Authors: Mateusz Zawisza McKinsey & Company

At McKinsey we strongly believe that machine learning and artificial intelligence (AI) aren't just about predicting the future: They're about shaping the future by making positive changes.

To date, machine learning and AI tools have been used to predict diverse business and social phenomena – and proved themselves excellent at the job. Their predictive capability is highly valued as it allows us to prepare for future events, for example, to make inventory decisions addressing future demand.

But many decision-based problems relating to issues that we face on a daily basis cannot be resolved purely by prediction. For instance, predicting the high probability of client churn doesn't mean unambiguously that a telco operator should target this client with a retention marketing campaign. Indeed, that may not be the right way to approach the clients at all. What we need is not so much a powerful predictive engine as the ability to assess the impact of our decisions.

Today's data scientists need the necessary knowledge and skills to assess the impact of their business and policy decisions. This gives them full control of decision-making processes and enables them to shape the future. Such methods had been studied in the past mostly within econometric and graphical models. Recently, these methods have been extended into to machine-learning models that are proving to be more robust and efficient than their predecessors.

The workshop introduces what is known as "causal machine learning". We then apply this approach in a team competition based on a real scenario. We end with a summary of lessons learned from the competition and key takeaways for participants.

1.Introduction to decision problems and methodology (60 minutes) 2.Team decision competition (90 minutes) 3.Summary: Scoreboard and lessons learned (30 minutes)

• Basic understanding of machine learning and models at an undergraduate level; see, for example, James et al. (2013)
• Basic knowledge of R programming language

• Angrist, J. D., & Pischke, J. S. (2008). Mostly harmless econometrics: An empiricist's companion. Princeton university press.
• Ascarza, E. (2018). Retention futility: Targeting high-risk customers might be ineffective. Journal of Marketing Research, 55(1), 80-98.
• Athey, S., Tibshirani, J., & Wager, S. (2019). Generalized random forests. The Annals of Statistics, 47(2), 1148-1178.
• James, G., Hastie, T., Witten, D., & Tibshirani, R. (2013). An introduction to statistical learning: with applications in R.
• Pearl, J., & Mackenzie, D. (2018). The book of why: the new science of cause and effect. Basic Books.
• Taddy, M. (2019). Business data science: Combining machine learning and economics to optimize, automate, and accelerate business decisions. McGraw Hill Professional.