Complexity: Absolute or Relative?
I spent the last few days in Santa Fe, absorbed in discussions of complexity, with particular emphasis on healthcare delivery. I have delved into this topic for quite some time. Three decades ago, we published our studies on the complexity of troubleshooting – figuring out the source of unfortunate symptoms, e.g., why your car won’t start.
Sponsors of our research asked us to devise a metric for the complexity of a troubleshooting task, which they intended to use to match to the complexity processing abilities of maintenance personnel. Pursuit of this goal led us to conclude that complexity is related to the intent of the person asking the question or performing the task, as well as the knowledge and skills of this person.
To illustrate, let’s say you purchased a Boeing 747 to use as a paperweight. From this perspective, this complicated airplane is just a large mass, pretty useful for keeping errant papers on a very large and structurally sufficient desk. In contrast, if you made this purchase with the intent of operating and maintaining the aircraft, the Boeing 747 is much more complex than your unwieldy paperweight.
This insight leads to a fundamental conclusion. Complexity has to be defined in terms of a relationship between an observer and an entity. The observer’s intentions, knowledge, and skills frame the assessment of the complexity of the entity. Thus, complexity is relative rather than absolute. Consequently, for example, we can only assess the complexity of a troubleshooting task relative to the personnel involved in the task.
I have discussed this conclusion in many of my talks over the past ten years or so. Roughly 90% of the people with some level of expertise in the topic agree with me. The other 10% say something like, “What you are saying makes sense, but what about real complexity?” These people are usually physicists who firmly believe in the absolute nature of complexity.
Many of those researching complexity construct network diagrams of the elements and relationships among elements of engineered, organizational, and natural systems of interest. They calculate various metrics associated with these network diagrams and then argue that these metrics reflect the inherent complexity of the systems of interest. I have done this as well, with the explicit acknowledgement that these network models reflect my intentions, for instance, to predict the difficulty of driving in different urban environments.
There are no intention-free models. Every model is constructed with the intent to analyze, assess, or predict some set of phenomena. Any properties of these models used as complexity metrics reflect the intentions of the modeler(s). This is as essential today as it was for Newton, Darwin, and Einstein in past centuries. Absolute complexity is a chimera.
