Using FRAM to Analyze Systems
If you’re considering using the FRAM (Functional Resonance Analysis Method) to model systems, it’s important to note that the term “FRAM model” is technically incorrect. Instead, the appropriate term is “a model built by the FRAM” or “a FRAM-built model.” This distinction is crucial because the “M” in FRAM stands for “Method,” not “Model.” The FRAM is a methodological tool without a predetermined visual representation, unlike models such as Bow Ties or Swiss Cheese diagrams. This flexibility makes the FRAM uniquely suitable for analyzing complex sociotechnical systems, where predefined structures can prejudice predictions of behaviour.
Systems Thinking: Complicated vs. Complex
Understanding real-life activities is a natural human curiosity, historically approached through rational analysis since the Renaissance. This method often involves:
- Reductionism: Breaking down a system into its component parts.
- Determinism: Understanding the nature and purpose of these parts.
- Mechanism: Reassembling the parts to recreate the original system.
While this approach works well for relatively simple or complicated systems (like a clock), it falls short when applied to complex systems. It’s essential to distinguish between “complicated” and “complex” systems:
- Complicated Systems: These systems can be broken down into smaller, manageable parts that follow specific rules or processes. With the right expertise and tools, such systems can be understood and solved deterministically.
- Complex Systems: These systems have many interconnected parts that interact in unpredictable ways. Changes in one part can significantly affect others, leading to unpredictable and emergent behaviors.
As Ackoff noted, “When you analyze a system, you learn something, but you don’t get understanding. The performance of a system doesn’t depend on how the parts perform separately; it depends on how they interact.” Therefore, the FRAM is invaluable for understanding complex systems because it focuses on these interactions and the behaviours they produce, rather than concentrating on the separate behaviour of isolated components.
Examples and Case Studies
To see the FRAM in action, consider our case studies, which provide practical examples of how the method has been used to analyze complex systems. Figure 1, for example, illustrates a FRAM-built model of primary care management for possible sepsis in a hospital ward. Despite its apparent complexity, the staff involved found it accurately represented their work processes. This highlights a common experience: those whose work is represented by a FRAM model often find it intuitive and insightful, even if it appears visually complex at first glance compared to more linear models.
Figure 1: primary care management of possible sepsis McNab et al. (2018)
Interpreting Lessons from Surprises: Incident Analysis
The FRAM was originally developed to address a specific problem: the difficulty of accounting for disproportionate (non-linear) outcomes using traditional accident models based on linear cause-effect reasoning. Conventional accident models, such as the domino, Swiss cheese, and bow-tie models, offer simplified analogies for how accidents happen. These models often blend method and analogy, making them easy to understand but limited in explaining complex causality.
In contrast, the FRAM differs by not relying on such fixed analogies. Instead, it focuses on understanding how systems are supposed to work, allowing us to infer how “normal” accidents might occur when systems deviate from expected performance. An example of this application can be seen in the Macondo Well Case Study.
Quantifying System Outcomes
The literature has yet to fully catch up with recent advancements in the FRAM methodology. Although it is often still seen as a purely qualitative human factors approach, recent developments have added quantification and dynamic capabilities to the FRAM. Traditional methods of quantification, such as fault trees or Bayesian networks, treat the FRAM as a static network of nodes and edges.
But all these current alternative attempts at delivering quantification have treated the FRAM model as a fixed network – nodes and edges of a digraph. (see Figure 2). It is not clear as to why you need a FRAM model to draw yet more boxes and arrows anyway – just draw them(!?). To calculate probabilities, they use FRAM models as networks to apply traditional network models such as Fault Trees and Event Trees, Random Walkers, Petri Nets or Bayesian Nets. However, this approach doesn’t leverage the unique strengths of the FRAM, which can model emergent behavior as the system progresses through various instantiations.
Figure 2 – Jigsaw pieces or cells?
A FRAM-built model represents a superposition of all possible instantiations, making traditional probability calculations less meaningful. For instance, calculating the mean passage time may be useful in time-constrained activities, like changing wheels during an F1 pit stop, but has limited value in most complex systems where activities are not time-constrained.
To properly quantify outcomes, the FRAM should be used to model each applicable instantiation, showing how the model is traversed from entry to exit functions. This stepwise evaluation process, described in detail in the paper below, considers how upstream and downstream functions interact dynamically, providing a more nuanced understanding of system behavior.
Interfacing with Current Tools
While it may not make sense to use the FRAM merely as a front-end for traditional network quantification methods, it is beneficial to use established system analysis procedures to identify functions, tasks, and processes in complex systems. Techniques like Hierarchical Task Analysis (HTA) and Systems Engineering Initiative for Patient Safety (SEIPS) have proven useful in this context. Presentations and papers on these applications provide further insights into their integration with the FRAM.
Recent developments have also explored using the Bow Tie method to identify essential prevention and mitigation functions as barriers, ensuring the safety of systems and subsystems.(Figure 3). Future planned developments will allow current Bow Tie software to interface with the FRAM, uploading data and definitions of barrier types and interdependencies into the FRAM model-building software.
Figure 3 – FRAM built model from a batch process Bow Tie
Summary
The FRAM offers a flexible and dynamic approach to understanding complex systems by focusing on interactions rather than isolated components. Its ability to model emergent behavior quantitatively, and interface with existing analytical tools, makes it a valuable method for analyzing and improving the performance of complex sociotechnical systems. For more detailed case studies and theoretical discussions, please refer to the papers and presentations listed below.