Emergent learning definition

The Systems Thinker – Emergent Learning: Taking "Learning From Experience" To a New Level

Afundamental paradox of working in today’s fast-paced organizations is that we don’t have time to make mistakes, but we don’t have time to avoid them either. Our jobs have become a blur. We cringe when we see ourselves falling into the same traps over and over. We groan in frustration when we find out that three business units are deep in the throes of reinventing the same wheel. Or we experience a stunning success, but we don’t have the time to figure out what made it possible.

In an attempt to capture learnings, we make our best efforts to take time out to reflect. For example, we may institutionalize project “postmortems,” or have an internal consultant study and document lessons learned. Or, we may focus on the “front end” by conducting training in balancing inquiry and advocacy, understanding systems archetypes, or engaging in dialogue.

All of these approaches have the potential to shift us out of our reactive ruts. But they do not automatically become part of an organization’s working habits; we must devote time, resources, and infrastructure to tend to and nurture them. More often than we care to admit, “lessons learned” collect dust on the shelf because we just don’t have the time to translate others’ hard-won insights into our next high-priority project. And sometimes our new reflection skills and techniques are just “out of sync” with our workflow — we don’t have time for them when we need them, and when we do have time, other priorities beckon us.

THE EMERGENT LEARNING PROCESS

“Learning from experience” is mostly done retrospectively. Engaging in emergent learning means taking an intentional, evolutionary approach to learning “through” experience — by conducting iterative experiments using a group’s real work as the experimental field. Taking this approach often produces new and powerful learning simultaneously to making headway on key business issues

Emergent learning practices offer us a pragmatic, low-overhead approach to making the time and space for organizational learning habits to grow. In the process, they help teams and business units develop “islands of mastery,” or growing areas of expertise in their increasingly complex working environments. And the practices help sponsors identify incremental wins and build a business case for the value of organizational learning.

What Is Emergent Learning?

Emergent learning is the ongoing exploration of a locally defined arena of action through intentional, iterative learning experiments. The goal of emergent learning is for a group of people — perhaps a team or business unit — to master performance in arenas of key importance to their business. The focus of these learning experiments might be improving the organization’s ability to fulfill its basic mission (such as, for a police department, reducing crime), managing escalating costs, creating successful strategic alliances, or bringing projects in on time and under budget. An experiment might involve comparing two recent strategic alliances, forming conclusions about these experiences, and testing the conclusions on a new project. Or for a group of project managers, an experiment might mean getting clients involved in projects at different times and in different ways to see how these variables affect the decision-making process But in each case, the two characteristics that distinguish emergent learning from how we usually approach simply “learning from experience” are that it is iterative and intentional. Teams repeat emergent learning experiments in parallel or in close enough succession to be able to compare and contrast performance from instance to instance. They purposefully define experiments in advance of the experience, not in retrospect, as in a “post-mortem.” These intentional iterations make learning from experience active and evolutionary, rather than a static, one-time review.

Simply put, today’s working environments are often too complex and fast moving to give us the time and space we need to focus our full attention on learning. Consequently, the practical reality for many of us is that only those learning practices that require little time will actually take root (see “Rethinking Time” by Peter M. Senge in The Dance of Change, Doubleday/Currency, 1999). By weaving learning into the real-time priorities and real work challenges of a business unit or team, an emergent learning approach bypasses the need to stop what we’re doing in order to learn

In fact, a team may develop extraordinary emergent learning practices without ever thinking of it as “learning.” Emergent learning often looks a lot more like locally driven strategic planning or problem-solving than like what we usually think of as training. Groups self-organize to focus on improving their performance, rather than stepping into a classroom setting where the attention centers on the instructor’s expertise. On the other hand, because of its iterative nature, it differs from what we traditionally think of as planning or problem-solving by focusing on mastery (performance over time), rather than on accomplishment (performance today) (see “Comparing Training, Planning, and Emergent Learning” on p. 3).

Emergent Learning in Practice

Here’s an example of an emergent learning process based on a group’s real work needs and conducted in real time: The executive team of a large regional vocational school expressed its frustration at once again needing to downsize because of escalating costs. In years past, members had rolled up their sleeves and done the painful work of identifying possible staffing and program cuts. When all was said and done, they had at least felt a sense of accomplishment at having taken hard but necessary steps to solve the problem.

After the third downsizing this decade, they made a determined effort to escape from what they had come to see as a vicious cycle by taking steps to shift their focus from short-term crisis resolution to developing long-term solutions through emergent learning.

