Uploaded December 2020 | Updated September 2026, 2 weeks ago
This IMA event explored the mathematical challenges and opportunities that the COVID-19 epidemic has highlighted, that will go on to be important into the future. Ricardo Silva (University College London) was one of the invited speakers at this event and presented a talk on 'Some Thoughts of Computationally Intensive Causal Inference.' His full abstract is below:
One of the primary goals of causal inference is to predict what happens under interventions, from vaccines to policy-making and scientific claims in general. In many cases, there is no available randomised controlled trial, and assumptions about the causal structure of the world will be necessary if observational data is to be used. There are many families of assumptions that one could be used to infer causation from association in observational studies. The machinery required for deriving the consequences of such assumptions may need to go beyond standard mathematical analysis and rely heavily on algorithms and simulation methods. In this talk, I will focus on a case known as the instrumental variable scenario, where an imperfect experiment informs the causal relation between two variables that are confounded by unobserved factors and the best we can do is to bound the unknown causal effect. We will describe a class of algorithms to compute such bounds while allowing the practitioner a flexible language to express assumptions about the source of such confounding.
Joint work with Niki Kilbertus and Matt Kusner
This IMA event explored the mathematical challenges and opportunities that the COVID-19 epidemic has highlighted, that will go on to be important into the future. Ricardo Silva (University College London) was one of the invited speakers at this event and presented a talk on 'Some Thoughts of Computationally Intensive Causal Inference.' His full abstract is below:
One of the primary goals of causal inference is to predict what happens under interventions, from vaccines to policy-making and scientific claims in general. In many cases, there is no available randomised controlled trial, and assumptions about the causal structure of the world will be necessary if observational data is to be used. There are many families of assumptions that one could be used to infer causation from association in observational studies. The machinery required for deriving the consequences of such assumptions may need to go beyond standard mathematical analysis and rely heavily on algorithms and simulation methods. In this talk, I will focus on a case known as the instrumental variable scenario, where an imperfect experiment informs the causal relation between two variables that are confounded by unobserved factors and the best we can do is to bound the unknown causal effect. We will describe a class of algorithms to compute such bounds while allowing the practitioner a flexible language to express assumptions about the source of such confounding.
Joint work with Niki Kilbertus and Matt Kusner
