Exchange Cybernetics: towards a Science of Agility & Adaptation

Predicting Successful Decision-Making

I believe that agility can become part of a scientific theory of adaptation, and I have been building simulations to explore some of these ideas. However, in order for agility to become a science, it must be shown to complement and add to the existing science of economic decision-making. In order to add something new to existing microeconomics and decision-making science, a science of agility would offer a different way of looking at familiar problems. It could be the case that the science of successful decision-making is not only predicted by assuming people are rational and having aligned incentives such as bonuses with the desired outcome as classical economics dictates. The more recent idea of behavioural economists that "nudging" people with inherent cognitive biases to make better decisions may also not be the full story. I believe that indeed there is a scientific "gap" in the market for predicting successful adaptation by individuals and organisations. So, it is my contention that successful decision-making is predicted by having the capacity for adaptation and using it wisely. I believe that this capacity for adaptation is nothing more than the ability to move resources around in order to take opportunities as they emerge. To be able to adapt in this way well is to behave with "agility," as the name implies. There is, however, some conceptual development needed to make good on the promise of this idea. So, let’s consider the ingredients of an agile theory of adaptation.

The Ingredients of Agile Adaptation

The first ingredient for a better model of adaptation is the concept of control. This already exists as a science called "cybernetics". Using these ideas, I argue that successful action (rather than simply decision-making) is largely predicted when the components for control over a situation are provided to a competent individual. The three main components of control are simple:

1. The resources, both information and physical to act, which comprise the "capacity to act"
2. The "opportunity to act" which comprises the often fleeting opportunity, that given sufficient capacity to act, it will reliably lead to success
3. A competent individual able to use these control resources to take the opportunity

These components, when all supplied, can be called the "locus of control" of an individual. When an opportunity falls within my locus of control, my contention is that as a competent individual, I will reliably take the opportunity. Nowhere does this type of analysis seem more common-place and clear than in team sports.

Football

Consider an example from my favourite sport, football. If, as a football team you collectively give a good footballing opponent too much time and space, they will take advantage of that and choose one of the better actions to exploit any weakness in your defence. Conversely, if you manage to deny, even very good, football players the resources in terms of time and space to act, they will reliably fail to win the game, as they are simply being denied the capacity to act. The game of adaptation is this: To supply agents with the locus of control to take their opportunities. In team ball sports the capacity to act is simply and obviously connected to passing the ball to where it needs to go. The idea of successful adaptation can then be said to be the facility to move the locus of control to where it is needed. As a more general way of analysing and predicting success in micro-economic decisions, this idea of supplying the ingredients of control is actually quite novel (so I would argue) in economics.
 
If this dimension of predicting successful decision-making and action is as important as I think, then it follows that we should aim to predict where the locus of control is and where it needs to be for a successful outcome because this best predicts success. I call the analysis of the adaptations required to move the locus of control to where it should be "Exchange Cybernetics". Agents must exchange control resources to move them where they need to go to control situations and convert opportunities. If they do this successfully, they will reliably succeed in their goals, just like the football team moving the ball to where it needs to go to eventually score a goal. This should also start to sound familiar to anyone who works in software development teams or manages teams, generally, because I think that this problem is simply a scientific description of the process or property of having "agility". By defining what this locus of control for each activity looks like, and measuring it, we should be able to reliably predict success or failure. If such measurement and prediction sounds impossible, consider football analytics.

Sports Analytics and Simulation

In the new science of sports analytics this type of analysis is already being done in football, a sport that until a few years ago was considered "unmeasurable". The results are tactical insights that don't just explain past events, but allow the devising of strategies to affect future choices, as you can see from this article Decomposing the Immeasurable Sport by Fernandez, Cervone and Bornn. Football is a game of extreme fluidity of attack and defence. In addition, the criticality of decisions to outcomes is high, as a single goal often decides a game. Nevertheless, data analysts can now use various methods to predict the optimal pass at any one moment, and also strategise about how to improve outcomes for a team by identifying when players could have made better choices. In my opinion, in sports analytics we are seeing a contribution to a new field of adaptation science where agility is actually near the heart. This is despite the data scientists in this football analytics domain perhaps being unaware that they may also be significant contributors to an "agile science". I am convinced that "exchange cybernetics" is the best way to do this sort of analysis in a general context, such as in business or indeed in other scientific domains. So, let’s consider an outline of a theory for how this might be done to inform tactics for a software development team:

The Concept of Redundant Control

The logic of a footballer’s decision to pass the ball is that the control of the ball she currently has can’t be made much use of, while the person she can pass the ball to looks as if they could do more damage to the opposition. Hence, one can encode in our model this general idea of "spare control" in one place "A" versus a potential locus of control in place "B" that is incomplete save for the resource that we don’t need so much of in place "A". We can think of this adaptation as an example of a change that almost regardless of what the new possessor of the control resource does, is far more likely to add value rather than subtract from it, because the new possessor of the control resource is using control that was redundant to the previous owner. It follows that to identify how to move control around, we should first identify where "spare" control is becoming available and then look at the areas where that now redundant control could be used to greatest effect.

