Make uncertainty work for you

I focus on the most important form of innumeracy in everyday life, statistical innumeracy – that is, the inability to reason about uncertainties and risk.

– Gerd Gigerenzer, Director of the Centre for Adaptive Behaviour and Cognition, Max Planck Institute

The understanding of uncertainty plays an increasingly important role in tactical and strategic mining studies and mine operational systems.

Fluctuations in the data within the typical mining supply chain are inevitable. Randomness, which is very much a standard geological feature of most ore bodies, can be a source of constant confusion and frustration. Beyond those variations of Mother Nature, we are confronted by the inevitable randomness of the human environment (automated or not).

The problem is that almost all prediction work done within most industries rely on the parameters of the Gaussian bell curve, which ignores large deviations and thus fails to take account of ‘Black Swans’. We as a species, even with our Bayesian brains, are bad at factoring in the possibility of randomness and uncertainty.

We forget about unpredictability when it is our turn to predict, and overestimate our own knowledge. We often desert the warning signs of our own intuition.

Uncertainty is the single greatest design flaw within the current mine operational mindset and predictive systems. We expect linearity in areas that are mostly non-linear. Our systems tend to follow the same philosophy that has defined our Tayloristic management systems which have dominated the corporate world since the beginnings of last century. We are now clearly out of date in so many areas within this new era of big data and complexity.

Whilst this area of discussion is highly relevant in today’s world, and discussion forums are very active and at times emotional, there are avenues to explore in working with uncertainty and natural randomness.

Risk and Uncertainty are not the same
Risk is the situation where there is a set of possible outcomes within a system, and the probability of each outcome is known. Uncertainty is the situation where there is an unknown set of possible outcomes within that system, and the probability of each one is not known.

Probabilities and Uncertainty
From a Bayesian viewpoint, probability is a measure that quantifies the uncertainty level of a statement. Put another way, in the absence of information, all values are equally likely to occur and uncertainty is at a maximum. Furthermore, the universe is an uncertain place and in general, the best we can do is to make probabilistic statements about it. Irrespective if the underlying reality of the world is deterministic or stochastic, we are using probability as a tool to quantify uncertainty (Osvaldo Martin).

Use analytical techniques that work within non-linearity environments
Highly complex “linear” modelling techniques in non-linear systems tend to be dangerous if given a greater credibility than they deserve. They can be useful in gaining a basic understanding, but inevitably they are often no more valuable than a simple cheap statistical study. To truly understand a dynamic non-linear system, you must adopt techniques that are built to handle such complexity and have a consciousness of history.

Prepare for multiple outcomes
Rather than trying to make the one “right guess” as to what will most likely happen. Use a range of techniques to make multiple predictions or scenarios. This is the way any truly innovative process works, and innovation is a good analogy for prediction. Also remember that predictions generate history, and history should be an input into all further prediction and decision processes.

Focus your evaluation of initiatives on the inputs, not just the outputs
Randomness will confound even the best efforts to produce trustworthy results. When assessing an initiative’s success, consider the quality of the decision to undertake it. Don’t rely solely on the actual outcome of the project (bad or good), but take into account the quality of the process that went into its planning and execution. Ensure your process(es) selection is democratic and not driven by internal politics.

Remain agile, and strive to respond quickly
There’s no substitute for intuition, awareness, listening, and detecting events as soon as they happen. Focus on “sense and respond” as an organisation, and empower people to act quickly and decisively. Have a corporate policy that is strong on principle but general enough to be flexible, remembering that actual results WILL vary.

As complexity increases, precise statements lose their meaning, and meaningful statements lose their precision.
– Lotfi Zadeh, Professor of Computer Science, University of California (Berkeley)