
Causal Reasoning and Decision Making
04th November, 2024
Tsinghua University, Beijing, China
- Time: 9:00 AM-12:30 PM, November 4th, 2024
- Venue: Room 329, Mengminwei Humanities Building (蒙民伟人文楼), Tsinghua University
Programme
Time |
Title |
Speaker |
Chair |
9:00-9:50 |
Capturing Cognitive Costs |
Thomas Icard (Stanford University) |
Fenrong Liu (Tsinghua University) |
9:50-10:40 |
Interpreting Causal Conditionals and Evidential Conditionals Based on Causal Plausibility Models |
Kaibo Xie (Wuhan University) |
|
10:40-10:50 |
Coffee Break |
||
10:50-11:40 |
On the Physics of Nested Markov Models: A Generalized Probabilistic Theory Perspective |
Yuhao Wang (Tsinghua University) |
Jialiang Yan (Tsinghua University) |
11:40-12:30 |
Towards Out-of-Distribution Generalization: Causality, Heterogeneity and Evaluation |
Peng Cui (Tsinghua University) |
Abstract

Numerous approaches to decision making with limited resources – going back many decades in disciplines like economics, computer science, and psychology – posit cost functions to capture resource usage and dependence. The goal of this talk will be to clarify what cognitive costs could be, such that they are commensurable with utility (or more broadly task performance) and can play the right theoretical in an account of rational decision making. Information-theoretic and automata-theoretic cost functions will be used as illustrations.

Causal reasoning and doxastic reasoning have a tight connection with conditionals. Although the causal modelling account is very successful in explaining many properties of conditionals, there is still a group of examples in the literature for which no pure causal modelling approach provides the desired results. Günther (2022) proposes that these examples should be analyzed based on the distinction between causal conditionals and evidential conditionals. In this talk, I will introduce a causal plausibility model which can be used to provide a logical account of the two kinds of conditionals.

The nested Markov model provides a simple description of the statistics of observed variables in a Bayesian network, without the need of characterizing the latent variable state space. However, its physical interpretation remains vague. In this work, we explore the nested Markov model in generalized probabilistic theories.