7 Implications of Complexity for Safety

One of my favourite articles is The Complexity of Failure written by Sidney Dekker, Paul Cilliers, and Jan-Hendrik Hofmeyr.  In this posting I’d like to shed more light on the contributions of Paul Cillliers.

Professor Cilliers was a pioneering thinker on complexity working across both the humanities and the sciences. In 1998 he published Complexity and Postmodernism: Understanding Complex Systems which offered implications of complexity theory for our understanding of biological and social systems. Sadly he suddenly passed away in 2011 at the much too early age of 55 due to a massive brain hemorrhage.

My spark for writing comes from a blog recently penned by a complexity colleague Sonja Bilgnaut.  I am following her spade work by exploring  the implications of complexity for safety. Cilliers’ original text is in italics.

1. Since the nature of a complex organization is determined by the interaction between its members, relationships are fundamental. This does not mean that everybody must be nice to each other; on the contrary. For example, for self-organization to take place, some form of competition is a requirement (Cilliers, 1998: 94-5). The point is merely that things happen during interaction, not in isolation.

• Because humans are natural storytellers, stories are a widely used  interaction between fellow workers, supervisors, management, and executives. We need to pay attention to  stories told about daily experiences since they provide a strong signal of the present safety culture.
• We should devote less time trying to change people and their behaviour and more time building relationships.  Despite what psychometric profiling offers, humans are too emotional and unpredictable to accurately figure out. In my case, I am not a trained psychologist so my dabbling trying to change how people tick might be dangerous, on the edge of practising pseudoscience.  I prefer to stay with the natural sciences (viz., physics, biology), the understanding of phenomena in Nature which have evolved over thousands of years.
• If two workers are in conflict, don’t demand that they both smarten up. Instead, change the nature of relationship so that their interactions are different or even extinguished. Simple examples are changing the task or moving one to another crew.
• Interactions go beyond people. Non-human agents include machines, ideas (rules, policies, regs) and events (meeting, incident). A worker following a safety rule can create a condition to enable safety to emerge. Too many safety rules can overwhelm and frustrate a worker enabling danger to emerge.

2. Complex organizations are open systems. This means that a great deal of energy and information flows through them, and that a stable state is not desirable.

• A company’s safety management system (SMS) is a closed system.  In the idealistic SMS world,  stability, certainty, and predictability are the norms. If a deviation occurs, it needs to be controlled and managed. Within the fixed boundaries, we apply reductionistic thinking and place information into a number of safety categories, typically ranging from 4 to 10. An organizational metaphor is sorting solid LEGO bricks under different labels.
  In an open system, it’s different. Think of boundary-less fog and irreducible mayonnaise. If you outsource to a contractor or partner with an external supplier, how open is your SMS? Will you insist on their compliance or draw borders between firms? Do their SMS safety categories blend with yours?
• All organisations are complex adaptive systems. Adaptation means not lagging behind and plunging into chaotic fire-fighting. It means looking ahead and not only trying to avoid things going wrong, but also trying to ensure that they go right. In the field, workers when confronted by unexpected varying conditions will adjust/adapt their performance to enable success (and safety) to emerge.
• When field adjustments occasionally fail, it results in a new learning to be shared as a story. This is also why a stable state is not desirable. In a stable state, very little learning is necessary. You just repeat doing what you know.

3. Being open more importantly also means that the boundaries of the organization are not clearly defined. Statements of “mission” and “vision” are often attempts to define the borders, and may work to the detriment of the organization if taken too literally. A vital organization interacts with the environment and other organizations. This may (or may not) lead to big changes in the way the organization understands itself. In short, no organization can be understood independently of its context.

• Mission and Vision statements are helpful in setting direction. A vector, North Arrow, if you like. They become detrimental if communicated as some idealistic future end state the organization must achieve.
• Being open is different than “thinking out of the box” because there really is no box to start with. It’s a contextual connection of relationships with other organizations. It’s also a foggy because some organizations are hidden. You can impact organizations that you don’t even know about and conversely, their unbeknownst actions can constrain you.
  The smart play is to be mindful by staying focused on the Present and monitor desirable and undesirable outcomes as they emerge.

4. Along with the context, the history of an organization co-determines its nature. Two similar-looking organizations with different histories are not the same. Such histories do not consist of the recounting of a number of specific, significant events. The history of an organization is contained in all the individual little interactions that take place all the time, distributed throughout the system.

• Don’t think about creating a new safety mission or vision by starting with a blank page, a clean sheet, a greenfield.  The organization has history that cannot be erased. The Past should be honoured, not forgotten.
• Conduct an ongoing challenge of best practices and Life-saving rules. Remember the historical reasons why these were first installed. Then question if these reasons remain valid.
• Be aware of the part History plays when rolling out a safety initiative across an organization.
  • If it’s something that everyone genuinely agrees to and wants, then just clone & replicate. Aggregation is the corollary of reductionism and it is the common approach to both scaling and integration. Liken it to putting things into org boxes and then fitting them together like a jigsaw. The whole is equal to the sum of its parts.
  • But what if the initiative is controversial? Concerns are voiced, pushback is felt, resistance is real. Then we’re facing complexity where the properties of the safety system as a whole is not the sum of the parts but are unique to the system as a whole.
    If we want to scale capabilities we can’t just add them together. We need to pay attention to history and understand reactions like “It won’t work here”, “We tried that before”, “Oh no! Not again!”
    The change method is not to clone & replicate.  Start by honouring local context. Then decompose into stories to make sense of the culture. Discover what attracts people to do what they do. Recombine to create a mutually coherent solution.