The team defined an arena on which to focus: its cost structure. Facing obvious and painful failures in trying to solve recurring financial problems, members recognized how little they really understood their costs. They made a commitment to “master” the cost arena — to develop a richer, shared understanding of what drives costs, and to be able to consistently manage them. They had a discussion to articulate the key variables or criteria that would indicate success in this arena.

The team then identified a few repeatable contexts that could easily provide opportunities for reflection: weekly staff meetings and executive reporting. Because these activities were already on their plates, they provided a relatively quick and easy way for team members to test their mental models about what was driving costs. Because they were recurring, the group could easily review the results of experiments that they planned to conduct on a regular basis, and gradually evolve a real mastery of the issue.

This process may look like nothing more than good problem-solving. But it demonstrates a subtle shift from accomplishment to mastery

To get started, team members shared their beliefs and understanding about what contributed to the school’s cost structure. Then they very deliberately turned these statements into hypotheses to test in learning experiments. Each member considered what projects he or she was involved in or what data he or she had that would serve as the basis for conducting experiments. For example, the head of programs was curious about whether his assumptions about the direct relationship between class size, perceived program quality, and costs would hold up. The head of facilities had questions about whether previous cuts in headcount might have actually resulted in increased maintenance costs.

Initially, they simply added brief reviews of cost trends (such as compensation, legal fees, and supplies) to their weekly meetings, and a discussion of 12-month cost patterns to the monthly and quarterly executive reports. Over time, through several iterations, they began to see new relationships and investigate such dynamics as the relationship between facilities maintenance, compensation, and legal costs. In staff meetings, they reflected on the potential causes of changes in costs and described experiments that they had tried. (At one meeting, the head of facilities reported about asking his team what they would do if he went on sabbatical for a year. The creative responses that he got inspired some of his peers to try the same experiment. )

At each iteration, the results of just-completed learning experiments became the “ground truth” on which they reflected in order to plan for the next learning experiments (see “The Emergent Learning Process” on p. 1). With the benefit of their peers’ perspectives, team members teased out unspoken assumptions, lessons learned, and so on. They began to question the measures that they had relied on in the past and realized that they needed more powerful and timely cost indicators. They acknowledged how delays in feedback — in the form of unanticipated cost increases — affected their ability to manage expenses. These sessions inevitably led to new questions and new experiments.

COMPARING TRAINING, PLANNING, AND EMERGENT LEARNING

Beyond Problem-Solving

This process may look like nothing more than good problem-solving. But it demonstrates a subtle shift from accomplishment to mastery. With this new mindset, everyone on the school’s executive team worked under the assumption that they would run through the learning cycle at least several times. Over time, as they cycled through iterations of this process, their learning experiments got more specific and they asked better and better questions. They also developed finer distinctions about costs and the dynamics that cause them to rise. In addition, they identified early indicators that a problem was brewing. As a result, their sense of confidence in being able to tackle something as complex as escalating cost structures grew.

On the other hand, if the team had continued to focus on problem-solving rather than on learning, they might have replaced downsizing with another, perhaps equally short-term, “solution.” By simply abandoning their first approach to the problem, they may have failed to develop a true understanding of why downsizing did not solve the problem. Or they might have chosen to “downsize harder,” triggering even steeper cost problems as the school struggled with the loss of skilled personnel

By taking an emergent learning approach, the team also created a compelling context for drawing on the tools of organizational learning. For example, they began to see that they had fallen into a “siege” mentality regarding saving their favorite function from the chopping block. So the group sought training in balancing inquiry and advocacy, recognizing that their ineffective communication habits were affecting their ability to explore alternative theories and solutions. They also studied systems thinking to begin to grasp the drivers of costs and to understand the behavior of reinforcing processes. In this way, they developed expertise as they needed it and as it made sense for addressing their current business challenges — not as it was deemed necessary by a training department or corporate mandate.

Simplicity and Localness

The best emergent learning practices track a few simple variables within an experimental field that is as local as possible. In the example above, the executive team initially tracked operating costs (variables) within the different departments (experimental fields). Each participant made a series of small changes to the work that they were already doing in these areas.

Over the long term, these intentional, iterative experiments at the operational level often generate new and unpredicted, but remarkably powerful, changes in behavior. For example, the Boston Police Department uses simple three-month charts of major crimes, district-by-district, to understand and influence crime trends, such as a rise in burglaries in a particular neighborhood. Over time,

this disciplined approach to managing crime has inspired district police to go out of their way to meet local teens and attend community meetings, not because it’s their job, but because they see that making a personal connection is critical to grasping what is fundamentally driving trends in crimes.

