Information ecology-complex adaptive organizations

A neuro-biological approach to modeling the information ecology of the Complex Adaptive Organization means to see organizations as living, conscious entities dedicated to problem solving. In this perspective organizations are concerned with efficient processing and treatment of different information stimuli. Patterns, composition and frequencies of information flows lead to evolutionary adaptation of the affected information circuits in the form of a re-wired organizational information network. The most efficient and probably most complex information processing system we know today is the human brain. It therefore seems appropriate to use some of the basic features and mechanisms that occur in neural networks of the brain to model dynamics of efficient information management. The simulation model we created could be called a “multi-agent based neural network simulation”: Imagine employees as agents, acting in a company just like an interconnected group of artificial nerve cells that affect each other in such a way as to arrive at a result based upon their inputs. This is adjusted in time until it best matches the required answer. Recurring processes related to information flows lead to evolutionary adaptation of the neural network, adaptation in reaction to the sensory inputs, in order to become more efficient in solving “standard problems” of information processing. In our conceptual model, the resources available to the agent - necessary to release organizational energy - are knowledge, time, attention and other agents. The model in general needed to include aspects of organizational culture, environmental influences and the behavior of information actors affected by their individual, dynamically changing characteristics.

Agents
Agents posses knowledge and the ability to communicate; they can therefore exchange certain information with each other. The exchange of information consumes the agents’ resources: time and attention. Agents exchange information preferentially, which means that some agents have a higher probability to mutually exchange information. This probability depends on the quality of past experiences during mutual interactions that were either positive or negative in regards to project completion. Agents learn either if their exchange partner shares information that is new to them, that is not yet known, or if, due to a successfully completed objective, they create new knowledge. Then their individual competence grows, since they are able to perform more efficiently in fulfilling organizational objectives.

Knowledge - The “knowledge profile”
To implement the knowledge that an agent holds, imagine an explicit visualization as a list that details areas in which the agent possesses knowledge, and the degree of mastery of that knowledge. This is what we call an agents “knowledge-profile”.

Knowledge- Area 1 2 3 4 5 6 7 8
Competency-
Level 0 4 6 2 0 0 3 6

The length of the knowledge-profile is chosen exemplary; here it contains 8 knowledge-areas. The competency levels are heterogeneous and variable, as they are constantly affected by agents’ interactions. Competence, understood as “enacted knowledge”, can grow if agents learn, and decay if the competence is not used. Knowledge is enacted, or used when agents communicate, i.e. actively share information.

Preferential exchange- the communication vector
Agents exchange information preferentially, that is, there is a probability of initiating communication with certain other agents. An agent is only conscious of those agents that he knows. To implement this idea in the simulation, we used vectors. The individual communication vector of an agent displays his individual preferences in the form of probability values for contacting other agents. Their length shows how many agents an individual agent can possibly connect to (respecting the connectivity restraints, see paragraph “Organizational objectives - the recipe”):

comm. -vector agent-C: [ 0.15 ; 0.06; 0.15 ; 0.15 ; 0.8 ; 0.19 ; 0.16 ; 0.06 ]

This is an exemplary communication vector of agent C. He knows 8 other agents that he contacts with different probabilities. We aggregate all individual communication vectors to a communication matrix that visualizes all inter-agent relations. This is what we called the “weighted information network map”. It adds qualitative and cultural aspects to a purely relational network map, as it integrates the agents’ dynamically changing preferences to activate certain “nodes” of the network. The inter-agent relations in this model can be a-symmetric, meaning that the probability of A contacting B is not equal to the probability of B contacting A.

“Weighted information network map”:

AGENT A AGENT B AGENT C AGENT D AGENT E AGENT F AGENT G AGENT H AGENT I
AGENT A 3 25 5 5 12 12 9 14
AGENT B 10 6 34 18 12 17 4 9
AGENT C 15 6 15 15 8 19 16 6
AGENT D 5 34 5 …
AGENT E 23 8 35 …
AGENT F 12 32 0
AGENT G 12 4 2
AGENT H 9 7 46
AGENT I 14 10 0

Strategic objectives - the recipe
The recipe is an abstract requirement list containing organizational objectives in the form of competency levels to be activated and released by the agent population. It underlies an evolutionary dynamic that reflects externally changing market requirements. Once the recipe has been created, it is postulated to a randomly assigned agent (= entry point). This agent checks how far his own knowledge profile and competencies cover the stated requirements. In most of the cases (and taking the realistic situation into account that the complexity of today’s organizations overwhelms individual capacities) the agent starts to spread the recipe to his colleagues that are available in his communication vector. He does this by creating a new, or activating an existing link between him and his exchange partner. The creation and maintenance of a connection between agents consumes an energy which we label “attention”. So communication is only possible if the exchange partner has sufficient “attention” in order to establish and maintain a connection. Once a connection between two agents is established, it allows information to flow. Each active connection, implying a process of active communication, consumes attention. The maximum number of simultaneously active connections is globally limited for all agents to the value “connectivity”. The minimum of connections an agent is obliged to maintain is fixed.

