Can simulation for decision support predict the future?
“The challenge in creativity is not to find new ideas but to abandon the old ones”.
Luc de Brabandere
Recently, scientists took some of the fundamental laws of physics, data on the formation of galaxies and images from the Hubble telescope, and created a simulation of our universe. The simulation has allowed them to study the makeup of dark matter and understand our universe in a much deeper way. One of the key results of the simulation has been the ability to make predictions based on observations.Our simulation of the universe may be fairly new, but human beings have always used model thinking. By creating a model or simulation of a system, we can see that the whole is greater than the sum of its parts; models let us combine the power of both reductionism and holism, allowing the individual components of a system or problem to predict the outcome.
The massive strides made in technology in recent years has fed into the production of high performing and affordable simulation technologies, which open a new world of decision support and prediction of outcomes, within both a civilian and military context.
Contributions simulation makes to decision support
Decisions can often be complex, with a variety of options. Simulation has a practical application to the process of decisionmaking. Simulation tools support us in the decisionmaking process. It gives us foreknowledge and with knowledge, comes control. Knowledge borne from simulation supports our decisions and makes them more accurate. Simulation tools allow us to take those options and see a big picture, without being blinded by the details.
Conditions of use for simulation in decision support
Because simulation tools can help you to predict future outcomes, they can also allow you to see alternative outcomes. By changing the variables of the model, you can see the impact each has on the whole. Simulation is useful across a wide spectrum of conditions. The validity of the information source you input will determine the accuracy of the outcomes and decisions made from the resultant simulation.
Prospects: what if we could predict the future?
Humankind has always toyed with futures. We have built it into our folklore; we use it daily in share trading, and design our technology to predict usage patterns for better user experiences. Model thinking and simulation allow us to predict the future for better decisionmaking, which gives us even more control over our environment. Being able to predict the future is the ultimate goal. We may not have an actual time machine… yet… but using simulation tools gives us the next best thing.
Why use simulation and not something else?
There is a saying, “there is more than one way to skin a cat” this
holds true for many areas of business decisionmaking. So why and when would you choose simulation tools?
If we look at three of the main technologies used to make decisions, we can see how simulation tools fit into the toolkit of choice:
∙ Business intelligence: this answers the questions ‘When?’ and ‘Where?’ if you have enormous amounts of data, using BI tools allows you to drill down into that data and to analyse it. It doesn’t predict, it analyses.
∙ Machine learning: this answers the question ‘Who?’. It uses data and information to find patterns about users. It doesn’t predict, it looks for pattern, and then uses them to adjust programs to obtain more accurate results about the knowledge of the users.
∙ Simulation: this answers the question ‘What if?’. Simulation uses data to go beyond BI and machine learning into the realms of prediction. It also answers the question ‘Why?’ as simulation helps us to understand a past event to build up knowledge from experience.
What is simulation made up from?
Simulation offers a mechanism for combining a number of different attributes not found in any of the technologies mentioned above. To generate a working model from simulation technologies, you require the following variables:
∙ something representative of the system for example, a tank could be represented by a lot of details such as a texture, or more roughly by a symbol.
∙ capabilities of the components of the model for example, a tank of this model has this type of ammunition and travels at speed x/y.
∙ behaviour of a component within the system. This is a characteristic unique to simulation. An example of a behaviour would be when a tank travels along the road at a certain speed and then reaches an uneven terrain, it will need to change its speed and movement.
All the above attributes are variable and can be manipulated within the model. It is this key feature of simulation that provides the insights into futures.
Simulation tools allow us to play out scenarios to see different outcomes. We can change the terrain and see how a variation in behaviour would impact the capability. Managing the variables will show changes across the system that allow us to understand how each impacts the other we can predict the future of changes in these attributes.
Simulation gives us an elegant way of testing out hypotheses simply by changing the input data/criteria. Implementing models, and observing the ways in which they interact, provides the information and arguments needed to evidence operator decisions and make those decisions more precise.
When to use simulation?
Simulation works best when focused on specific tasks and requirements. Keeping it simple will generate more accurate results. Simulation is there to support your decisionmaking processes.
If you are unsure of how a decision will impact on a system or design, then simulation allows you to observe the consequences of an action if I change X, Y will happen. However, simulation does not tell the operator what to do. Nor does it tell them how they should act, or how they should predict the future.