The U. S. Army’s After Action Reviews (AARs), which emerged from its intensive two-week training simulations in the Mojave Desert, are another example of a practice that is so simple and local in design that it spread on its own, without being mandated from above. In an AAR, soldiers take an hour after a military encounter (simulated or real) to analyze what caused any differences between what they intended to accomplish and what actually happened. In addition, they identify strengths to sustain and weaknesses to improve in the next encounter. AARs have become so ingrained in the organization’s culture that almost anything is now seen as a learning opportunity—, “Let’s AAR that.”

Committing to Learning Experiments

As shown in these examples, opportunities for emergent learning are everywhere. The seeds for it can be found in what Barry Dym calls “forays” — small, local initiatives that are exceptions to the more established patterns of working together (see “Forays: The Power of Small Changes” by Barry Dym, V9N7). They can also spring up in “communities of practice” — informal groups that join together to develop a shared repertoire of resources. To reap the benefits of emergent learning, members of these groups must shift from following the traditional professional association model — holding abstract conversations based on expert presentations — to making the commitment to study their own performance in a concretely defined field of experiments (see “Communities of Practice: Learning as a Social System” by Etienne Wenger, V9N5).

Nortel Networks’ Competitive Analysis Guild (CA Guild) is an apt example of a self-organized community of practice that has been able to make that shift. The CA Guild gathers members from across organizational boundaries to share knowledge about Nortel’s competitors and build their competitive intelligence skills. Guild membership outlives project assignments and creates a “virtual neighborhood” of likeminded individuals.

Some Guild practices look like those of traditional professional associations: monthly meetings with formal presentations and a Web site with announcements of upcoming events. But the Guild has also created some activities that are developing emergent qualities. For example, any Nortel Networks employee can use the Guild Web site to seek information about competitors from members. The sharing of questions and answers through the network is an iterative process. Participants have reported that they have become more sensitive to early indicators of important actions by competitors.

The Guild also views industry trade shows as a natural experimental field. At any given industry trade show, there may be 30 or more Nortel Networks employees wandering the floor. The Guild developed a procedure to focus these employees on a learning agenda. After each show, not only does the Guild take away good data, but it also reflects on and refines its trade-show practices. Over time and with iterations, this approach turns good intelligence-gathering into emergent learning.

Islands of Mastery

Peter Senge has commented that, “I have never seen a successful organizational learning program rolled out from the top. Not a single one. Conversely, every change process that I’ve seen that was sustained and that spread has started small. Usually these programs start with just one team” (Fast Company, May 1999). Emergent learning builds the organizational learning “habit” from the bottom up, by focusing a team on mastering performance in an arena that is important to them. The venue may be big and “strategic,” such as demonstrating leadership during a merger, or it may be small and “tactical,” such as planning food for faculty meetings. Whatever the level, as the team disciplines itself to focus its attention on its performance in this one arena in an iterative way, a lot of what previously seemed like erratic, unpredictable results can begin to make sense (see “Conducting Learning Experiments”).

Emergent learning builds the organizational learning “habit” from the bottom up

And so, an island of mastery begins to emerge from the sea of complexity. And as one arena of action starts to make sense, the group naturally expands its field of inquiry into other arenas. In turn, team members’ confidence in being able to master their business challenges grows. They become better able to clarify their priorities, articulate their own theory of success, test their hypotheses, and make a strong case in support of their thinking.

This self-reinforcing cycle of curiosity and growing competence can have an almost addictive quality — it makes people thirsty to learn more. As people develop a learning discipline and begin to search for fundamental solutions, they almost automatically take a systems perspective, collaborate more effectively with others, and challenge their existing mental models.

In this way, pairing emergent learning practices with traditional training can help the tools and techniques of organizational learning find a natural home. As internal and external practitioners, we can look for opportunities to turn events and projects that we are currently working on into learning experiments. We can do more to identify and support naturally occurring emergent learning practices, and make it a priority to notice and publicize results. And we can also help business units, teams, and communities of practice create new emergent learning practices. In the process, we will build natural advocates for organizational learning, complete with their own compelling stories to tell.

CONDUCTING LEARNING EXPERIMENTS

Practices like these can be found germinating in many corners of any corporation. You may be able to identify — and build on — many naturally occurring examples of emergent learning in your own organization. But you can also begin the process of developing your own emergent learning discipline by following these steps: 1. Identify an arena of action that is critical to the success of your business unit or team; for example, having effective meetings, given that your team members are spread across time zones and rarely meet face-to-face.

