So who’s got the answer?

If like me you are always up for an argument and enjoy the pub brawl that is Twitter more than you wish you did, you might have tuned into one or more of the Conservative Party leadership election debates that are ongoing as I write this post.

If you did, you quite possibly found that, like most broadcast political debate, it was a fairly unedifying, if unsurprising spectacle.  These were Tory leadership candidates, but I suspect if they were Labour or Lib Dem, Republican or Democrat we would have witnessed something similar.

As I watched each candidate struggle to give their short, simple answers to large complex problems, it struck me how seldom, if ever, a politician or other public figure, responds to a question by saying “I don’t know?”.

Instead, they seem to prefer to stumble through off-the-cuff responses or offer a series of platitudes, or – because they are trained in this stuff – simply reframe or avoid the question.

Why they do this may be in part explained by work done by University College London in 2016 (  As part of the study, researchers from UCL’s School of Neurology wired volunteers to two machines: one that measured their stress levels and one that could administer an electric shock.  They then asked them to play a simple computer game in which they had to guess the likelihood of finding a snake under a number of rocks and shocked them when they guessed incorrectly.

As the game progressed the volunteers learned the clues by which they could identify the rocks harbouring snakes, but every so often the parameters of the game were changed and the learning process had to start again, creating fluctuating levels of uncertainty for the players.

The experiment showed that players’ levels of stress where highest when their level of uncertainty was highest, higher even than when there was certainty of receiving a shock.  In other words, it turns out that it’s much worse not knowing if a bad thing is going to happen or not than knowing it definitely will.

So maybe our politicians have worked out that a voter will feel more positive towards them if they know what their proposed policies are, even if the voter disagrees with them, then if they admit that they haven’t decided yet what their policy is going to be.

In data science we work with clients to help them solve complex problems by building predictive models.  We know that no model can ever be completely ‘right’ and that we only find the best answers to complex problems by learning from the actions we take in response to them.  But models are useful because they help us explore options and avoid the riskiest courses of action.

So while we don’t want to suggest to clients and colleagues that our models can produce answers that are 100% ‘right’, we also need to help them recognise the anxiety they may be feeling in the face of this truth for what it is, a predictable evolutionary response to uncertainty.  We don’t want to ignore these feeling, but equally they don’t all need to be completely resolved before important work can progress.

By Kevin Cornwell
Categorized as blog

Leave a comment

Your email address will not be published. Required fields are marked *