Simulation is also a powerful tool in complex systems, where it may takes excessive time to see a change occur, and where it would be too costly to see how making a change would impact the outcome.
An example of this in action is in the use of simulation within a military context:
Situation: a tank company is moving from one Coordination Line to another.
Variables: the movement of the tank is not as straightforward as it may seem; there are interactions between the tanks and the terrain, and tactical behaviours need to be taken into account. This in turn leads to more specific timelines.
Simulation model power lies in the ability to bring multiple variables together to understand their interactions and how they impact on each other to create future events.
Simulation tools act as indicators for decisionmaking.
They are decision enablement frameworks empowered through information, and can rationalize and confirm a decision within a given context. You can liken them to a ‘super calculator’, which can take in the data entered and come up with a result.
How can I trust the result given by simulation?
As mentioned, simulation has been used by human beings for a very long time and allows us, amongst other things, to plan a mission and visualise the coordination points. Simulation models can be validated using a sandbox; you can populate the sandbox, maneuver it, see it from all angles and make changes. But a sandbox is not reality. Because of advances in computer power and performance, digital simulation has moved the goalposts of reality now it is much closer to the real thing, making simulation outcomes much more realistic.
A real sandbox is far from reality. Models are rough. Items are represented by natural components. A leaf becomes a tree, a heap of earth becomes a hill, a piece of wood becomes a building, and pebbles are friends and foes; an operator implicitly understands this simulation is
not reality. Because of this, the operator can integrate differences between the simulation and the real world in the analysis.
Conversely, digital simulation can be close to reality; it is like reality but not quite. This kind of simulation lets an operator assume reality and handle it as such. But all the same, it is not real.
The operator knows it is wrong but doesn’t know how to explain it. An expert is required to discuss the validity of models and data.
Simulation results and the person who will use them (mostly operators) remain fundamental.
Technicians would not necessarily be the ones who use simulation, the most important aspect to consider is to know how best to use and interpret the results.
Creating an indicator of trust: degree/level of trust
Applying a trust indicator helps to interpret simulation results. For example, in the weather forecasting field, indicators are applied to every forecast, such that:
a. if the indicator is low and the forecast is incorrect, there’s nothing unusual. The trust indicator applied to that simulation was poorly applied.
b. if the indicator is low and the forecast is right, the forecast is likely down to luck so again the trust indicator was poorly applied.
c. if the indicator is at its maximum and the forecast is right, it is a compliant trust indicator.
d. however, if the indicator is at its maximum and the forecast is wrong, there’s a contradiction. The indicator of trust should therefore be questioned rather than the simulation.
This indicator is built in accordance with operational criteria, and the knowledge of simulation methods. The operator would help build this indicator of trust.
What would the consequences be?
If the interpretation of a simulation is based on its level of trust alone, then it is not enough. It must be supplemented by assessing the impact of the simulation:
1. Accuracy of important simulations will require additional studies to confirm or overturn the results of the simulation (for decision support, for instance) regardless of the level of trust of the simulation.
2. Even seemingly small outcomes from a simulation can also provide information that can direct or close assumptions, and therefore limit costs.
Integrating simulation tools into the operator’s work environment
If the operator is the key stakeholder in the use of simulation tools, then the prime concern is usability. In other words, the operator should be able to make use of decision support without a complex simulation interface. This tool should be integrated into the operator’s work environment and should be easily accessible.
To create a usable and effective simulation you must determine the best possible integration scenario. This means that the operator environment must be properly analysed, with use cases and scenarios mapped out to optimize experience with the tool and ensure that the operator can work effectively. Mapped to this must be interoperability with the operator’s current tools. Using simulation is about decision support this must be done in a seamless manner with little disruption to the operator’s usual working environment.
Switching between simulation tools and basic interfaces has to be seamless. Complete integration is one of the main features of simulation. The other one is the automated transmission of data from one system to another, and vice versa. An example of this in action is the use in smartphones where simulation apps like Waze, a realtime community drive navigation app, are completely integrated into the smartphone; Waze reuses data derived from the smartphone or other systems.
Predictive analysis using simulation is possible. We already use it in fact in our weather predictions. The weather forecasts use a 5 level index to show the reliability of their weather simulation. However, further work needs to be done to create truly powerful predictive analytics, which have to be fully integrated and seamless. Creating an intrinsic system makes the use of simulation invisible to the operator, improving decisions and giving us better outcomes to problem solving.