#52: Problem Solving – Critical Thinking

Here’s a thought experiment: you get to create a dream team of problem solvers in your organization. What would the team look like? Who would you have on your team? Over the last few weeks, I have talked about various dimensions of problem solving: this one and the next few will focus on the importance of critical thinking, and perhaps most importantly, asking questions. And for this, we need to turn to an unlikely source (at least in modern education) – philosophy. I would recommend at least a couple of philosophers in your dream team – and while you are at it, you might want to consider David Hume and Socrates

 David (Who?) Hume

Hume was one of the many 18^th century philosophers who thought and wrote extensively on a variety of topics – the one that interests us today is empirical thought. He pushed the idea of empirical evidence as the primary source of knowledge, and tried to debunk the notion that previous experiences should influence our current decisions. Which goes smack against the Bayesians (and the priors!) – and the reason why you want to counter-balance your data scientists (who mostly lean towards a Bayesian mindset) with someone like Hume. Here is what he had to say in his ‘Treatise on Human Nature’:

All the objects of human reason or enquiry may naturally be divided into two kinds, to wit, relations of ideas, and matters of fact. Of the first kind are the sciences of Geometry, Algebra etc. … [which are] discoverable by the mere operation of thought … Matters of fact, which are the second object of human reason, are not ascertained in the same manner; nor is our evidence of their truth, however great, of a like nature with the foregoing.

Someone coined the term ‘Hume’s fork’, which is often stated in such a way that statements are divided up into two types – which are distinct, separate and you cannot induce one from the other:

• Statements about ideas. These are analytic and knowable a priori. In other words, these are statements that belong to logic and mathematics, which we all know to be true. For instance, “a triangle has three sides”, “all bachelors are unmarried”, and so on.
• Statements about the world. These are synthetic and knowable a posteriori. This is where it gets interesting: Hume claimed that our belief in cause-and-effect relationships is not always grounded in reason, but rather arises merely from habit or I suspect, intellectual laziness. For instance, “all swans are white” (which was true until someone actually spotted a black swan, which entirely falsified the premise). Nassim Taleb’s ‘Black Swan’ must have been influenced by Hume’s work.

He also articulated this as the ‘is-ought’ problem: if someone has access to ‘non-moral and non-evaluative factual premises, the reasoner cannot logically infer the truth of moral statements.’ In other words, the leap from facts (‘is’) to inferences (‘ought’) is not straightforward, if possible at all.

So what?

You might say this is all metaphysical claptrap – but there are two very important points that are relevant to building effective problem solving teams:

Mistaking Data for Insights

Obvious though this may sound, the entire ‘big data’ movement has led many organizations down this path. The idea was this – build a data warehouse, or even more expansive, a data lake; install a BI/reporting environment on top of the data; hire Analysts who churn out dashboards and reports and voila, you have a data-driven enterprise. This is the classic mistaking of the ‘is’ to the ‘ought’ – and it goes much deeper than a poor return on investments. It often ends up taking organizations down rabbit holes that are often difficult to get out of – especially, if they end up carrying legacy data and practices. A large B2B firm never bothered to ask of its customer data: “What ought to be our Install Base?” – instead, the question posed was “What is our Installed Base?” This might sound like a semantic difference but it ended up with the creation of a massive enterprise data lake with customer data from multiple product teams, each with their own siloed definitions. Taking that and stitching a complete customer install base view turned out to be a gargantuan problem – and through that process, the company lost millions of potential revenue generation opportunities (of cross-sell, bundling of service options etc.) As Hume would have pointed out, the idea of crossing from the ‘is’ to the ‘ought’ on Hume’s fork is an exercise in futility. In an alternative reality, had the question started with “what ought to be our install base?”, it might have begun with a much better starting point: a conscious effort to pull datasets from individual systems that works towards stitching together an enterprise customer view. As they say, the ‘choice at origin’ matters.

The Falsifiability Principle

The idea is simple, but pivotal: every theory or insight must have the capacity to be contradicted by evidence. And when that happens, be ready to throw it out. Clearly, not enough of this happens – and analysts are especially prone to fall into this trap. And even when this does happen, people don’t bother to focus on the right evidence. Back to our B2B example: contract renewal is perhaps one of the biggest focus areas for the Account managers. Propensity models are built to predict the probability of non-renewal – and on an ongoing basis, the evidence of the success of these models is the reduction in churn rates. And any evidence to the contrary should invoke the falsifiability principle – send the teams to the drawing board to re-evaluate the models. While Hume would have approved of that, he would also have been quick to point out that the cause-and-effect here is tenuous (how do you know that the model recommended actions actually impact the churn rates?) and in any case, isn’t it too late to figure out the design flaws in the barn gate after the horse has bolted? Instead, it might be better to look at leading indicators with a shorter feedback loop – so that it becomes possible to recalibrate insights on an ongoing basis as evidence keeps coming in. It might be better to think of action recommendations (Next Best Action) based on intermediate signals – e.g. is there a jump in the number of issues logged by the customers? Are the customers declining to meet the Account Manager or even more fine-grained, is there a change in how the declines are happening? The point that Hume (and a whole lot of subscribers to the skepticism school of thought) have made: it is better to rely on “skeptical solutions” and always challenge them with “skeptical doubts”.

So there we have – Hume in the team would definitely keep everyone honest. One thing with having a philosopher though, is that they can be annoying – any philosopher worth his or her salt enjoys being a gadfly. Next week, we will look at the ‘original’ gadfly – Socrates and what it would be like to have him on your team. And can you end up questioning too much and not getting on with it? Socrates was certainly accused of that – what if you had someone like Socrates on your team?

PS: this thought experiment is not particularly original (although I wish). I am reading a hugely enjoyable book called ‘Socrates Express’ by Eric Weiner where he imagines how the greatest philosophers from history would react to the current times. Fantastic premise, and much needed too!