The key to this type of measurement analysis can be seen by analysing PERT charts and adapting them to look at the concept of "Exchange Cybernetics".

Figures 1 & 2

The PERT chart in Figure 1 is usually used to do Critical Path Analysis but can instead be utilised to explain “exchange cybernetics”. A PERT shows us where the critical path of a project is. Other paths which are "sub-critical" have "float" of a certain amount, meaning that they can be delayed by the float amount without delaying the whole project. Of course, if there are risks that things take longer than expected, then this also has no negative effect on the project if it only uses the float. In terms of exchange cybernetics, one can also think of the activities with float, more positively, not just as having contingency but also as having "adaptive float".  This means that adaptive float is also just "spare control" as just mentioned. The extra functionality one can build using only float should have no negative effects which would threaten to outweigh the positives, and so is far more likely to be beneficial overall; see Figure 2.

The next step needed however, is to model how adaptive float emerges in a situation to be taken advantage of. Such adaptive float will not necessarily show up in the PERT at the start of the project, but can appear according to circumstances, creating the potential for agile decision-making.

A Tactical Example

In an exchange cybernetic model of a software development project we can consider the tactical options for the team when they need to refocus the locus of control to successfully complete a task by taking advantage of emerging adaptive float. One important dimension of control is timing; you can try to achieve control early on in a process, or, if things change, later on. The best tactic depends on the answer to a couple of key questions about what might be happening in the process as it progresses. To see why, let’s consider this example in more detail:

We have to decide if some of the control we should have early on in a process could actually become redundant as the tasks progress. Recall that the idea of the "locus of control" consists of physical and information resources that give the capacity to act, as well as opportunity to take action. Some control early on in a process might become redundant if some of the remaining necessary components of the locus of control continue to be missing from that phase. For example, early on into a process we might discover that reliable information about what might have business value is not forthcoming, even if we go to some considerable trouble to try to obtain it. If so, it now makes more sense to try to refocus control from some of the now redundant control resources in the early phase to the later stages when reliable information can be obtained. This can happen in the later phase via more reliable feedback on what is missing from what has already been built, for example. Given enough other resources to act, we should then be able to dramatically increase the odds of a successful outcome, as the locus of control will be complete in that later phase and a competent person will be able to take that opportunity (see top of Figure 3).

Figure 3

Tactical Planning

In practical terms, such a tactical adaptation could also be planned for as a potential scenario envisaged in the beginning, so that should this scenario develop, one can take a pre-arranged decision to cut short the earlier phase when it is realised that the necessary reliable information is not forthcoming, despite reasonable efforts. What I mean by this, is that tactical adaptation can be planned for, just as in a sporting context where tactical moves are planned and then triggered by a context that emerges which dictates when they become the better option. In this case, the trigger for reliable information not forthcoming might be, for example, that we find that after a defined amount of time, the people we need to confirm business value are still unavailable, or if available, they are unable to confirm the business value without seeing a prototype first, creating the adaptive float. This criterion can be pre-judged as sufficient for making the tactical switch to the process. So the tactical decision would be taken in reaction to an emerging situation, just as a footballer reacts to the unfolding situation to change their plan about who to pass to at the last second, yet as a rehearsed tactic. This implies that as agile practitioners we should also plan tactical moves and mentally rehearse. By clearly defining the conditions for such a tactic to be successful and tracking whether they have been met we can then be ready to execute the tactic flawlessly when the situation arises. We can also be entirely accountable due to our planning for this scenario, and also transparent to the business about our rationale.

But there’s another side to this particular tactical equation: a later phase might come after a "cliff-edge" decision-point where any re-work becomes very, very expensive. For example, if a release of the software delivers critical components which contain bugs, this could severely affect key customers and the reputation of the business. In such a scenario (bottom of Figure 3), it might be judged that the locus of control cannot be focused after that decision-point because the capacity to act is missing after the cliff-edge. In such a case, this is an argument that adaptive float is emerging in the later phase, and so we should, in fact, push forward the locus of control to an earlier phase, before the cliff-edge decision-point (see bottom of Figure 3). Such a decision would again need to be taken after careful planning and scenario analysis.