2. Articulate a few simple key variables or criteria for success in that arena; for example, shared understanding, measured by tracking the agreements that are kept and those that fall apart.

3. Identify processes or events that are already on your plate and that repeat on a fairly regular basis, such as video-conferenced project meetings.

4. Start with a hypothesis, mental model, or question about success in that arena; for example, “If we actively make room for dissenting opinions up front, the quality of follow-through on agreements will increase. ”

5. Define a simple experiment to test your hypothesis that you can “slip” into an existing event or project without a lot of extra design effort; for instance, each time a decision is about to be reached, you (as a team member) can ask, “Is there anyone who doesn’t feel heard on this yet?” Make some predictions about what you expect to see as results; for example, within two meetings there will be an absence of the usual “Well, I didn’t really agree with that anyway” when a slip-up is discovered.

6. Plan when, how, and with whom you will study the results. Meet between repetitions of selected experiments so that you can assess the results and apply what you learn to the next iteration. For example, as a part of planning each meeting, three project managers may briefly review the “ground truth” from the last experiment and discuss their conclusions. In this case, the number of agreements kept may have improved, but now the meetings run long.

7. Iterate the process, starting with step four. “So, given our understanding of how time constraints and the keeping of agreements are related, how can we adjust our hypothesis about how to achieve both?”

What is Emergent Learning?

Welcome to the era of emergence

In systems theory, emergence refers to the presence of any system of special properties that are not inherent in its elements (both individually and in their sum). Elizabeth Comstock, vice chairman of General Electric, talks about how emergence affects our lives and work, and what information technology has to do with it.

The speed of information movement increases, and we move faster after it. More and more people are going online (more than 2.3 billion in 2016-2017 alone), and the result is a fundamental and spontaneous restructuring of our collective behavior. The top layer of our evolving digital nervous system on a planetary scale has absorbed the eternal factors of change - human needs, politics, geography and culture - and created new schemes and patterns based on them. Each of us, especially those who run companies, needs a new frame of reference.

An important part of this coordinate system is emergence, the principle of self-organization, which explains the development of the complex from the simple and the emergence of order from chaos.

Emergence explains complex phenomena in various fields such as thermodynamics, biology, and digital technology. This concept is used to explain the principles of urban development and why some institutions function well in the long term, while others disappear. This is the basic principle behind the complex collective behavior of natural systems, which partly explains how computers achieve their useful complexity despite the fact that their work is based on a few simple processes.

We live in an era that I call the "Age of Emergence."

This time is characterized by the rapid decline of traditional institutions, despite the fact that their replacement has not yet taken shape. We live in a chaotic, sometimes disturbing and uncertain space between the old and the new.

If we successfully pass this period, we may find ourselves in the historical equivalent of a moment in the creative process when chaos gives way to order and a new pattern, solution, or structure becomes apparent.

By adopting and supporting emergent systems in business, politics, and technology, we will be able to unleash human potential to a greater extent than ever before in history, and achieve benefits never before possible.

I will not claim that these are purely my ideas. I synthesized them from common themes and views found in the work of other specialists in various fields and industries, adding my own conclusions to them. In fact, in this new age, it is especially important to develop the ability for such a synthesis of individual points of view, data sets and conclusions. The concept of the "Age of Emergence" appeared in my thinking just as spontaneously and unexpectedly as the phenomenon it explains.

Two sides of emergence

Emergence, as I call it, has two key aspects.

The first is that we are currently dealing with a huge number of emergent systems and technologies. There is not a single area of ​​human activity that is not now at a crossroads. Each of them has many options for future development, represented by new discoveries, business models or inventions.

The second aspect of emergence is stranger and better describes our historical reality. An emergent system, as stated above, is a system in which order is born out of chaos. It is also a system, the power and structure of which is created by a network organization, and not determined by a hierarchy from above. The individual components of any emergent system operate on the basis of relatively simple rules, but when they interact with each other en masse, they can develop into complex adaptive structures.

The countless individual neurons that connect in the human brain, the many birds in a flock, the individual ants in a colony, and the rows of electrical circuits inside a computer are all examples of simple objects working together to create something more than the sum of the parts.

In the business world, emergence explains why some tech start-ups suddenly get super-large market valuations and expand their functionality incredibly quickly. It's not about the app or site that creates value, but about the collective behavior of the people who use it.

By the end of 2016, the number of Facebook users has grown so much that it is no longer considered just a social network. With more than 2 billion users, the company has become a de facto political organizer, a major (albeit highly problematic) source of information, a global video distribution platform, a marketplace for goods, and even jobs.

