Prospect evaluation is an activity that often deals with very large uncertainties because of incomplete information and/or fragmented knowledge on critical parameters (or worse we might not even be aware of what the critical elements are). To resolve this conundrum, we embark on many studies and build models by which we hope to constrain these uncertainties and derive better predictions. These activities lead in general to greater complexity of our predictions and this very complexity can be harmful.
William Heath Robinson (31 May 1872 – 13 September 1944) was an English cartoonist and illustrator best known for drawings of ridiculously complicated machines for achieving simple objectives. His delightful inventions illustrate this better than any words:
by William Heath Robinson
The problem with complexity is that it negatively affects Efficiency and Accuracy:
- It reduces the Efficiency of our work, because building and checking complex models/evaluations will always take a lot of time.
- It reduces the Accuracy of our predictions, because we tend to lose sight of the objective in the face of increasing complexity, amplifying the scope for human error.
In addition, complexity has the potential to disguise ignorance (it works like a magician’s sleight of hand and can de-focus our attention from the things that really matter). We should not confuse Complexity for Predictive Value, which is the real objective of scientific inquiry. In fact, focusing on Predictive Value often requires shedding Complexity. Making complex models is as easy as it is appealing to many of us, however simplifying complex models can be a difficult task. To avoid the trap of Bizarre Sophistications (BS), a good workflow for creating predictive models is to:
- Start simple.
- Keep it simple by always asking the question if added complexity increases the predictive value (it often does not).
- Simplify. Given that most models tend to become more complex with time, investigate once in a while if simplifications (that do not effect erode the predictive value) can be made.
Managing the human tension between Accuracy versus Complexity is an essential ingredient of our GoExplore Philosophy. Our software tools, are inherently designed to focus on accuracy and only add more precision (and the related complexity) if it demonstrably improves the overall Predictive Capabilities of the application.