The Why Conundrum - Practical causal inference and discovery with python





When we process raw data, it transforms into "information", and when information is processed, we get "knowledge" and when knowledge is processed, we get "wisdom". In this journey, the question "why" acts as the compass that guides a statistician to travel from mere knowledge land to profound understanding.

In this talk, we will take a hands-on and practical approach towards explaining the concepts of Causal Inference and Discovery with python.

In this talk's context, Causal Inference is the process of applying data science/ML libraries along with CI libraries such as doWhy and econML to determine whether a change in one attribute causes a change in another thing. This is a generic deduction process moving from generics to specifics.

Causal Discovery is the process of applying data science/ML libraries to identify and explore causal relationships without prior hypothesis. This is a induction process moving from specifics to generics. We will highlight how we can use observational data and analyze it with domain knowledge to come up with causal models.

In the spirit of Plausible reasoning, the theme of this talk will be the following: "Experience modifies beliefs. A sentinel being learns from experience. A good scientist endeavors to extract the most correct belief from a given experience”

In the demo section: We will show case the open source python libraries being applied on practical causal experiments to answer questions such as: 1. Why does a computer system get infected by a virus? 2. Why was a virus not able to infect a system running a particular software? 3. Why is a software running sluggishly?

and we will empower the audience with the practical knowledge to use the framework to use against their own data corresponding to the problem domain they are working on to answer causal questions.

In the process, we will also cover Prof. Judea Pearl's ladder of causation and explorer Pearlian perspective of causal inference and discovery with dowhy calculus.


The session will cover everything that needs to be known to understand the concepts. Basic statistical inference concepts Basic programming concepts Basic Numpy, Scipy knowledge is prefered

Content URLs:

Speaker Info:

Dinesh Venkatesan is a Logician & Mathematician presently working as security researcher at Microsoft. He has been in the cyber security industry for over 17 years working with Google, Symantec and HCL Technologies and has published numerous blog posts on malware analysis. He is a specialist on the mobile threat landscape and desktop security threats and has discovered multiple vulnerabilities in Android framework layer, responsibly reporting it to Android and helping to make the OS secure. He has hands-on expertise in writing generic detection and cure routines for prevalent malware families. He is on an active lookout for collecting threat intel about sophisticated attacks and keen on researching various threat actors and developing useful insights into malware evolution.

Speaker Links:

RSA conference : Android Security Summit: eBPF Summit 2020: eBPF Summit 2022: Microsoft blog:

Section: Data Science, AI & ML
Type: Talks
Target Audience: Intermediate
Last Updated: