The weather in Dublin was unseasonably splendid last weekend. Taking the DART to Bray, which becomes ever more scenic as the carriages trundle past Dalkey, I decided instead to dip back into Daniel Kahneman’s Thinking Fast, and Slow.
The chapters were to do with a series of fascinating experiments that aimed to pit logic against emotion.
In a sense, it had to do with base probability and set theory. You would have to read the chapter, but in short: it is far less probable that, in Berkley College in the 1970s, that a women would be a librarian and a militant feminist than solely a librarian.
The discovery from the experiments were that, contrary to both logic and probability, people tended to believe more in the least probable option.
In other words, richer descriptions feel more true.
On one level, this was about stereotypes – they are powerful. The profile written for Kahneman’s experiment described a particular kind of woman without any reference to profession – it was only the options that suggested these – yet, despite no evidence being provided for it, most people tended to choose the more detailed option because it seemed to confirm their stereotypes.
Interestingly, the experiments also revealed that when stereotypes are deployed this way, we become especially confused between probability and likelihood. Probability is numeric – it can be calculated. I once asked my data science colleagues the probability of being on a bus with a person who had experienced domestic or sexual abuse – the answer was shocking. But likelihood is different – this is how most people think about probability, which means we call on other sources of information, and cognitive systems – lazy systems – to reach a conclusion.
The learning is this: believability is additive. Our cognitive biases impel us to believe more conjunctions – in this case, being a librarian and a feminist, is more likely because we can, perhaps, picture such a person who, as it happens, plays into our pre-existing stereotypes about certain people. Our heuristics.
The chapter made me think about the ‘consumer personas’ we so often write in this industry. Our goal, always, in writing them is to give as rich and vibrant a description of a ‘typical’ person and their world. They are indented as stimulus to give direction in marketing and advertising campaigns. They attempt to crystallise our mountains of information, data, intuitions into a digestible form that can be used.
But how likely are they, really? How data-driven are they? How statistically representative are they, and does it matter?
But isn’t this just the same thing as our librarian feminist?
As I’m beginning to suspect, a great many of them are convincing exactly because they exploit the cognitive bias identified by Kahneman and his colleagues.
The more rich detail we add, the more ‘alive’ and therefore the more likely these people are to not only ‘exist’ but to represent reality.
This shouldn’t be a problem. We all enjoy fiction, identify with fictional characters as if they were real. It’s a basic, human trait. And we’re a species programmed for storytelling. But apart from the logical and epistemological conundrum the experiments expose, it can also be a problem for business.
So much of business is about numbers – trends, data, profit/loss – placing a bet on a consumer profile that feels convincing but isn’t backed up by research and data can be a serious problem.
At the same time, data-driven personas can be incredibly uninspiring if not injected with a good dose of creativity.
Provided we are aware of our biases, and tread carefully when conjuring ‘consumers’ out of the data cloud as much as thin air, albeit, perhaps intuitively blended with personal experience, hopefully we can strike the right balance.
Because: if the personas we create don’t link with business reality, but equally, if they don’t inspire ourselves and business to bring campaigns to life, the world of marketing could be a very boring place.