The versatility of Facebook relies on the size of this network. Once a system that is capable of reproducing emergent behavior grows to a certain extent, applying the same rules that governed its operation before begins to lead to very different and sometimes useful results.

A classic example in nature is an anthill. As noted biologist Edward Osborne Wilson, the social behavior of ants is the most difficult after humans. At the same time, these social structures arise as a result of simple behavior. A colony in which ants track and redraw the hormonal traces left by other individuals during their movement, creates anthills, develops an effective form of using all available food supplies, learns to protect their homes and territory, migrate and survive in extreme conditions. Some ants do not obey the rules, but when the population grows to a certain size, its collective intelligence and power become a commanding force of a much higher order. If we talk about ants, then the more of them, the smarter they are. The emergent mind has allowed them to survive virtually unchanged for the past 9 years.0 million years.

But here's what's weird. We understand how emergent systems work in nature only because we first noticed that similar principles make our computers smart. As Steven Johnson writes in his book Emergence, the discovery of the properties of emergence occurred at the intersection of the ideas of Alan Turing, the founding father of modern computing, and Ilya Prigogine, the pioneering chemist. Digital systems are also emergent systems by structure.

This means that as we move our information, money, goods, and services through digital systems, those systems are more likely to have beneficial emergence.

Residents of Europe and North America may consider that such a collective reorganization around the flow of digital information has already occurred to them. But this is a misperception caused by the high visibility of the industries that were the first to go digital — music, finance, and journalism.

In reality, the digital transformation has only just begun. According to the McKinsey consulting agency, 82% of US businesses have not fully realized the potential of digital technologies. Some of our largest industries (such as manufacturing and healthcare) are actually the least digitized. The potential for digital-driven growth is even greater in developing countries.

The scale of the digital transformation we are about to experience is staggering. The number of Internet users is projected to rise from the current 3.6 billion to 5.8 billion over the next year and a half, with mobile devices accounting for most of this increase. In other words, 1/3 of the world's population will go online for the first time in the next 18-24 months. Add to that the estimated 50 billion machines that will be constantly communicating with each other over the same network by 2020.

On the business side, the environment created by this large-scale shift will allow the most valuable companies to lead the reorganization of existing assets and experiences built around the flow of digital information indefinitely.

Two of Silicon Valley's most successful startups, Uber (worth $62 billion at the time of writing) and AirBnB (worth $25 billion) are just such companies. None of them actually produce the assets they sell, but each of these companies has created a new flow of information that has allowed individual agents (people with smartphones) to respond in new ways to the location of these assets in real time. The interaction of users en masse has led to the emergence of new types of behavior and new markets based on these types.

It's like the ants have a new set of pheromone commands, or like the brain cells have developed a new way of signaling. Emergent systems are extremely sensitive to the slightest change in the original operating conditions. The reconfiguration of these conditions will lead to the spontaneous appearance of new structures.

If you know what these structures will look like or, even better, if you can create and shape the disruptive forces that will define them, you will gain an invaluable strategic advantage.

When it comes to ubiquitous and more easily understood assets like cars and real estate, startups will have this advantage. But for heavy and specialized industries, it is extremely important to possess not only specific specialized tangible assets, but also in-depth knowledge of their production and operation.

Acceleration and (re)distribution

According to the World Economic Forum, the technological advances of the past few decades have lifted half of the world's population out of poverty.

This is an extremely rapid change, which we also see in real time. It has positive results, but at the same time it increases the level of anxiety in society. Increased migration of people and the escape of the population from poverty with the help of capital can also have unpredictable and obscure consequences, like BREXIT and recent US election results. In a bizarre combination of motives, the positive and negative impulses of life in the Age of Emergence are reflected.

On the one hand, people seem to want to get rid of the government bureaucracy, which for the most part is unable to keep up with the times. On the other hand, there is a reactionary desire to separate from the rapidly developing global economy, which often also separates people from familiar values ​​and institutions, even if they do not want to or do not realize it.

We have already gone through similar periods of instability caused by information incentives, although not on the scale and at the speed that we have seen in the last two or three decades. The sudden return of interest in ancient literature and its translations, as well as the powerful flow of information associated with them from the monasteries of Europe, led to the onset of the Renaissance. In the 19th century, the world economy spontaneously built itself around the new flows of information generated by the widespread use of printed literature and the ability to transmit data at the speed of ships and trains, as well as by telegraph. The results of these processes were the transition from monarchy to democracy, which took almost a century, the greater recognition of the rights of women and minorities, the elimination of the global slave trade and the technological revolution.

