On Analyzing Self-Driving Networks: A Systems Thinking Approach

Editor's Notes

• #8: Social systems are not linear but belong to the class of systems called multi-loop nonlinear feedback systems. Level two chaotic systems (Harari) “History cannot be explained deterministically and it cannot be predicted because it is chaotic. So many forces are at work and their interactions are so complex that extremely small variations in the strength of the forces and the way they interact produce huge differences in outcomes. Not only that, but history is what is called a ‘level two’ chaotic system. Chaotic systems come in two shapes. Level one chaos is chaos that does not react to predictions about it. The weather, for example, is a level one chaotic system. Though it is influenced by myriad factors, we can build computer models that take more and more of them into consideration, and produce better and better weather forecasts. Level two chaos is chaos that reacts to predictions about it, and therefore can never be predicted accurately. Markets, for example, are a level two chaotic system.  A complex adaptive system, as distinguished from a complex system in general, is one that can understand itself and change based on that understanding. Complex adaptive systems are social systems. The difference is best illustrated by thinking about weather prediction contrasted to stock market prediction. The weather will not change based on an important forecaster’s opinion, but the stock market might. Complex adaptive systems are thus fundamentally not predictable.
• #9: Us vs. Them Mentality; Vested Interests The systems thinking worldview dispels the “us versus them” mentality by expanding the boundary of our thinking. Within the framework of systems thinking, “us” and “them” are part of the same system and thus responsible for both the problems and their solutions.
• #12: “From a very early age, we are taught to break apart problems, to fragment the world. This apparently makes complex tasks and subjects more manageable, but we pay a hidden, enormous price. But, as physicist David Bohm says, the task is futile—similar to trying to reassemble the fragments of a broken mirror to see a true reflection. Thus, after a while we give up trying to see the whole altogether.”
• #15: Unfortunately, our problem-solving instinct also creates a number of followup problems and networking systems (including future self-driving networks) are not immune to this tendency [40]. Systems thinking can help us anticipate and avoid the negative consequences of well-intentioned solutions. This can be done both prospectively by anticipating unintended consequences during strategic planning or retrospectively by understanding more deeply the non-obvious causes of existing chronic complex social problems. 
• #16: As we move our perspective from the event level to the structural level, we have a better understanding of what is really going on in a system. We also increase our ability to influence and change the system’s behaviour, i.e., we can make adjustments to the structure which are consistent with the behaviour we would like to produce.
• #17: As we move our perspective from the event level to the structural level, we have a better understanding of what is really going on in a system. We also increase our ability to influence and change the system’s behaviour, i.e., we can make adjustments to the structure which are consistent with the behaviour we would like to produce. As pointed out by H. L. Mencken,
• #18: 2) The harder you push, the harder the system pushes back.  This is, in systems thinking parlance, “compensating feedback”. This happens in conversation, government programs, business and personal wellness efforts. In conversations we often try to argue our point by disagreeing with the other person. Our “push” helps them strengthen their position. We make them think and fight harder. Social programs are rampant with examples of community improvement initiatives that resulted in worse conditions. Government aims to improve the living conditions for a group in one part of town result is more people moving into the area, placing an unsupportable burden on the systems in place.  4) The easy way out usually leads back in.  Kaplan called this the law of the instrument saying ”Give a small boy a hammer, and he will find that everything he encounters needs pounding”. Maslow reframed this saying “If all you have is a hammer, everything looks like a nail”.  “In a modern version of an ancient Sufi story, a passerby encounters a drunk on his hands and knees under a street lamp. He offers to help and finds out that the drunk is looking for his house keys. After several minutes, he asks, "Where did you drop them?" The drunk replies that he dropped them outside his front door. "Then why look for them here?" asks the passerby. "Because," says the drunk, "there is no light by my doorway.” Excerpt From: Peter M Senge. “The Fifth Discipline: The Art and Practice of the Learning Organization: First Edition.” iBooks.  “Pushing harder and harder on familiar solutions, while fundamental problems persist or worsen, is a reliable indicator of nonsystemic thinking—what we often call the "what we need here is a bigger hammer" syndrome.”  5) The cure can be worse than the disease.  