Knowledge
Area
Entry point 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Competence
level Agent C 3 5 9 0 0 4 1 0 10 13 5 0 0 11 2

Communication, knowledge exchange and learning
In the run of communication, agents are combining their knowledge and competence by exchanging information with each other. Agents learn from each other in the course of recipe completion. This is the case, if one or both exchange partners have heterogeneous knowledge profiles. Then it is possible that some of the information exchanged raises the knowledge of the agent. Of course agents learn only in those areas, in which the competence of their exchange partner is higher. Learning takes place, when an agent discovers that his competence is lower. In that case he can pick up one unit of competence as long as he has sufficient attention to “keep up the line”. Agents don’t lose knowledge related to the competency that they share, they rather “keep it up to date” and stay on their level of competence whilst using it. On the other hand, connections and competence levels in knowledge areas that are not used in the exchange process, decline exponentially. Relations and knowledge areas are not kept up to date in regards to the ever changing organizational environment. The rates of knowledge- and connection decline can be fixed. Once all knowledge areas are covered by interconnected, communicating agents that have sufficient competence levels, we say the recipe has been accomplished. Agents that collaborated successfully in the moment of recipe completion create new knowledge, which raises their competence.

System adaptation and network evolution
In order to create a truly adaptive system, a “learning organization”, certain feedback loops needed to be implemented. Those active inter-agent communication configurations that lead to recipe completion are strengthened: the probability to re-contact the colleagues with whom the agent has successfully performed raises. The weights in the weighted information network map are shifted to reinforce “good practice” patterns. This feedback-loop is the fitness function that the social network follows. It leads to an evolutionary network adaptation. At steadily changing market conditions it forces the organization to adapt their strategy. Thus evolution is in fact co-evolution. We design a simulation with “mutating” recipes that randomly change their composition. The speed of market change dictates how urgent the necessity of adaptation is. A drastically changing content is a sign of fundamental changes on the market landscape.

Organizational energy release
Efficiency is measured via a ratio of an input-output relation. In this conceptual model agents provide time, attention, knowledge and readiness to share information and generate in collaboration with other agents accomplished recipes. Using a ratio which has been proposed by Simon and Davila (2000), we can measure the managerial efficiency with which information is managed. Like similar ratios, for example the ROI (Return on Investment), the Return on Management (ROM) describes how efficiently a manager has chosen amongst various information sources, given limited managerial time and attention, to release organizational energy. Simon and Davila defined the return on management as follows:

ROM = Organizational energy released / time and attention invested

We expanded this ratio by integrating the number of agents necessary to fulfill the strategic objectives. In this perspective, information is managed efficiently if a minimum of time, attention and involvement of agents is necessary in order to accomplish the problem solving process. In our simulation the ROM is consequently defined as follows:

ROM = Recipes completed / time and attention invested + number of agents involved

In our conceptual approach we followed the idea of the networked organization. Any organization as an entity is part of a larger network of other organizations, which reflects an adaptive and also networked market for the outputs that organizations generate. Organizations compete for certain resources, co-operate to face certain market challenges and thus constantly change the market-landscape. In a sense any market is a fitness-landscape (Kauffman, 1996) in which organizations are placed on different fitness-levels, depending on the capacity that an organization has in coping with restraints and in exploiting opportunities. The necessity of organizational adaptation is subject to the evaluation of the current market-landscape topology. Observable market changes lead to the (re-)formulation of corporate strategy. Strategies are problem-solving algorithms or “cooking recipes” formulated in form of operational objectives that need to be accomplished by managers and employees. The efficiency with which an organization is acting in a market depends on the capacity that an organization can generate in reaching these strategic objectives. This in turn depends on the efficiency of the internal information network. Efficiency and competitively of the networked organization in this perspective is measured in terms of information diffusion and absorption speed, reactivity, flexibility and the adaptability of network structures to change. The network structure directly affects the ability of the agents to synchronize their various resources. The general goal is to release the organizational energy necessary to maintain or improve the current organizational positioning on the fitness landscape “market”.

Complexity theory is an increasingly popular research field that has its roots in several disciplines, such as physics, mathematics, biology, economics and artificial intelligence. The increasing popularity is due to the fact that most modern science is concerned with specialized fields of research and thus unable to offer an integrative and transversal approach to understanding reality; it still follows the paradigm of reductionism. Modern science is trying to detect patterns that allow predictability, control, stability, repeatability by identifying fundamental causal relationships amongst the systems elements. Complexity in contrast emerges in the non-linear interplay of a multitude of elements that need to be studied holistically. “A non-linear system is one in which the evolution of the phenomenon does not take place by adding elements to each period, but rather by multiplying them. Probabilistic, non linear dynamic systems are still considered to be deterministic…At different times the same complex system can produce chaotic or orderly behavior” (Baets and Van der Linden, 2000). Thus reductionism is unable to explain emerging system dynamics. To understand the emerging behavior of units that include individual organisms as well as the largest economic, political or social institutions, complexity theory often uses agent based approaches to grow this behavior “bottom up”. “Research in artificial context provided us with the insight that, instead of reducing the complex world to simple simulation models which are never correct, one could equally define some simple rules, which then produce complex behavior” (Simon, 2004).This approach allows to program agents that co-ordinate, synchronize and auto-organize without the help of a central leader. Characteristics of emergence, auto-organization, dependency on initial conditions and time boundedness can be modeled with the help of agent based simulation tools (further called ABS) that rely on the power of parallel programming. Agents, equipped with simple rule-sets interact semi-autonomously while trying to optimize their individual fitness given certain restraining environmental conditions the agent is able to perceive. The implementation of certain feedback loops allows the agents to learn, adapt and evolve in the run of the simulation. The application of different algorithms enables us to model evolutionary systems with the help of ABS. In general, the utilization of ABS in the field of research on organizational structures and the dynamics of diffusion of innovation, knowledge and learning are very promising research areas and have witnessed huge interest lately (Carley, 2001).

