Causal Falling Rule Lists

By Fulton Wang, Cynthia Rudin. 2015

21 citations recorded by the source corpus.

A causal falling rule list (CFRL) is a sequence of if-then rules that specifies heterogeneous treatment effects, where (i) the order of rules determines the treatment effect subgroup a subject belongs to, and (ii) the treatment effect decreases monotonically down the list. A given CFRL parameterizes a hierarchical bayesian regression model in which the treatment effects are incorporated as parameters, and assumed constant within model-specific subgroups. We formulate the search for the CFRL best supported by the data as a Bayesian model selection problem, where we perform a search over the space of CFRL models, and approximate the evidence for a given CFRL model using standard variational techniques. We apply CFRL to a census wage dataset to identify subgroups of differing wage inequalities between men and women.

Subfields: Artificial Intelligence, Statistics and Probability, Causal Inference & Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Cynthia Rudin
2. Been Kim
3. Kush R. Varshney
4. Mohit Bansal
5. Adrian Weller
6. Justin Gilmer
7. Nicolas Papernot
8. Margo I. Seltzer

In Statistics and Probability

1. Kush R. Varshney
2. Been Kim
3. Margo I. Seltzer
4. Nicholas Larus-Stone
5. Daniel Alabi
6. Elaine Angelino
7. Julius Adebayo
8. Ian J. Goodfellow

In Causal Inference & Discovery

1. Kush R. Varshney
2. Julius Adebayo
3. Moritz Hardt
4. Been Kim
5. Cynthia Rudin
6. Alexander Volfovsky
7. Atoosa Kasirzadeh

Citations in the corpus

• Sanity Checks for Saliency Maps (2018, background): “Explanations could potentially help satisfy regulatory requirements [3], help practitioners debug their model [4, 5], and perhaps, reveal bias or other unintended effects learned by a model [6, 7].”
• On the Safety of Machine Learning: Cyber-Physical Systems, Decision Sciences, and Data Products (2016, background): “[31], and causality may be used to induce interpretability, e.”
• Learning Certifiably Optimal Rule Lists (2017, background): “For example, Falling Rule Lists (Wang and Rudin, 2015) are constrained to have decreasing probabilities down the list as are rule lists for dynamic treat-ment regimes (Zhang et al., 2015) and cost-sensitive dynamic treatment regimes (Lakkaraju and Rudin, 2017).”
• Scalable Bayesian Rule Lists (2016, neutral): “A Causal Falling Rule List (CFRL) (Wang & Rudin, 2015a) is another such example.”
• Engineering safety in machine learning (2016, background): “For a rare combination of features in a test sample [20], a safe fail mechanism is to always go for manual examination.”