If it’s felt like the excitement around people analytics has died down, well just enjoy this moment to catch your breath, because important new approaches are in the works.
Perhaps the most tangibly exciting of these new approaches is round the area of what’s known as ‘causal mapping’.
Causal mapping combines the insights about correlation from machine learning with both human and machine-generated insights into causation.
The result will be much more robust models to guide effective interventions.
The idea of causal modeling is self-evident even to children.
They understand that tipping over a glass of water will cause a spill and the fact you took an umbrella did not cause the rain.
However, this intuitive understanding has – thus far – been hard to model mathematically, but it is advances in mathematics that make causal modeling important for analytics.
We all know that correlation is not causation, but we may be unaware that many statisticians took this so seriously that they wouldn’t talk about causation at all.
This lack of an ability to create causal models has limited the effectiveness of analytics to provide advice on which interventions will achieve the desired results.
The link to machine learning
Causal modeling is usually mentioned in conjunction with machine learning.
I believe this is because causal modeling will help unleash the power of machine learning.
One way it helps is that it gets rid of spurious correlations that undermine the accuracy of machine learning predictions.
It also provides explainability, so that we can be more confident about what the machine-learning model is predicting.
Finally, it allows us to combine human insight with the findings from machine learning to help plug gaps where we do not have good data.
All in all, causal modeling and machine learning are a natural match.
Where you can learn more
An excellent book on the birth of the science of causal modeling is “The Book of Why: The new science of cause and effect” by Judea Pearl and Dana Mackenzie.
Pearl won the Turing Prize for his work on causal reasoning.
Yes, there is a moderate amount of math in the book, but you don’t need to follow the math closely to understand the main points.
There are also good YouTube videos available, search for Causal AI videos by CausaLens.
Conclusions
We already have the theory and tools that we can start doing some useful work in causal modeling.
The exciting thing is that this field is still in its early stages.
We can expect new and better tools to help us apply this in the years ahead.