This type of analysis, whichever way the argument about viable tactics goes, is the foundation of a general form of exchange cybernetics. While real life is always more complex, it holds out the promise of enabling managers and teams to make better informed and pre-planned tactical decisions, and to be agile in a way under-pinned by a sound theory. This also means to do so more transparently and with more accountability to stakeholders and management. But in order to achieve that higher accountability and transparency, we also need to define a strategy for tactical planning.

Building the Capacity for Tactical Planning

The key to building capacity for the execution of tactical switches to a process is to plan by envisaging the scenarios that might occur and preparing potential tactical switches before the process begins. This would involve four parts:

1. Reviewing history of the team and using their experience to identify commonly occurring scenarios that have previously occurred, in the current organisation, or elsewhere.
2. Discussing the tactics that could be executed to deal with that emerging scenario using the exchange cybernetic logic of utilising emerging adaptive float.
3. Where possible, defining in advance the checkpoints and conditions for triggering a tactical switch, ideally using measurable departures from planned progress expectations tracked during the process phases. The scenario planning context would also inform on what expectations and potential departures to track. This would give further accountability for tactical switches.
4. Drawing up a list or analysis of substitute activities. These are potential substitute forms of the same activities that you initially perform, such that switching to the substitute form of the activity, task or product component allows you to recover time and resources created by the emerging adaptive float and exchange them with some other activity as part of the tactic in any given scenario. An analysis of substitute activities would be drawn up to complement scenarios and tactics envisaged. They would ensure that careful thought goes into how to execute a tactic without causing undue extra work downstream. For example, what requirements might become redundant, and what requirements or delivery methods could be altered in a given scenario to safely release emerging adaptive float?

The hallmark of a more scientific approach to adaptation would be this tactical planning. The strategy is applicable to any kind of team process or project planning or plan for the delivery of engineered or built functionality. It rests on the recognition that executing tactical switches in teams themselves also needs planning to increase the chances of success, because without planning the time for decision-making is often very compressed and mistakes are likely, e.g. adopting hasty changes to complex activities can result in substitute forms of activities that are not actually good substitutes. Yet, there are also further implications for the strategy in this approach.

Stress-Testing Strategy

Firstly, from a strategic point of view we can also ask ourselves whether we are likely to have at our disposal the required substitute activities to safely release the adaptive float in order to make tactical switches. By considering lots of possible scenarios and the different tactical switches we would need to execute, we are performing a type of "stress-testing" of the process for a team, before we start a project or continue in an agile process. In this way we show how we intend to deal with uncertainty and manage risk by showing how we would control it by showing the capacity to adapt, accordingly. Part of managing risk in this way is to understand how uncertainty creates new scenarios and demand for adaptive float and viable substitute activities to release it. My previous article at InfoQ about my simulations focused exclusively on this kind of risk, and described "Lazy Stopping", which is where team members progress activities without realising that some work is actually still undone, creating the need for adaptive capacity to react and recover. This risk is higher when executing work of higher novelty to the team.

Adaptive Opportunity Cost

A second strategic aspect is the potential to theorise and then understand what can be termed “adaptive opportunity cost”. This is the cost in terms of opportunities that were missed. Due to the focus on adapting to taking a certain opportunity that was presented, we can miss out on a better opportunity. Again, in team sports this idea is commonplace and obvious: if I make the wrong pass or choose to pass when I had an easy shot, then even if that pass was OK, the opportunity cost of that decision is high because a better choice was available; it is part of what makes it a bad decision. In business terms, we can also analyse this opportunity cost, and adapted PERT charts can also be used to illustrate this analysis. I am actively working on further simulations and analysis to show how we can evaluate agile strategies for a business by doing such opportunity cost analysis.

Conclusion

Some basic ideas in exchange cybernetics have been explained to give the reader a flavour of how this idea works to model opportunity created by identifying emerging "adaptive float" and moving it to where it can better be utilised as a model of agility. Concrete examples were given for how to do tactical planning in order to execute agility. There are many more aspects to this theory, but this is the bedrock of the idea. Everything else that can be done with it is an extension of this basic concept. Yet surprisingly sophisticated ideas about tactical planning and the estimation of adaptive opportunity cost follow from it. The plan is to apply these ideas where I work and continue working on the theory behind it. I also continue to blog about these ideas.

About the Author

Adam Timlett has a research masters in philosophy from the University of East Anglia, and has published on complexity and commonsense. He currently works as the analytics manager at the company PPL, and lives in London, United Kingdom. He has a keen interest in science, philosophy, technology, and innovation. He blogs on the subject of “complexity” and carries out research in biology and economics.