The role of information and knowledge sharing in today’s organizations is well researched. However, employees and managers face fundamental problems performing in a fast changing and increasingly complex working environment. On a personal level they are overwhelmed with a multitude of incoming information in form of e-mails, meetings, phone calls, visio-conferences, letters, regulations, presentations and documents. All this information needs to be dealt with in the frame of the limiting factors of time and attention-span. On the corporate level the co-ordination, increasing specialization and segmentation of the different organizational functionalities require intelligent and efficient management of work- and information flows.

Core business processes, hierarchical structures, human behavior, regulatory frameworks, information and communication technologies, structures of human and social capital in combination with influences of organizational culture and “politics” are interdependent factors, that shape the underlying dynamics of organizational adaptation and change. In other words, the interplay of these multiple interdependent factors means that organizations are dynamic, non-linear, hence “complex” adaptive systems (CAS). Complexity theory and its implications on current management techniques and paradigms is not a new field of investigation (Baets, 1998, 2005, 2006; Simon, 2004). The particular focus of efficiency of information use, though it has been labeled by Davenport and Prusak (1997) a while ago as an “ecological” information approach, has not yet been understood within the frame of an ever changing company, indeed a complex adaptive system. Focusing on the dynamics of information that flows inside the complex adaptive organization (CAO) and the systemic interconnectedness of its influencing factors, it is appropriate to speak about an “information ecological” approach to information management.

Corporate information ecology, that is, a more ecological study (less resource consuming, less inefficient waste of energy) of the information environment of an organization, is consisting of diverse, interdependent, cultural, social, technical and political subsystems. The metaphor of ecology highlights the notion of (co-) evolution and the embedded, networked nature of information in the organizational overall system. In complex - and especially in social – networks, the behavior of actors emerges in autonomous, non-linear interaction with a variety of other actors. The ecological view-point puts the focus back on the behavior of employees as information actors. They participate as key constituent elements in a larger organizational network of subsystems in which information is created, in which it flows and in which it is being used. Thus, an ecological paradigm seems to be promising for identifying a new approach to deal with the problems related to information overflow and other dysfunctionalities related to information management.

This paper attempts to study this in more detail. After a brief introduction in complexity theory and agent based simulation, the third section develops the concept of a model for information ecology. Section four deals with a conceptual model that is next translated into an agent based simulation, described in detail in section five. Section six discusses the results of the simulations and draws some first conclusions.

Despite the importance for organizational effectiveness, competiveness, innovative potential and adaptivity, our understanding of the basic laws governing human communication and especially decision making in organizations (hierarchic systems) remains limited.

There is a lack of tools to monitor dynamics of human interaction processes, knowledge sharing and the quality of decision making. This paper is an attempt to model these communication processes inside complex dynamic systems with the help of an agent based simulation.

From a complexity perspective a system is developed, consisting of a network of information agents with individual traits and communicative competencies that interact with each other in order to solve problems imposed onto them by an external dynamically changing competitive environment. In several simulation runs emergence of characteristic interaction (collaboration) patterns and communication structures could be witnessed.

Certain network constellations seem to correlate with distinct levels of organizational performance, measured in terms communicative efficiency and robustness, in terms of speed of problem solving, human- resources involved and quality of decision making. Interestingly there seem to be general patterns of successful and efficient communication, independent of the size and complexity of the organization, quality and competencies of its employees and the dynamics of external market change. In regards to a fundamental understanding of the functioning of social communication processes inside organizations this paper promotes an ecological approach to human systems management.

I have long thought about how to manage and represent my research on the phenomenon of what I like to call "complex adaptive organizations". Since I first got in touch with "out of the box thinkers" during my stay in Marseille, South of france, the interest in understanding the dynamics of complex systems and emergence has not ceased to grow. In order to introduce you to my way of thinking (which is pretty interdisciplinary, courageous and sometimes naive), this blog will contain results of my research on the concept of information ecologies. How can we discover the most efficient levers to foster innovation, conduct successful change management and create knowledge management strategies that "work". What are the ingredients to a operational approach to manage complexity? What is a people based, truly social way to do KM? Well, I try to shed some of my insights into the subject matter and I hope that those that like what they see leave a little message