Menu Close

Can human intuition ever trump AI?

Tim Lewis AI

Well-designed AI provides a level of objectivity of which humans are incapable. But while there are times when AI knows best, human input is the magic ingredient.

Humans make decisions based on heuristics – simple rules of thumb. For most of our existence as a species, we’ve been hunter gatherers for whom the ability to make split-second “fight or flight” decisions was the difference between life and death. But because human decision making bypasses complexity, it’s notoriously fickle, often depending on little more than sentiment, personal biases, and “gut instinct”.

Our reaction to big data has been to humanise it – to incorporate it into our very human decision-making processes. We’ve developed ever-more sophisticated ways to extrapolate and summarise our data feeds, but Eric Colson, former VP of data science and engineering at Netflix, says it’s time we stood aside and let AI make some of the choices for us.


“The mind can handle sales numbers and average selling price rolled up to a regional level. It struggles or shuts down once you start to think about the full distribution of values and, crucially, the relationships between data elements,” writes Colson for Harvard Business Review. Summarising complex data destroys its nuance or clouds its true meaning, and sometimes the categories we create for our summaries are meaningless. 

We like simple linear relationships because they’re easy for us to understand, but sometimes the data suggests other, more complicated patterns of connection. Sometimes, too, we think up creative explanations for an apparent trend, where there is, in fact, no relationship. The way we summarise our data plays into our biases.

“For routine decisions that rely only on structured data, we’re better off delegating decisions to AI.” It’s better at interpreting micro-level data, better able to identify market segments too small for us to distinguish, and better able to deal with non-linear relationships, be they “exponential, power laws, geometric series, binomial distributions, or otherwise”. 


As long as we scrutinise data-generation parameters for inbuilt biases which could lead to unfair decisions, AI has the ability to read the data objectively. But humans have the vision – there may be times, for example, when AI suggests investing in sales to maximise ROI, but the better strategy is to maintain quality in line with the company mission statement. 

AI is about generating alternative choices based on a true evaluation of the data. “Values, strategy and culture is our way to reconcile our decisions with objective reality.”

AI offers the very real opportunity to circumvent the weak spots in human decision making. As long as data sets are well designed, AI can provide unbiased insights which humans can choose to act on – or not.

Source Article: What AI-Driven Decision-Making Looks Like
Author(s): Eric Colson