Baseball has always had statistics. They’ve been studied to excess and its auditors have immense data directly related to every player. Not so with CSA BASICs.
The FMCSA built CSA and it routinely lacks the complete data tov reveal the behaviors the Agency – and insurers, and shippers – are interested in. What the Agency does instead is substitute stand-in data, or what we might call proxy data. They draw statistical correlations between a motor carrier’s type of operations and its potential for safety. These correlations discriminate. Whereas baseball stats pour in daily for more than 6 months a year; and they can feed back inconsistencies into the model, redefining it as they progress, CSA scores are static; there is no mechanism to correct errors (let’s not even bring up dataQ appeals). Conditions and outcomes change or evolve in court somewhat closer to the truth, so must the model the penalties are based on.
CSA, cloaked as it is in a great deal of mystery, with only chance encounters delivering outsized results, relies heavily on a handful of “test” results, which is so very far from algorithmic modeling. But yet, CSA purports to predict outcomes – i.e., crashes. These “predictions,” unfortunately, guide the discussions of shippers, DOT inspectors, and insurers.
There will always be miscalculations in CSA evaluations because the models used are just simplifications. No model can include all the world’s complexities or nuances of human behavior. Inevitably, a lot of important stuff gets left out – like communications in operations, variable ground conditions, and interactions with other parties, namely shippers and receivers.
To be frank, CSA BASICs is a toy algorithmic model that abuses truckers who all operate on the slimmest of margins. The Agency makes choices about what’s important enough to include, simplifying the world of trucking into its own version of reality, not a true life version where real life decisions and actions are and can be made every day. CSA reflects the goals and ideology that the Agency imposes onto its “safe trucking” mandate. CSA scores are opinions embedded in arbitrary mathematics. What the Agency is trying to accomplish – saving lives – the very definition for its existence, is, of course, more than simply admirable. But by blending in arbitrary measures of “success,” the CSA model hunts down arbitrarily determined data. The CSA model itself becomes a belief, relied upon by those who serve and use the trucking industry. It has not eliminated DOT-based human bias; it’s only camouflaging its bias with its mysterious predictive math calculations.
To sum up, CSA presents three elements of false beliefs: it’s opaque; it uses false scales to measure safety, and reliance upon it definitely causes damage. When the volume of data multiplies, CSA scores-can’t decipher a meaning. They just generate more inaccuracies. Carriers are being coerced, threatened with livelihood – and no recourse exists in the system that shuts down a carrier or imposes an onorous penalty.
People – not algorithms – still decipher meaning from events. CSA, for all its supposed magical abilities, is still not equal to a group of people who can sift through false indicators and bypass CSA’s wishful thinking.
CSA, with great fanfare, was simply a human-derived formula deliberately wielded to impress, rather than clarify the data it receives. In the insurance world, its created a flock of underwriters who are in the rubber stamp business, formulating opinions by proxy, and which opinions become self-reinforcing. When you create models from proxies, it’s easy for users to game it to their desires.
CSA suggests it relies on its efficiencies. It feeds off data that can be counted and measured. But not necessarily fairly or accurately because that data are incomplete qualities and subjective concepts that were built into the original algorithm, and now are essentially unassailable DOT “truths.” Fairness is not calculated; instead CSA calculates great unfairness and causes collateral damage. Many truckers are not singled out as others are, either because of the overly zealous commercial enforcements efforts in known jurisdictions, or perhaps the size (small or large) of the operation doesn’t warrant the time needed to truly gauge a safe operation.
These primitive algorithms have a real, deleterious effect on motor carriers, shutting some down; causing others to pay exorbitant insurance premiums and fines. All based on incomplete, human-biased algorithms. Knowledgeable persons can and do a much better job at determining safe operations and crash predictability. The scores should come with an explanation that they are entirely based on opinion, innuendo, influence groups, and are not to be taken literally, as there is no scientific basis to the conclusions.
Call for help. 602 900 1800