Search
Buy your pass
OnlineDiscovery - From Molecules to Models Badge available

Learn how to do from what you see: Causal effect estimation

  • Register

    Watch

     

    Watch and get your discovery badge

    * Register (or log in) to the Neural Network to add this session to your agenda or watch the replay

    • Date
    29 October 2025
    Timeframe
    17:00 - 18:00 CEST Geneva
    Duration
    60 minutes

    During this talk, I will describe methods to find the real effect of an action, whether it is a new medical drug, a ticket price change, or a specific web design, on a final result such as patient health or sales. This task, called causal modeling, is challenging. First, both the actions and the context are often highly complex. Second, we cannot always run new experiments and are instead limited to studying historical data. The biggest hurdle is hidden confounding: unseen factors that influence both the action taken and the final result, making it appear as though the action caused the result when it did not.

    I will introduce new, practical machine learning methods to address these challenges. These tools use AI, such as kernel methods or deep neural networks, to better understand the effect of actions on outcomes. Crucially, if hidden factors are confounding our observations, specialized techniques such as instrumental variable regression can be used to mathematically correct for the confounding. I will demonstrate these methods by analyzing the success of the US Job Corps program for disadvantaged youth and, in reinforcement learning, in predicting the reward of different action policies.

     

    Learning Objectives
    By the end of this session, participants will be able to:

    • Explain the challenge of hidden confounding in causal modeling and why it complicates the assessment of an action’s true effect on an outcome.
    • Recognize the (conditional) average treatment effect and instrumental variable regression as specialized techniques used to mathematically correct for confounding when analyzing historical (observational) data.
    • Describe how modern machine learning models, such as deep neural networks, can be applied to model the complex relationships between actions, context, and outcomes in causal inference tasks.
    • Analyze the application of causal modeling to evaluate the effectiveness of a social program, like the US Job Corps, and in determining the value of a policy in reinforcement learning.

     

    Recommended Mastery Level / Prerequisites:

    Intermediate to Advanced – suitable for graduate students, early career researchers, and professionals in climate science, oceanography, or environmental data science.

    • Discovery badge available
    Showcase it on your LinkedIn profile or your professional portfolio
    Verified badge is available for CHF 10
    Learn more
    Share this session with your network
    Discover more from programme stream
    Discovery - From Molecules to Models
    Generative AI and magical thinking

    Generative AI and magical thinking

    12 January 2026 - 16:00 to 17:00

    Related stories

    Are you sure you want to remove this speaker?