Correlations can be interesting, even enlightening; however, scientists understand that correlation isn’t the same thing as causation. To highlight that point, Tyler Vigen, a Partner at Boston Consulting Group (BCG), created a site called Spurious Correlations that highlights some interesting, if meaningless, correlations. For example, Vigen shows that, between 1999 and 2009, there was a strong correlation between U.S. crude oil imports from Norway and the number of drivers killed in collisions with railway trains. Nobody believes, however, that dozens of lives could have been saved had the U.S. avoided importing crude oil from Norway. Because correlations can be spurious, making decisions based on those correlations can prove foolish. Taco Cohen, a machine learning research scientist at Qualcomm Research Netherlands, observes, “At its core, science is all about finding causal relations. For example, it is not enough to know that smoking and cancer are correlated; what is important is to know that if we start or stop smoking, then that will change our probability of getting cancer. Machine learning (ML), as it exists today, usually does not take causality into account, and instead is merely concerned with prediction based on statistical associations. This can lead to problems when using such models for decision-making.”[1]
Improving Decision-making
Good decision-making in business is crucial. Bain analysts, Michael C. Mankins and Lori Sherer, explain, “The best way to understand any company’s operations is to view them as a series of decisions.”[2] They add, “We know from extensive research that decisions matter — a lot. Companies that make better decisions, make them faster and execute them more effectively than rivals nearly always turn in better financial performance. Not surprisingly, companies that employ advanced analytics to improve decision making and execution have the results to show for it.”
Good decisions generally begin by asking good questions. Great questions lead to better answers and better solutions. Cohen observes, however, that there are three types of questions that rely on finding causation if they are to be answered effectively. “Being able to answer such questions,” he notes, “is very important in many AI tasks.” Those question types are:
• Associational: If I see X, what is the probability that I will see Y?
• Interventional: If I do X, what is the probability that I will see Y?
• Counterfactual: What would Y have been, had I done X?
Cohen concludes, “Since causal reasoning is key both to our scientific understanding of the world, and our common-sense understanding of it, it stands to reason that causal reasoning is also critical in the quest to build AI.” The staff at Rebellion Research explain, “To correctly predict causality, causal AI must identify the root causes of outcomes, model interventions that could change outcomes, and answer what-if questions. … The ability to reproduce results with ensured high accuracy is [a] benefit of causal AI in machine learning. … These causal AI benefits have so far been employed in different sectors, including asset management, capital markets, manufacturing, retail banking, healthcare, insurance, and telecommunications.”[3] To that list of benefiting sectors, I would add the retail sector. The Enterra Global Insights and Decision Superiority System™ (EGIDS™), using causal AI algorithms, allows clients to generate unparalleled insights, see beyond the horizon, and gain competitive advantage in critical markets leveraging an outside-in approach. The Rebellion Research staff concludes, “Causal AI makes machine learning robust, explainable, and valuable.”
Pursuing Causality
Rafi Katanasho, APAC Chief Technology Officer and Solution Sales Vice President at Dynatrace, observes, “The pace of change in IT environments is now faster — recent research shows Digital Transformation accelerated for 90% of organizations in the past year alone. The complexity of modern environments was already judged as being beyond human capabilities alone in 2020. Add another three years of change and it’s perhaps no surprise to see organizations leaning even more heavily on AI to bridge operational gaps while enabling the current pace of innovation to proceed. What this is driving, however, is a review of the AI — particularly Machine Learning — algorithms in use, particularly the extent to which they’re able to create the accurate insights needed to reduce time and effort of operations teams in diagnosing and remediating issues or performance bottlenecks.”[4]
The CityLife staff notes that, as effective as machine learning can be, it does have limitations. They explain, “Traditionally, AI has been focused on identifying patterns and correlations in data, which has proven to be highly effective in many applications, such as image recognition, natural language processing, and recommendation systems. However, correlation does not necessarily imply causation, and this limitation has become increasingly apparent as AI researchers strive to create more intelligent and autonomous systems.”[5] This limitation, they insist, motivated analysts to pursue causal AI and the effort has accelerated. They believe it has created a revolution. They explain, “The Causal Inference Revolution seeks to [develop] AI algorithms that can not only identify correlations but also understand the underlying causal relationships between variables.”
As the Enterra Solutions® team pursues the advancement of Autonomous Decision Science™ (ADS®), we understand that finding causation is extremely important. Because selecting the right algorithm can be difficult, we use Massive Dynamics™ Representational Learning Machine™ (RLM). The RLM helps determine what type of analysis is best-suited for the data involved in a high-dimensional environment and it accomplishes this in a “glass box” rather than “black box” fashion (i.e., it makes decisions explainable). One reason causal AI remains in its infancy is because finding causation is not easy. The CityLife staff notes, “It is often difficult to obtain data that explicitly demonstrates cause and effect relationships. This has led researchers to explore alternative approaches, such as using observational data, which is more readily available but may be subject to confounding factors and biases. To overcome these challenges, researchers have been developing new methods and techniques for causal inference in AI.”
Concluding Thoughts
The CityLife staff concludes, “The Causal Inference Revolution has already begun to show promising results in various applications. … As AI continues to advance, the importance of understanding causality will only grow. By incorporating causal inference into AI algorithms, researchers can create more intelligent and versatile systems that can better understand the complex relationships that govern our world. This, in turn, will enable AI to make more informed decisions and predictions, ultimately leading to more effective and beneficial applications across various domains.” Katanasho adds, “Causal AI can answer conceptual and counterfactual questions to model a wide range of potential outcomes. This form of predictive analytics enables organizations to anticipate situations and prepare contingencies.” In a fast-paced business environment, the more rapidly and accurately businesses can make decisions the more successful they will be. Causal AI will prove to be an essential tool in the emerging business environment.
Footnotes
[1] Taco Cohen, “Is causality the missing piece of the AI puzzle?” OnQ Blog, 6 September 2022.
[2] Michael C. Mankins and Lori Sherer, “Creating value through advanced analytics,” Bain Brief, 11 February 2015.
[3] Staff, “What Is Causality In Artificial Intelligence?” Rebellion Research, 8 September 2022.
[4] Rafi Katanasho, “Causal AI makes for more accurate decision-making,” Intelligent CIO, 12 May 2023.
[5] Staff, “The Causal Inference Revolution: How AI is Learning to Understand Cause and Effect,” CityLife, 25 June 2023.