This is also called “shifting the burden” and is easy to confuse with the push/push back law. It’s slightly different though and can occur at the same time. The “cure” in this case is an intervention that is enabling and becomes addictive. As dependence on the intervention increases the system’s ability to cure itself lessens. This is about the difference between giving a man a fish and teaching him how to fish. If an intervention is needed then we have to make sure the intervention doesn’t weaken the entire system causing more and more dependence. In some ways public education shifted the burden of teaching children from parents to teachers. Engaging stakeholders in defining problems and finding solutions keeps the burden where it belongs, shared across the entire system and not just on one part of it.  “ The phenomenon of short-term improvements leading to long-term dependency is so common, it has its own name among systems thinkers—it's called "Shifting the Burden to the Intervenor." The intervenor may be federal assistance to cities, food relief agencies, or welfare programs. All "help" a host system, only to leave the system fundamentally weaker than before and more in need of further help.”  7) Cause and effect are not closely related in time and space. How many times do you push an elevator button? How often have you over compensated for the amount of cold water in the shower? We tend to believe that when we do something there should be an effect that we can see within a set amount of time. All of our testing, funding and business practices reflect this belief. The challenge is that sometimes there is a clear and present relationship between cause and effect. Just not all the time. When you actively inform, engage and include your community you provide them with an opportunity to see the real space between cause and effect.  8) Small changes can produce big results – but the areas of highest leverage are often the least obvious. Butterfly wings and hurricanes. This is the law of leverage. Small, focused actions at the right place in the system can produce the biggest and best changes. The challenge is that the “right place” is not obvious and can seem counterintuitive. Donella Meadows describes 12 places of leverage in a system and close to the top of the list is the goals of the system. Some say that the education system isn’t working. I think it works quite well given the original goal of producing factory workers. Right now the goal is changing, in part because of changes in technology that allow education system stakeholders to collaborate and cooperate and influence the goals of the system. See #6 and 7 above.  A great example of small, counterintuitive actions are the use of insects to control insects. Wasps are introduced in many a greenhouse as a way to control other insects that feed on the greenhouse crop. Brilliant, effective and not the most intuitive solution. The key to being able to use leverage in a system is knowing the structure of the system. In education the structure is massive and very few people know it well enough to intuit where the leverage points are.  11) There is no blame. In a complex adaptive system there is no separate “other”. Everything and everyone is connected and together we co-create the whole system. Sometimes we have difficulty with this. We reflex to blame, we deflect, and deny. Its hard to take full responsibility for something that seems to be outside of our control without trying to control everything. It can feel like two competing ideas and for many that feeling is uncomfortable. Senge suggests:  The cure lies with the relationships with the very people we typically blame for the problems we are trying to solve. “Systems thinking shows us that there is no outside; that you and the cause of your problems are part of a single system. The cure lies in your relationship with your "enemy.” “The feedback perspective suggests that everyone shares responsibility for problems generated by a system. That doesn't necessarily imply that everyone involved can exert equal leverage in changing the system.”
• #20: © Professor Ockie Bosch and Dr Nam Nguyen https://www.slideshare.net/Think2Impact/
• #21: Feedback is so commonplace that MIT’s Jay Forrester, the pioneer of the System Dynamics discipline, said,
• #22: systems thinking worldview is that problems are internally generated—we often create our own “worst nightmares.”
• #26: Most intuitively obvious policy interventions in complex social systems are low-leverage (i.e., they do not produce significant long-run change and will likely also create other problems) and only a few policy interventions are high-leverage (i.e., capable of producing substantial long-run change). Previous research in system dynam- ics has shown that most intuitively obvious policy interventions in complex social systems are low-leverage (i.e., they do not pro- duce signi cant long-run change and will likely also create other problems) and only a few policy interventions are high-leverage (i.e., capable of producing substantial long-run change). System dynamics research has consistently highlighted the counterintu- itive nature of complex social systems in that the high-leverage points are not where most people expect, and if even these points are identi ed, they are prone to be altered in the wrong direction by people [11].
• #27:  In addition, the process reinforces the stock and flow way of thinking by emphasizing the difference between information and material flows, and the importance of unit consistency throughout a diagram.
