Recycled Bullsh-IT

We’ve heard a lot over the last year about ‘following the science’ or ‘following the data’. This is normally spouted by po-faced politicians or their spooky scientific sidekicks while they grip their lecterns and gurn into the camera, interspersed with forlorn and random ‘next slide please’ requests.

 We Are Scientists

Sometimes this has morphed into following the (data) scientists. This has been entertaining (discovering illegal trips to second homes or infidelity secrets), but not so illuminating, as it transpires that scientists are human too. And what do humans do? Make mistakes. Hold prejudices. Have their own agendas. Be rude to others. You start to wonder who and what to believe. Other than 5G gave us COVID, of course.

 New Order

Data are important and should be the foundation of all important decisions. It’s also vital to apply the Scientific Method to the data.  And use critical mathematical techniques to glean unambiguous information from the data. However, as I touched on in It’s Not Data Science, big data can be difficult to understand and use effectively. Not only are there staggeringly large amounts of noise, but also the data are rarely complete, consistent or correct.  Conclusions should be given as probabilities, with margins of error, and all credible answers documented.

 No Doubt

Most people crave certainty, so they look for unequivocal facts giving clear guidance to Do the Right Thing. Unfortunately, many decisions have to be made on imperfect data, with subjective interpretation by a series of humans all of whom have differing perspectives and desired outcomes. From imperfect data, decision-makers feel the need to state ‘facts’, rather than explain the balance of possible vs probable.  It’s no coincidence that the best-selling book on statistics ever is “How to Lie with Statistics” by Darrell Huff (1954).

D-ream

Add in the current fad for people wanting to state ‘personal truths’ (i.e., opinions with no basis in fact, made famous in some high profile interviews), then getting to an objective truth is nigh on impossible. It’s fine to have reasons for a decision, but these are rarely irrefutable truths, but a highly subjective lens on the subset of data considered.

 Black Box

What about AI? Surely with all that machine-learning artificial intelligence, the information will be fully objective, and correct? Unfortunately, we’re finding that machine learning picks up our conscious and unconscious biases, making AI pronouncements every bit as partial as those of we humans. But without the compassion or common sense. Beware ‘computer says no’.

 Altered Images

Many leaders believe that they need to appear decisive, and mistakenly assert fuzzy findings and unclear conclusions as firm facts. If and when the opposite proves to be the case, they will blame the data (which may not have changed) or their experts (who warned them of the range of scenarios). Obviously, leaders are paid to make decisions, particularly in difficult situations with no clear solutions. The best leaders state up front that there were insufficient accurate data to draw any firm conclusions, but on the balance of risk and benefit they would pursue course A for now.

 The Pretenders

Mark Twain popularised the phrase: “Lies, damned lies, and statistics” over 100 years ago. Perhaps we should change this to “data, big data, and dunno”.

John ‘Maybe, Who’s Asking?’ Moe