strange

attractor) • Time-dependency and irreversible

dynamics

• Catastrophic failures (e.g., systemic risk, cyber- security breaches)

• Escalation of causes leading to disastrous societal consequences (e.g., disrupting lives on a large scale)

Scalable Dynamics • Instability and large variations • Single cause leading to a cascade of

interconnected events

• Self-similarity across scales • Positive or negative extreme outcomes • Fractal dynamics • Power laws

Emergence

Emergence is a dynamic process of interactions among heterogeneous agents that unfolds and evolves over time, resulting in various kinds of unexpected novel individual- and group-level configurations and/or broader social structures (Benbya and McKelvey 2016). Complexity and organization scholars have theorized such a dynamic process for some time (Kozlowski et al. 2013; Plowman et al. 2007).

Systems-wide changes in natural open systems revealed how unorganized entities in a given system, subjected to an exter- nally imposed tension, can engage in far-from-equilibrium dynamics. The entities can therefore self-organize into dis- tinct phase transitions leading to a new higher-level order (Prigogine and Stengers 1984).

Social systems put under tension, through recession, crisis, organizational change, and so forth, can exhibit similar phase transitions and emergent outcomes. As such, many social scientists have made a direct mathematical parallel between physical and social systems to deduce the process mech- anisms inherent in micro interaction dynamics that yield the higher-level order and its emergent novel outcomes. They have identified two forms of emergence: composition or compilation (Kozlowski and Klein 2000). In composition models, emergent processes allow individuals’ perceptions, feelings, and behaviors to become similar to one another.

Compilation models, on the other hand, capture divergence. They characterize processes in which lower-level phenomena are combined in complex and nonlinear ways to reflect unit- level phenomena that are not reducible to their constituent parts. The discovery of emergence involves either a post hoc analysis of time series data (e.g., system behavior) and conceptual tools that allow scholars to verify the existence of emergence dynamics in systems, or an analytical mapping of the sequential phases of emergence dynamics (e.g., Plowman et al. 2007).

Interactions among sociotechnical entities yield many emergent outcomes in information systems. Examples include the collaborative creation of online order and tech- nology affordances (e.g., Nan and Lu 2014), IS alignment (Benbya et al. 2019), and new configurations among organi- zation, platform, and participant dimensions (Benbya and Leidner 2018). An emergence perspective offers a lens to understand many unpredictable sociotechnical phenomena that span individual, group, organizational, and societal levels in the context of widening digitalization.

Coevolution

Coevolution refers to the simultaneous evolution of entities and their environments, whether these entities being organisms or organizations (McKelvey 2004). Ehrlich and

MIS Quarterly Vol. 44 No. 1/March 2020 5

Benbya et al./Introduction: Complexity & IS Research

Raven (1964) introduced the term coevolution to characterize the mutual genetic evolution of butterflies, and associated plant species. Such a process encompasses the twin notions of interdependency and mutual adaptation, with the idea that species or organizations evolve in relation to their environ- ments, while at the same time these environments evolve in relation to them.

In addition, to the above characteristics, coevolutionary processes have three main properties. First, coevolutionary phenomena are multilevel. They encompass at least two dif- ferent levels of analysis. Second, coevolutionary phenomena take time to manifest. This implies that longitudinal designs are necessary to understand coevolutionary processes. Third, bidirectional causality or two-way relationships (e.g., Yan et al. 2019) are central to coevolutionary processes.