5. Unpredictable and novel characteristics may emerge from an organization. These may or may not be desirable, but they are not by definition an indication of malfunctioning. For example, a totally unexpected loss of interest in a well-established product may emerge. Management may not understand what caused it, but it should not be surprising that such things are possible. Novel features can, on the other hand, be extremely beneficial. They should not be suppressed because they were not anticipated.

• In the world of safety, failures are unpredictable and undesirable. They come as surprises because they unexpectedly emerge when a hidden tipping point is reached. Many organizations have in place Emergency Preparedness plans which designate Recovery crews with well-defined roles. Their job is to fix the system as quickly as possible and safely restore it to its previous stable state.
• The other type of surprise is Serendipity, an unintended but highly desirable consequence. This implies an organization should have an Opportunity crew ready to activate. Their job is to explore the safety opportunity, discover new patterns which may lead to a new solution, and exploit their benefits.
  At a tactical level, the new solution may be a better way of achieving the Mission and Vision. In the same direction but a different path or route.
  At a strategic level, the huge implication is that new opportunity may lead to a better future state than the existing carefully crafted, well-intentioned one. Decision-makers are faced with a dilemma: do we stay the course or will we adapt and change our vector?
• Avoid introducing novel safety initiatives as big events kicked off with a major announcement. These tend to breed cynicism especially if the company history includes past blemished efforts. Novelty means you honestly don’t know what the outcomes will be since it will be a new experience to those you know (identified stakeholders) and those you don’t know in the foggy network.
  Launch as a small experiment.
  If desirable consequences are observed, accelerate the impact by widening the scope.
  If unintended negative consequences emerge, quickly dampen the impact or even shut it down.
  As noted in (2), constructively de-stabilize the system in order to learn.

6. Because of the nonlinearity of the interactions, small causes can have large effects. The reverse is, of course, also true. The point is that the magnitude of the outcome is not only determined by the size of the cause, but also by the context and by the history of the system. This is another way of saying that we should be prepared for the unexpected. It also implies that we have to be very careful. Something we may think to be insignificant (a casual remark, a joke, a tone of voice) may change everything. Conversely, the grand five-year plan, the result of huge effort, may retrospectively turn out to be meaningless. This is not an argument against proper planning; we have to plan. The point is just that we cannot predict the outcome of a certain cause with absolute clarity.

• The Butterfly effect is a phenomenon of a complex adaptive system. I’m sure many blog writers like myself are hoping that our safetydifferently cause will go viral, “cross the chasm”, and be adopted by the majority. Sonja in her blog refers to a small rudder that determines the direction of even the largest ship. Perhaps that’s what we are: trimtabs!
• On the negative side, think of a time when an elected official or CEO made a casual remark about a safety disaster only to have it go viral and backfire. In 2010 Deep Horizon disaster then CEO Tony Hayward called the amount of oil and dispersant “relatively tiny” in comparison with the “very big ocean”.  Hayward’s involvement has left him a highly controversial public figure.
• Question: Could a long-term safety plan to progress through the linear stages of a Safety Culture Maturity model be a candidate as a meaningless five-year plan?
  If a company conducts an employee early retirement or buy-out program, does it regress and fall down a stage or two?
  If a company deploys external contractors with high turnover, does it ever get off the bottom rung?
  Instead of a linear progression model, stay in the Present and listen to the stories internal and external workers are telling. With the safety Vision in mind, ask what can we do to hear more stories like these, fewer stories like those.
  As the stories change, so will the safety culture.  Proper planning is launching small experiments to shape the culture.

7. Complex organizations cannot thrive when there is too much central control. This certainly does not imply that there should be no control, but rather that control should be distributed throughout the system. One should not go overboard with the notions of self-organization and distributed control. This can be an excuse not to accept the responsibility for decisions when firm decisions are demanded by the context. A good example here is the fact that managers are often keen to “distribute” the responsibility when there are unpopular decisions to be made—like retrenchments—but keen to centralize decisions when they are popular.

• I’ve noticed safety professionals are frequent candidates for organization pendulum swings. One day you’re in Corporate Safety. Then an accident occurs and in the ensuing investigation a recommendation is made to move you into the field to be closer to the action. Later a new Director of Safety is appointed and she chooses to centralize Safety.
  Pendulum swings are what Robert Fritz calls Corporate Tides, the natural ebb and flow of org structure evolution.
• Central v distributed control changes are more about governance/audit rather than workflow purposes. No matter what control mechanism is in vogue, it should enable stigmergic behaviour, the natural forming of network clusters to share knowledge, processes, and practices.
• In a complex adaptive system, each worker is an autonomous decision-maker, a solution not a problem. Decisions made are based on information at hand (aka tacit knowledge) and if not available, knowing who, where, how to access it. Every worker has a knowledge cluster in the network. A safety professional positioned in the field can mean quicker access but more importantly, stronger in-person interactions. This doesn’t discount a person in Head Office who has a trusting relationship from being a “go to” guy. Today’s video conferencing tools can place the Corp Safety person virtually on site in a matter of minutes.