Let us do a thought exercise and quickly examine energy management from a data science perspective. We would be well-served to first make the distinction between data and parameters:

1. Data is empirical, meaning it is metered or measured and is continuous in nature, subject to changing value over time.
2. Parameters are artificial thresholds laid upon the data as one sheet of overhead projector transparency film would overlay another in order to gauge if performance conforms to a desired outcome. Parameters do not reside on the same plane as data.
   

You would be surprised how little this distinction is acknowledged or understood and what the implications can be.

In the field of data science, an arbitrary parameter is commonly accepted as a reasonable starting place in the early stages of data examination (defining cluster boundaries by “eyeballing it” for example). It is not however accepted as a destination for data examination (ultimately the data should express through the model where the boundaries are among clusters) and I think that gets lost in the energy management domain. There is just enough cross pollination of talent among the energy management and data science domains for an individual to achieve but also impair because one can slip seamlessly from starting data examination with an arbitrary parameter into full blown Cognitive Bias unconsciously baked into an Energy Control Measure.

Cognitive Bias is akin to an invasive species that can overtake the habitat of a native species and measures must be taken to eradicate it while preserving a proper energy management habitat. It has many manifestations as there are about 170 varieties of Cognitive Bias hence the invasive species analogy. We’ll just touch upon just a few and organize them in a simple hierarchy to illustrate.

The essence of Cognitive Bias is to make skewed decisions based on pre-existing factors such as personal experience or preference rather than on the data and other hard evidence. Let us now imagine playing the role of Energy Manager and that as part of as part our R&D duties we are expected to prospect and model an Energy Control Measure affecting our national portfolio and see how Cognitive Bias can contaminate our ECM.

• [an error occurred while processing this directive]Sampling Bias is a bias in which a sample is collected in such a way that some members of the intended population are less likely to be included than others, thereby leading to skewed conclusions when applied to the whole population. For our ECM the median age of the facility in our portfolio is 23 years old but the sampling unintentionally reflects a disproportionately high number of facilities less than six years old within the portfolio.
  

• Selection Bias is introduced by the selection of individuals, groups or data for analysis in such a way that proper randomization is not achieved, thereby ensuring that the sample obtained is not representative of the population intended to be analyzed. For our ECM, we have five different building prototypes to account for but we did not properly classify for prototype so our selection was not as random as it should have been.
  

• Survivor Bias is the logical error of concentrating on the subjects that made it past some selection process and overlooking those that did not, typically because of their lack of visibility. For our ECM we model only on facilities that met a certain threshold (choose your measure) with the theory that meeting this threshold is indicative of being causal rather than coincidental. It is a potential false assumption that can lead to overly optimistic model.
  

• Confirmation Bias occurs when people actively search for and favor information or evidence that confirms their preconceptions or hypotheses while ignoring or slighting adverse or mitigating evidence. For our ECM we succumb to our lesser selves and take a low rung approach and intentionally search for and assign more weight to evidence that confirms our hypothesis, while ignoring or assigning less weight to evidence that could prove to be dissuasive.

Sampling and Selection biases occur in data preparation prior to modeling. These are simple biases of omission where perhaps we skipped a step or got lost among the nuts and bolts. The solution is to apologize, backtrack and reapproach. Embarrassing but not fatal.

Survivor and Confirmation biases are biases of commission. Survivor Bias is based on a false assumption and its solution may entail starting all over again and perhaps we will have to answer for extending the deadline. Confirmation Bias is a conscious choice, perhaps intended to accelerate the pace to meet a deadline while masking evidence that the ECM is contraindicated. It is the lid atop the boiling pot.