The Internet of Things (GE's terminology for the Industrial Internet) is the printing press of our generation. This technology weaves a stream of digital information into the material world and promises to endow this material world with the same adaptive emergent functions that we see in ants, neurons, and digital systems.

Some believe that the growing complexity of connected machines is the beginning of a threat that will manifest itself in the future. I prefer to see it not as a threat, but as an opportunity for learning. Our expanded global nervous system is growing and evolving, and we are developing new relationships with it.

As with any new relationship, we have a choice. We can develop them or slow down this process. The best choice in this situation - one that can free us from hard labor and poverty in the future - is to develop, move forward, innovate with digital systems and explore previously unfamiliar areas. This prospect may be daunting, but it holds enormous potential.

Mental states as emergent properties. From Walk to Consciousness. Part 1 (Introduction)

Authors:

Cruz Holke [email protected] University of Bielefeld, Bielefeld, Germany

Schilling Malte [email protected] University of Bielefeld, Bielefeld, Germany

Translated from English Shuravin A.P.

Annotation .

In this article, we propose a bottom-up approach to higher mental states such as emotion, attention, intention, will, or consciousness. The idea of ​​such a bottom-up approach is that higher states can arise as emergent properties, that is, they can occur without an explicit implementation of the material under study. Using a neural architecture that shows the capabilities of autonomous agents, we want to arrive at quantitative hypotheses about cognitive mechanisms, i.e. come up with verifiable predictions about the basic structure and functioning of an autonomous system that can be tested in a robot control system.

We don't want to build an artificial system like consciousness. On the contrary, we want to create a system capable of controlling behavior. Only then will this system be used as a tool to test whether the descriptions of the degree of mental phenomena used in psychology or philosophy of mind can be applied to such an artificial system. Initially, these phenomena must be defined using verbal formulations that allow them to be interpreted in different ways. Functional defined, in contrast, does not suffer from ambiguity because it can be expressed using mathematical formulas that can be tested in quantitative modeling. It is important to note that we are not connected with the "hard" problem of consciousness, that is, with the subjective side of mental phenomena. This approach is possible because, by adopting a monistic view, we believe that we can get around the "hard" problem without losing information about the possible function of these phenomena. In other words, we propose that phenomenality is an inherent property of both full awareness and metacognition (reflexive consciousness). Following these arguments, we argue that our network can not only show emergent properties at a reactive level, it also shows that mental states such as emotions, attention, intentions, will, and consciousness can be observed. With regard to consciousness, we argue that the properties that partly make up total awareness are present in our network, including global properties. This means that elements of procedural memory can be applied even if they are not relevant to the current content. Further expansion of the discussion shows that it is possible to recognize properties that are attributes of metacognition or reflective consciousness.

Keywords.

Total awareness, consciousness, emergent properties, model inner case, minimal cognitive system, motor control, ReaCog, recurrent neural networks, reflective consciousness, robot architecture.

1. Introduction
   

In this article, we propose to approach from below upwards to higher mental states, such as, for example, consciousness. Unlike most related approaches, we do not take consciousness as our starting point, our goal is to first build a system that has the basic properties of a reactive system. In the second stage, this system will be expanded to obtain cognitive properties in the sense of the ability to plan ahead. Only after that the work will be completed, we want to see that this system is to some extent equipped with higher properties, such as, for example, emotions or consciousness. While other approaches require a precise definition, such as a definition of the term "consciousness", in our case we should not start with a clear definition and try to fit a model into it. We follow this alternative route because there are no commonly accepted definitions regarding such higher order phenomena. Thus, we hope to determine the basic elements needed to create, for example, an instance of consciousness.

The idea behind this approach is that higher order properties can arise as emergent properties, i.e. may arise without the need for an explicit implementation of the phenomenon under consideration, instead they arise due to the interaction of elements of the lower level. Some authors distinguish between "strong" and "weak" emergence of emergent properties (eg Laughlin and Pines 2000). By "strong" emergent properties, we mean those properties for which there is no way in principle to explain their occurrence by known properties of the elements of the system and their interaction. Here we are dealing with a "weak" appearance. A property is recognized as such if, looking at the whole system, we can, at first glance, not trace (or, perhaps, only partially trace) the connection of the property with the known properties of the elements and their compounds. Auxiliary assumptions are often used to explain a property as a global property, that is, as a property attributed to the system as a whole. On closer examination, however, it can be shown that such ancillary assumptions are not required. Instead, the emergent property derives from the properties of the elements and the specific ways in which they interact (have causal relationships). This insight allows us to understand the property in the sense that it can be predicted, although we could not understand why it arises, and that someone is able to build a new system due to this property.

