Fil Salustri's Design Site

Site Tools


Systems Thinking

Systems thinking and its relation to design.

General Notes

A system can be thought of as a transformer that converts inputs to outputs. Some examples include:

  • an elevator takes people at one floor/altitude as input and transforms them into people at a different floor/altitude.
  • a blender takes separated foodstuffs and transforms them into blended foodstuffs.
  • a stapler transforms loose papers into bound papers by the addition of a staple.
  • a cantilever beam has certain forces and moments as inputs and transforms them into reaction forces, moments, and deflections as outputs.
  • a bottle of water transforms the person who drinks from it, so the drinker is input and output to the water-bottle system.

We can write this “mathematically” as: $O = S(I)$

where O are the outputs, I are the inputs, and S is the system that does the transformation.

Different engineering tasks are indicated by which of the three variables are unknown.

$O = x(I)$

  • The system is unknown, but the inputs and outputs are known. This represents design. The inputs are derived from the operational environment in which the finished product is expected to function. The outputs are the user's requirements for what the product must do.

$O = S(x)$

  • The inputs are unknown, but the outputs and system are known. This represents analysis. The system is a product design, and the outputs are the behaviour of the product (e.g. the mathematical description of its mechanical properties). The goal is often to find things like “maximum stress” - that is, we're looking for the maximum allowable inputs under which the system still behaves as desired.

$x = S(I)$

  • The output is unknown, but we know the inputs and the system. This represents both experimentation and validation. For example, we have a design, and we conduct a FEA to determine if its behaviour (outputs) satisfies our expectations. Also, in scientific studies, the inputs are the hypothesis and experimental conditions, the system is the experimental specimen or theory, and the outputs are the results of observing the system/experiment behaving in some way.

Adaptive systems

“Adaptive systems are composed of different heterogeneous parts or entities that interact and perform actions favouring the emergence of global desired behavior. In this type of systems entities might join or leave without disturbing the collective, and the system should self-organize and continue performing their goals. Furthermore, entities must self-evolve and self-improve by learning from their interactions with the environment.” [BG14]

Generally, this definition takes in societies, biomes, and organizations as well as the usual complex artificial systems.

The problems in the work - and presumably in the field of complex adaptive systems (CAS) - are:

  • The notion of global desired behaviour. Desired by whom? For artificial CASs, the desires are those of the designer. In nature, however, there is no such global agent. What is the equivalent, if there is one, for natural systems? Without understanding this, bio-inspired design will not help CASs.
  • It appears that a research theme in CASs is to develop and study the distributed algorithms that “allow system entities to select the best suitable strategy/action and drive the system to the best suitable behavior….” [BG14] But again there is no overarching control system or controlling agent in a natural system with a sufficiently global view to direct the development of such algorithms.

Adaptive systems ought to be designed bottom-up [BG14].

Superorganisms (which [BG14] model as CASs), like bee colonies, can reconfigure themselves (a property of CASs), but their elements are static over their lifespans. The individual bee doesn't really change much at all over it's life, even though the colony can change dramatically.

So, it is not necessary for the CAS elements to change for the CAS itself to change.

[BG14] offers only a very superficial treatment of natural systems. I disagree strongly with the paper's assertions about the similarities between natural and artificial systems. See notes in hardcopy for details.

Research Issues

  • Polysystem theory: check this google for info.
  • You can't identify the environment of an organism without knowing what the organism is.
    • This is flawed – it's about the boundaries between the organism and the environment, because I believe we perceive the boundaries and fill in the rest.
  • Every organism constructs & alters its environment. This environment likewise affects the organism.
    • This is okay, but it must be re-derived from a different line of reasoning given the previous item.
  • Paper [HO08] describes a course in interdisciplinary system dynamics between social work and electrical/systems engineering. It's awful. There's no quantification; no grounding of goals; no analysis beyond simple anecdotal recounting of students; no direct quotes; lots of bald assertions.

See Also


BG14. a, b, c, d, e M. Bakhouya and J. Gaber. 2014. Bio-inspired approaches for engineering adaptive systems. Procedia Computer Science 32:862-869.
research/systems_thinking.txt · Last modified: 2020.03.12 13:30 (external edit)