• #35: Breaking a “Fixes that Fail” cycle usually requires acknowledging that the fix is merely alleviating a symptom, and making a commitment to solve the real problem now. A two-pronged attack of applying the fix and planning out the solution will help ensure that you don’t get caught in a perpetual cycle of solving yesterdays “solutions.” “People “shift the burden” of their problem to other solutions—well-intentioned, easy fixes which seem extremely efficient.” ― Peter M. Senge, The Fifth Discipline: The Art and Practice of the Learning Organization: First edition Fixes that Backfire This system archetype is associated with the concept of unintended consequences. Fixes that back re are characterized by the use of a quick x to reduce a problem symptom that works in the short run but at the cost of long-term consequences (which people often fail to see due to long system delays). This is a common pitfall in networks with some networking examples including (1) increasing queue buffers to decrease packet loss but instead causing bufferbloat [14], and (2) introducing additional links to an existing system only to see overall performance drop (Braess’ paradox) [24]. Shifting the Burden This archetype is associated with the concept of unintended dependence. This arises from dependence on a quick x, which is resorted to when the more fundamental solution is too expensive or too di cult to be implemented. This archetype di ers from fixes that back re” since the fundamental solution may not be apparent or applicable in the latter. The best example is seen in network capacity planning, where operators would rather over-dimension the network than implement more complex long-term solutions. Shifting the burden • Problem symptoms are usually easier to recognize than the other elements of the structure. • If the side-effect has become the problem, you may be dealing with an “Addiction” structure. • Whether a solution is “symptomatic” or “fundamental” often depends on one‘s perspective. Explore the problem from a dif- fering perspective in order to come to a more comprehensive understanding of what the fundamental solution may be.
• #36: Limits to Growth This archetype describes the concept of unanticipated constraints, based on the insight that no physical system can sustain growth inde nitely. Any engine of growth, however successful, will in- evitably be constrained by internal and external bottlenecks and constraints, e.g., Meadows [32] showed that we cannot sustainably support perpetual growth in a nite world. This is a long-standing worry in communications networks. For example, researchers are now exploring how a permanent energy crisis scenarios may fun- damentally limit our ability to maintain the current-day Internet architecture and what our response should be in such an eventuality [41] [39].
• #37: Effective solutions for “Tragedy of the Commons” scenario never lie at the individual level. Ask questions such as: “What are the incentives for individuals to persist in their actions?” “Can the long-term collective loss be made more real and immediate to the individual actors?” Find ways to reconcile short-term cumulative consequences. A governing body that is chartered with the sustain- ability of the resources limit can help. Of all the system structures that affect our lives and world, escalation takes the prize for its negative impact. In situations such as price wars and arms races—prime examples of the escalation dynamic—two competing entities operate, each with the goal of staying ahead of the other. As shown in “Escalation Template,” the basic causal loop structure of escalation has two balancing loops, as each party takes action to achieve its goal and move a bit ahead of its opponent. It is important to note that reinforcing loops do not have to be made up of all, or even any, “s” links. This diagram has two “o” links and six “s” links, so it is a reinforcing loop.
• #42: We can use the systems thinking concept of system-as-a-cause to explain how the perennial Internet nuisances (such as spam and lack of privacy, security and QoS) are not isolated problems but, as noted by Keshav [25] follow endogenously as the byproducts of the Internet’s design preferences. This work points out that para- doxically the Internet’s architectural elements most responsible for its success are also responsible for its most vexing problems. It is clear that if we want to x these ancillary problems, this cannot be achieved super cially without changing the systemic causes.
• #43: Thus, there is “not a single happy family of people” on the Inter- net with aligned goals [7]. Apart from tussles and conflicts, Internet protocols and applications also often face dilemmas in which the goals of the subsystem and the overall system conflict. One of the major insights of systems thinking is that the best way to optimize a system is not to independently optimize each subsystem but to optimize the relationships among the parts (which often is the bottleneck).  “Addiction is finding a quick and dirty solution to the symptom of the problem, which prevents or distracts one from the harder and longer-term task of solving the real problem.” ― Donella H. Meadows, Thinking in Systems: A Primer
• #44: Ben Shneiderman (University of Maryland) argues against autonomous systems. His point is that it is essential to keep a human in the loop. If not you run the risk of abdicating ethical responsibility for system design.
• #45: https://arxiv.org/pdf/1611.04135.pdf