PhD Position Causal Inference & Machine Learning

Challenge: Causal inference relies on strong assumptions

Change: Methods to detect and mitigate assumption violations

Impact: Trustworthy causal claims

This PhD position focuses on causal inference and machine learning, with the goal of developing methods to detect and mitigate violations of causal assumptions.

Job description

Progress in statistics and machine learning research has led to great strides in using data to uncover relationships between things in the world we observe around us. Many questions of scientific and societal (e.g. medical) importance, however, are causal questions. Causal questions ask what the effects of new actions will be, rather than how things are related in a world in which we keep things as is.

Given the importance of answering such causal questions, many data-driven methods have been developed to estimate causal effects in machine learning as well as statistics, epidemiology, econometrics, psychometrics, and many others. However, these methods all rely on causal assumptions, such as positivity and unconfoundedness. When such assumptions are violated, these methods will return incorrect causal answers, which leads to flawed and potentially even harmful decision-making. Currently, however, there is a lack of data-driven methods to detect assumption violations, leading to problems that go undetected, and ultimately, untrustworthy causal claims.

In this project, our goal is to better understand the limits of the detection of causal assumption violations, develop new detection methods, incorporate these violations into causal models, and design procedures to mitigate their influence. By offering these tools to practitioners, we will contribute to a new, safer approach to addressing causal questions that leads to more trustworthy causal answers.

The project particularly focusses on positivity violations, that is the assumption that it is possible for each treatment of interest to have occurred for each relevant unit in the data. We are interested in developing methods to detect whether this assumption is violated, to find out for which treatments or subpopulations it holds, and to properly account for the uncertainty caused by limited support in the data. We want to investigate this in both simple and more complicated (e.g. high-dimensional or longitudinal) settings.

The project is part of the "Safe Causal Inference" consortium, a multi-disciplinary (computer science, mathematics, biostatistics, epidemiology) consortium spanning eight PhD positions to improve the trustworthiness of causal methods. Through the consortium, you will be able to collaborate with experts and peers from multiple causal inference disciplines to strengthen and inspire your work.

We are looking for an enthusiastic and self-motivated person with a passion for research. We offer a stimulating scientific research environment and the embedding in a multi-disciplinary consortium that allows you to develop skills that go beyond discipline-specific boundaries. You will be supervised by Jesse Krijthe (computer science, TU Delft) and Nan van Geloven (biostatistics, Leiden University Medical Center) and be primarily based in the Pattern Recognition and Bioinformatics group of TU Delft, which includes researchers working on the methodology of machine learning, bioinformatics, computer vision, and socially perceptive computing.

Job requirements

• A critical thinker with an open mind
• Strong interest in understanding and developing statistical and machine learning methods
• Master’s degree in a relevant quantitative field (such as statistics, computer science, mathematics, econometrics, psychometrics, physics, etc.)
• Ability to program in Python, R, Julia or a similar (scientific) programming language
• Proficient in spoken and written English

TU Delft (Delft University of Technology)

At TU Delft, you will work in an international research environment focused on science, engineering and design, addressing challenges in areas such as energy, climate, mobility, health and digital society. TU Delft offers a stimulating research environment within the Faculty of Electrical Engineering, Mathematics and Computer Science and the Pattern Recognition and Bioinformatics group, with opportunities for ground-breaking research and collaboration across disciplines.

Faculty of Electrical Engineering, Mathematics and Computer Science

The Faculty of Electrical Engineering, Mathematics and Computer Science (EEMCS) brings together three scientific disciplines that reinforce each other and contribute to research and technology development across areas including sustainable electricity systems, chips and sensors, AI, and applied mathematics. The faculty offers an innovative research environment with excellent labs and facilities.

Conditions of employment

Doctoral candidates will be offered a 4-year period of employment in principle, but in the form of 2 employment contracts. An initial 1,5 year contract with an official go/no go progress assessment within 15 months, followed by an additional contract for the remaining 2,5 years assuming everything goes well and performance requirements are met.

As a PhD candidate you will be enrolled in the TU Delft Graduate School. The TU Delft Graduate School provides an inspiring research environment with an excellent team of supervisors, academic staff and a mentor. The Doctoral Education Programme is aimed at developing your transferable, discipline-related and research skills.

TU Delft offers a customisable compensation package, discounts on health insurance, and a monthly work costs contribution. Flexible work schedules can be arranged.