Causal Inference is a Minefield — Here’s How to Navigate It with DoWhy

Author(s): Torty Sivill

Originally published on Towards AI.

Causal inference helps us move beyond prediction and into decision-making but it’s dangerously easy to get wrong. In this article, we explore a real-world business example and useMicrosoft’s DoWhy library to answer causal questions. We’ll learn how to structure a causal analysis and how to avoid common pitfalls that can sabotage your interpretation.

Image from Dewang Gupta, Unsplash

Causal Inference: From Prediction To Decision-Making

AI’s scope of application is accelerating at an unprecedented pace. Over the past decade we have seen countless success stories: AI being used to detect cancer, predict financial trends and weather forecast. Amazon’s recent deployment of AI systems which autonomously route products is an example of a successful application of AI, not just to predict, but to make decisions.

In order to make decisions, we need to know what the causal, not just predictive, effect of one variable is on another, “Does discount actually reduce churn?” In my last article I showed how traditional, predictive machine learning, cannot be trusted to answer these sorts of questions as correlative patterns in data do not guarantee causation.

The Dangers of Causal Inference

Causal inference is the study of the true causal effects between variables. However, the world of causal inference is complex. To correctly… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI