Learning abstract causes from text

Consider a question that a policymaker might have: which economic factors have causal effects on the median housing price in a region? Answering this question requires gathering historical observations of house prices and economic factors of interest and performing statistical analysis to asses causal effects. But what if the policymaker does not exactly know which economic factors are relevant? What if they cannot afford to measure some of them? The goal of this project is to develop new methodological tools to facilitate causal analyses when experts cannot measure all the potential causes of a phenomenon but have access to relevant sources of knowledge in textual form, e.g., news articles, forums, social media, etc. Building on recent work on causal feature learning, we aim to develop methodology to extract implicit information on the state of potentially relevant causal variables, which we term “abstract causes”, and use these to enable downstream causal analysis.

Faculty Supervisor:

Pascal Germain

Student:

Partner:

ServiceNow Canada

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

Université de Montréal; Université Laval

Program: