The Effectiveness and Strengths of Simulations with Intentional Uncertainties

Most things in this world are represented as a nonlinear, dynamic, and complex system. A few models that Guastello (2013) mentioned, including the complex adaptive system of the human cognition and decision process, the dynamics of the human learning process, the circular causation network of risk assessments, and complex system on an organizational level, are just a few examples representing the nonlinearity of most systems in the real world. When training in a healthcare setting, the goal of simulations is to prepare physicians for expected and unexpected events that could occur during surgery, allowing them to retrieve any knowledge under any circumstances. This is tightly related to how the human brain learns. Although both types of learning, intentional uncertainty, and repeated simulation, are consequential, it is best for training approaches to introduce intentional uncertainty as part of the simulation design to simulate the real world better and make the preparation and practice process more effective. 

Human neural network

It is essential first to identify and understand how the human learning and decision model operates to design an enhanced training simulation. Currently, many different models attempt to model human cognition in the field of semantic cognition (McClelland & Rogers, 2003). Although no single model can fully address the human mind, according to McClelland and Rogers (2003), the use of parallel distributed processing (PDP) models can address several aspects of human cognition that other models could not encompass.

When humans encounter a new piece of information during training and learning, a new node is added to the neural network. It connects with the existing networks as McClelland and Rogers (2003) analogize, just like how children first acquire general concepts and progress to finer topics and details, connecting with the existing knowledge that follows the shortest paths, thus, expanding and building a more intricate neural network. Finally, Guastello (2013) finds that noise in a stimulus allows the neural network to make adaptive systems more readily than noise-free training. This comment is compatible with the idea that an expanded web with nodes provides faster access to knowledge and enhanced training results. In the context of healthcare, when physicians gather a new piece of knowledge, they will not remember that piece of knowledge as is. Instead, they will remember the piece of knowledge with the context of the patient’s condition or demographics, integrating the knowledge to what he already knows.

Complex systems in real-life

From experience, most works in real-life are not a one-person job. There are often multiple subunits, systems that are intertwined in a decision-making process. For example, Pype et al. (2018) demonstrate the complexity of healthcare teams and how the complex adaptive system model can best describe and explain different team and individual behaviors. Their findings suggest that most systems in the real-world are nonlinear and often dynamic. For example, a nurse may just have a bad day and be unable to provide the physicians with the level of help they expect. Therefore, training should not only focus on a personal level, such as hard knowledge, but also interpersonal level that can account for behaviors of various individuals or groups. One way of doing such is to introduce uncertainty into the training to better prepare individuals and teams for unforeseen situations. As Guastello (2013) states, if training were going to simulate a complex adaptive system, it would also need the adaptive capability. 

Psychological aspect of intentionally uncertain simulations

Opponents may argue that human neural networks need repeated simulations to strengthen and reinforce the knowledge. However, there are upsides to intentionally uncertain simulation even if it cannot reinforce the knowledge as in a repeated simulation training. Suppose a group of physicians is training with repeated simulated training. They will know a set of skills well. However, simulation designers cannot possibly program all possible scenarios into the training. As mentioned, all systems are complex, dynamic, and nonlinear in real life. Each patient will have a different condition and health history; the on-call specialist will also differ each day. In other words, physicians will still encounter unknown situations, as in an intentionally uncertain simulation training. When physicians encounter an unexpected problem, for example, during an operation, and are not used to the stress under time pressure, their performance could be impaired by the pressure. 

On the other hand, physicians with intentionally uncertain simulations are used to solving unexpected issues on the spot because their training prepares them for unforeseen events. Compared to physicians who train with repeated simulations, these physicians will calmly address the situation, as they were trained, and maintain performance. Ceschi et al. (2017) note that decision-making ability can act as a predictor of work performance. When decision-making is impaired by stress induced by a lack of knowledge, work performance also decreases.

Conclusion

Although both types of training, whether intentionally uncertain or repeated, are important in preparing operators for real life operations, due to the nature of how the world operates, which is always dynamic and nonlinear, and how the human cognition incorporates new knowledge, a training is more effective if  intentional uncertainties are introduced. Similar ideas can be seen in machine learning, where the artificial neural network is derived from the human neural network. Machines do not learn with the same set of data on repeat, but numerous sets of large data, in an attempt to understand the world. This is the only way to prepare for any operation in the real world, where each decision is determined by infinite paths.

References

Ceschi, A., Demerouti, E., Sartori, R., & Weller, J. (2017). Decision-Making Processes in the Workplace: How Exhaustion, Lack of Resources and Job Demands Impair Them and Affect Performance. Frontiers in Psychology, 8, 1–14. https://doi.org/10.3389/fpsyg.2017.00313

Guastello, S. J. (2013). Human Factors Engineering and Ergonomics: A Systems Approach, Second Edition (2nd ed.). CRC Press.

McClelland, J. L., & Rogers, T. T. (2003). The parallel distributed processing approach to semantic cognition. Nature Reviews Neuroscience, 4(4), 310–322. https://doi.org/10.1038/nrn1076

Pype, P., Mertens, F., Helewaut, F., & Krystallidou, D. (2018). Healthcare teams as complex adaptive systems: understanding team behaviour through team members’ perception of interpersonal interaction. BMC Health Services Research, 18(1), 1–13. https://doi.org/10.1186/s12913-018-3392-3