Figure 1: Arrangement of foot controllers (boxes: FL front left, ML middle left, HL rear left, FR front right, HL rear right) water strider. The arrows show the coordination influences (1-4) that operate between adjacent leg controllers.

Following this approach, a key decision will be made first regarding the detailing of the lower order elements. In our approach, we start from a behavioral point of view and consider the nervous system as the central control of actions. Thus, we use neural modules as the main elements of our model and also for analysis. In particular, we use artificial neural network modules with analog activation values ​​and dynamic properties (low pass filter). That is, our nerve elements are qualitatively compatible with non-pulse neurons. Although there are arguments that consciousness may require synchronously oscillating impulses in order to emerge (Singler and Gray 1995), we argue that the level applied here is sufficiently general to allow such mental processes to be understood. As a side effect, this level of abstraction covers various evolutionary groups, such as insects and mammals, for example. Although much of our discussion below focuses on the example of insects, we are not going to argue that insects have all of the higher-level properties that we will explore later in this article. We only claim that they share the same basic functions used in engine management. This is, at this level, comparable structures.

Using these simple neural elements, we will begin by implementing the basic functions that include the ability to move one's own body and allow orientation and navigation in a partially known environment. To do this, we use a body with six limbs, like insects. This means that we are dealing with at least eighteen active degrees of freedom (DOF), and not two, as is the case for many robots, which are limited in moving in a circle on a two-dimensional plane. This means that the controller is dealing with a lot of excess degrees of freedom. To control the behavior of the robot, we use a reactive and embodied neural controller, as available in earlier work on insect behavior (Schilling et al. 2013a). Later, a slight expansion of the network will allow the realization of cognitive abilities.

What are the properties of the reactive/cognitive system considered here? The reactive system is called "Walknet" and it is based on biological principles drawn from experiments on the locomotion behavior of sticky insects (Dürr et al. 2004; Bläsing 2006; Schilling et al. 2013b). As will be explained in Section 2, Walknet was created as a system to control the behavior of a walking six-legged robot in an unpredictable environment, such as cluttered terrain or steep climbs, when performed in a realistic, natural environment. This is a non-trivial task. Already at this level we can observe emergent properties. The number of feet on the ground varies depending on the speed of the walker (in case of slow walking, more feet are on the ground). As a consequence, the phase relationships between different legs differ depending on the speed of the walker. It is important to note that the results of stepping patterns (“gaits”) are not explicitly encoded in the control network, but are the result of the interaction of the control network with the environment, as mediated through the body (1st order in the Metzinger 2014 variant). In the next step, the reactive controller is extended to be able to handle navigation tasks. This complementary network, called "NaviNet," makes it possible to model a range of experimental outcomes seen in desert ants and bees, such as the ability to find food sources using path integration and visual cues.

Both networks are decentralized. These networks are composed of procedural (reactive) elements, which are called small neural networks, which are connected to sensory input and motor output, thereby forming a procedural memory. Inspired by these procedural elements (Maes, 1991), coupled with a motivational module and a feedback neural network, we formed the basis of a complete system. This type of architecture has been called MUBCA (Motivation Unit Based Columnar Architecture). The motivational block of the neural network allows you to select different behaviors that take on different states of the attractor, where each attractor represents a group of motivational blocks that are active, they include control over procedural elements. As different groups do in partial overlap, albeit in different ways, the network makes it possible to represent a heterarchical (heterogeneous) structure (for example, see the top left of Figure 2).

Figure 2: Reactive system control network. The motivational module (red) is formed by RNNs that can take on different attractor states (only two leg controllers are shown). Arrows show excitatory influences, T-junctions show inhibitory influences (fw forward, bw backward, r1 rule 1 coordination). Motivational blocks in the control procedure of the lower level (boxes, for example, swings, racks). The procedures include an internal body model (blue). The body is marked with a dotted line (“legs”). The network specified here controls the direction of NaviNet movement (see Figure 4 for more details).

