Causal Python with Alex

I find it surprising that by now there are probably thousands (tens of thousands?) of people on the internet who are willing to tell you that food like this is a "poison" and you should eat something like raw meat spiced with a bunch of supplements instead to be healthy.

What do you think about this trend?

Málaga vibes con 💙❤️

One of my favorite cities of all time with my favorite person ☀️❤️🌏🌊

🤤

With at

Causal Inference and AI community?

Time for the fifth out of our 5 January announcements.

In January we'll release a beta version of the Causal Bandits Community.

If you want to learn with others, exchange information, discuss particular topics (causality in Python, R, deployment of causal models in production, and more), this place is for you!

Later this month, we'll start with invitation-only access for 100 people and then -- if there's enough interest -- we'll scale it further.

Answering the requests from an earlier survey, I want to make the core functionalities free with paid access to some more premium features.

What do you think about this approach?

🚀 If you want to be updated on the community, leave your email here: https://vist.ly/3ms3u7c

If you want to learn the basics of causal thinking, here's something for you.

Time for announcement 4 out of our 5 January 2025 announcements!

In January, we'll release a free mini-course on causality.

It's a high-level introduction to causality with elements of Python and references to modern AI.

If you ever wondered why a machine learning model failed in production despite excellent cross-validation performance or why so many factors found in investment literature fail to bring expected returns, this course is for you.

If you're more advanced, I hope that the insights about agents, active learning, and the limitations of randomization will be valuable to you.

If that sounds interesting, join the waitlist here: https://vist.ly/3mrz7sh

You'll receive an email when the course is available.

Hope to see you in the course!

Hey folks!

Today it's time for the 3rd out of our five January announcements.

In January, I’ll be starting work on the second edition of "Causal Inference & Discovery in Python".

What to expect?

Updated code, refreshed core content, and exciting new sections and chapters!

Here are a few of the topics you can expect to find in the new edition:

🔥 A section on LLMs with axiomatic causal training

🔥 A new chapter on causal inference under hidden confounding

🔥 New sections on causal discovery: benchmarking, working under violated assumptions, causal discovery with Transformers and score-matching

🔥 An entirely new chapter on causality in business

And if there's a topic that is not listed here and you'd like to see it in the new edition, share your proposals here: https://causalpython.formberry.io/cidp-2nd-ed

Hi folks!

It's time for the 2nd of our 5 announcements for January 2025!

I had three realizations recently:

1. Our causal community has many great and talented people who don't necessarily have a large platform to share their thoughts

2. Our Causal Python Weekly newsletter is mainly written by me personally, and so the content reflects my perspective on causality which is just one of many possible

3. One of the core goals of the newsletter is to facilitate building bridges between people with different backgrounds in causality and causal modeling.

These three thoughts inspired me to open the newsletter to community contributions.

I already shared this idea with our subscribers, and we already have first submissions, which I am very excited about!

Community contributions are a great way to share, learn, teach, and create a lasting impact on our community!

You can also get feedback on your writing, and add the contributions to the open-source section of your CV ❤️

What's next?

If you're interested in causality (no matter your level of expertise) and would like to contribute to Causal Python Weekly, apply here:
👉🏼 https://causal-python-io.formberry.io/community-contrib

To learn more about community contributions and next steps read this:
👉🏼 https://preview.mailerlite.io/emails/webview/516251/141446504455865866 (scroll down to the Community Contrib section)

To test the newsletter, sign up here (you can always unsubscribe with 1 click):
👉🏼 https://causalpython.io/

I am looking forward to seeing your application on the list! 🚀

Hey folks!

Time for the first of our five announcements this week!

In January, we'll kick off the new season of the Causal Bandits Podcast.

Our first two guests have something in common: they are both researchers who actively apply causality in business.

Prof. Stefan Feuerriegel leads the Institute of Artificial Intelligence (AI) in Management at LMU Munich.

His team consistently showcases their work at top AI conferences (NeurIPS, ICML, ICLR, ...), while also helping businesses make better decisions using causal methods.

Reducing semiconductor manufacturing yield loss by 49% is just one of their success stories.

Dr. Ciarán Gilligan-Lee is the Head of the Causal Inference Research Lab at Spotify and an Honorary Associate Professor at University College London. A quantum physicist by training, his causal journey began during his studies.

Today, he leads causal research at Spotify, focusing on applications and uncertainty quantification.

Recently, he helped one of his students use causal methods to tackle problems in astrophysics.

Can causal models help us infer something about the origins of our universe?

Stay tuned!

I hope you'll find these episodes interesting and maybe even inspiring. 🙏🏼

🔥 Don't miss an episode—leave your email here: https://vist.ly/3mrrbgk

Hey folks!

No meme today, but a quick update instead (straight from my home gym 😅).

Next week, I’ll share 5 announcements about 5 initiatives aimed at strengthening, inspiring, and supporting our causal/causal AI community!

We’re starting on Monday!

Wishing you an awesome first weekend of 2025—see you next week!

In 2025, burn the pyramid.

Climb the ladder instead.

Pyramids of evidence are a popular device for conveying the idea that different types of studies provide different strengths of evidence for causal claims.

They can be very helpful in a high-level conversation.

But when we take a deeper look at what causal questions really are, it becomes painfully clear that pyramids of evidence can do more harm than good.

A pyramid of evidence does not offer any insight into what type of causal questions, with what precision and under what circumstances can be answered using which methods.

The consequence of this is that many practitioners get confused regarding what conclusions can be drawn from which types of studies*.

Pyramids also occlude from us the results produced by the last 3 decades of research in causality.

Where should we put evidence obtained by combining interventional and observational data on a pyramid?

How to explain to a student or a practitioner a difference between counterfactual and interventional questions?

How to talk about bounds over counterfactuals?

How to define a situation where we indeed have enough information (even if incomplete) to choose an optimal course of action based on observational data and expert knowledge?

All of these questions are answerable with clarity when we look at causality as a framework.

So what to do if you want to learn more?

In my personal view Pearl's framework, that brought us the notion of "the Ladder of Causation", offers the most clarity in this regard**

A great first step would be to read "The Book of Why" (https://vist.ly/3mrc6j7)

You can also consider:
🔥 Joining the waitlist for a free mini-course on causality: https://vist.ly/3mrc6j5
🔥 Staying up to date with the latest in causal inference, ML & discovery: https://vist.ly/3mrc6j6

*Elizabeth Tipton has a great presentation with real-world examples of how the results of randomized trials can be misapplied in practice

**Potential outcomes and SWIG frameworks offer us similar benefits (in fact using a mixture of Pearlian and potential outcomes concepts is often very convenient in practice).