Causal Artificial Intelligence? Causal Analytics!
Fundamental of Data Driven decision
Recently, I was reading “Causal Artificial Intelligence” by Judith S. Hurwitz and John K. Thompson. The “Artificial Intelligence” caught my attention. I knew the authors comes from vendor/consulting background and was expecting the content to be more sale-sy. Did not really expect it to inspire this issue but I do gain more ideas and thoughts from the book, plus to be fair the book is technically grounded to a sufficient extent. I admit that I feel the book title is click-bait to me rather.
The book make me look back on the development on Analytics. I am a strong believer in value, as in projects involving data in any business should lead to value. In fact, if I were to be more precise, it should lead to projects that provide values large enough that the company is willing to undertake the costs of building and maintenance costs.
There is this term that keeps popping up in any literature discussing what analytics should lead to to generate value. The term is “actionable insights”. Now I have a problem with this, similar to “AI Governance”, as it does not seem to hit the nail on the head on what Analytics should be doing.
The epiphany comes when I was listening to “Causal Artificial Intelligence”. What Analytics should be working on is not “actionable insights” but rather using data to find and establish causal chains. What do I mean?
At the end of the day, doing Analytics is about making an impact in certain business area, for instance, increasing revenue in a particular niche market or reducing costs without a significant reduction in customer loyalty. In order to solve these challenges without disrupting other parts of business, we will need to understand what are the sphere of influence and how these spheres of influence interact each other. Only upon understanding what and how A causes X effect happen, can we either increase (strengthen the input through the causal chain) or decrease the impact (breaking the chain at appropriate links), depending on business challenges.
Thus the term “Causal Analytics”. What we are doing is through internal, and later external data, to establish these causal chains that are happening throughout the business environment. Through understanding the various causal chains happening across businesses, and keeping a measure on the different links and impact, we establish a list of “levers” in the business. These “levers” can be managed to, hopefully, reach the desired impact, results and value.
Conclusion
When we do Analytics, we must aim to determine the causality, the how and extent of impact. This is where most of the value of doing Analytics is, leading to the term “Causal Analytics”. “Causal Analytics” will lead to data-driven decisions that can be impactful and can be acted upon. Business teams should always keep in mind, that if after doing the analytics, one cannot see the possible causal chain, then it might be good to see if more can be done, so as to continuously and consistently contribute value to business through data.
Your thoughts?
Thank you for reading till here! You can make a “book” donation and drop me some wisdom! Or consider liking or forwarding this email to friends who might benefit from the newsletter. Thank you!
There is also subtlety in the granularity of the causal chain that we should pursue. The cause and effect can be superficially understood (e.g. weather) or going very deep into the biological and psychological level (e.g. marketing). Knowing what decisions you can take to influence is an important stopping rule on how granular one should attempt to go with causal analytics.