As the next "evolutionary" step, this reactive network will be expanded to be able to cover cognitive properties (sects 3 and 6). The concept of cognition is often used in a broad and sometimes not very specific way. In what follows, we will build on the definition given by McFarland & Bösser (1993), who believe that the cognitive system is characterized by the possibility of forward planning. We prefer this clear definition of cognition over many others found in the literature, as the latter is generally rather weak (in extreme cases, cognitive properties have even been attributed to bacteria, which would, in our opinion, make the term cognition meaningless). While such a specific definition may seem too narrow, in our understanding it captures the essence of cognition. Focusing on forward planning, realized through mental modeling (Hesslow 2002), allows us to expand this concept of cognition and easily include other phenomena of a higher order, while still relying on the same internal mechanism. Thus, in this article, apart from section 10c (Metacognition), we will use the term "cognition" in the strict sense proposed by McFarland & Bösser (1993).

The pre-planning state implies the ability to internally mimic behavior, which basically means the ability to mimic the movements of one's own body in a given environment. This ability requires, as a first step, having a flexible internal body model. Forward planning is interesting in a situation where the implementation of reactive behavior does not actually achieve the pending goal. Thus, further expansion requires the ability to invent new behavioral strategies. Together with the ability to plan ahead, the system can test newly invented behavior strategies by applying internal simulations (internal "trial and error") in order to find a solution for new problems for which the solution is not currently known to the system. [3]

This system, called "reaCog", is the basic version of the cognitive system in the strict sense of intent from McFarland & Bösser (1993). Thus, cognitive extensions do not work on their own, but only as an add-on on top of reactive structures—this view has been maintained for a long time (Norman & Shallice 1986). The cognitive system depends on its reactive basis (which is why it is called reaCog). Thus, the evolution of cognitive abilities fundamentally requires a correspondingly capacious (procedural) memory.

In order to increase the memory richness of the entire system, in section 5 we will introduce memory perceptions and complete the system by implementing a "Semantic Web", a specific form of process perception and memory. Thus, the entire system is equipped with elements of semantic memory, and they can be required to represent a minimal cognitive system. We are not engaged in training, but only discussing the properties of the finished network. The assimilation of some aspects, however, have been tested previously (Hoinville et al. 2012; Cruz & Schilling 2010a).

After introducing reaCog in sections 2-6, we will discuss in sections 7-11 how more abstract functions, such as those described in psychology, for example, can be based on such a simple structured network. The fundamental problem lies in the desire to understand phenomena, such as emotions or consciousness, with regard to the phenomenal aspect. The phenomenal aspect, often characterized as a difficult problem (Chalmers 1997), refers to a strange, inexplicable phenomenon: physical systems, in our case in the form of specific dynamics of neural structures, can be accompanied by subjective experience. Mostly, they may experience pain, a sense of color, or an internal state of emotion (eg, joy, fear). In Section 7 we will discuss this aspect in some detail and assume that the phenomenon is an emergent property. As already mentioned, we do not seek to solve a difficult problem (Chalmers 1997), but we think it's enough to focus on the functional aspect.

In particular, we focus on the phenomena of emotions and consciousness. According to a number of authors (for example, Valdez and Mehrabyan 1994), they are considered an essential property for some cognitive systems. Therefore, although we do not claim that emotions (Section 8), attention, will, intention (Section 9), and consciousness (Section 10) must necessarily be referred to our system in any sense, we are going to discuss to what extent properties characterized by different levels of description can be observed in our model.

Given that emotions are defined at different levels in the literature, there is no clear, generally accepted distinction between concepts such as emotions, moods, motivations, dominants, etc. , which seem to form a continuum of overlapping, distinctly non-separable concepts (Perez et al. 2012). Focusing on individual examples, in Section 8 we will show how these phenomena can be related to our system, for example by referring to basic emotions as proposed by Ekman (1999).

Concerning consciousness, as noted in Cliremans (2005), this phenomenon should be considered by differentiating various aspects and treating these aspects separately. To do this, the next block (1995, 2001), Cliremans (2005), introduces a distinction between access to consciousness, metacognition, and phenomenal consciousness. In sections 10a (consciousness access) and 10c (metacognition) we will focus in a certain way on the presented model, which can be related to various aspects that Cliremansa (2005) describes, such as consciousness access and metacognition. From our point of view, the simple control of the structure is the performance of some aspects of both consciousness and metacognition.

Learn more

• 
  Robot abc songs
• 
  Fun spelling games for 3rd grade
• 
  List of adjective in english
• 
  Teaching short vowel sounds kindergarten
• 
  Creative activities to do at home
• 
  Story telling nursery
• 
  The three bear song
• 
  Letter recognition for preschoolers
• 
  Fairy tales songs preschool
• 
  Goldie and three bears
• 
  Learn to read for children