As computing systems become more actively involved in societally essential areas such as healthcare, education, and government, it is crucial to accurately forecast and comprehend these interventions’ causal repercussions. Traditional machine learning algorithms based on pattern recognition and correlational analyses are insufficient for decision-making without an A/B test.
To fill this gap, Microsoft researchers created a platform that executes the process of causal inference analysis from start to finish to assist data scientists in better understanding and applying causal inference. They developed the DoWhy in 2018. Since then, the library has been doing precisely that, cultivating a community committed to using causal inference principles in data science. “DoWhy” is a Python package that attempts to encourage causal thinking and analysis, many ways machine learning libraries have done for prediction. DoWhy provides a four-step interface for causal inference that focuses on clearly modeling and confirming causal assumptions as feasible.
Traditional machine learning approaches aim to anticipate a result. Consider a public utility business that wants to minimize their customers’ water use using a marketing and incentives campaign. The success of a rewards program is difficult to assess since any drop in water consumption by participating consumers is masked by their decision to engage in the program.
Suppose we see that a rewards program member uses less water than others. How do we know if the program motivates their lower water consumption or if consumers who were already expecting to cut their water usage choose to join it? Given information about the determinants of consumer behavior, causal approaches may separate confusing variables and determine the impact of this incentive program.
This collaboration between Microsoft and AWS applies concepts of Causal Inference and The Book of Why: The New Science of Cause and Effect of professor Judea Pearl. He is a Global Enlightenment Leader.
Boston Global Forum honored Professor Judea Pearl with the 2020 World Leader in AIWS Award.
