Publications Search
Sam Witty, David Jensen, Vikash Mansinghka
A Simulation-Based Test of Identifiability for Bayesian Causal Inference Miscellaneous
2021.
Abstract | Links | BibTeX | Tags: Causal Modeling
@misc{DBLP:journals/corr/abs-2102-11761,
title = {A Simulation-Based Test of Identifiability for Bayesian Causal Inference},
author = {Sam Witty and David Jensen and Vikash Mansinghka},
url = {https://arxiv.org/abs/2102.11761},
year = {2021},
date = {2021-01-01},
journal = {CoRR},
volume = {abs/2102.11761},
abstract = {This paper introduces a procedure for testing the identifiability of Bayesian models for causal inference. Although the do-calculus is sound and complete given a causal graph, many practical assumptions cannot be expressed in terms of graph structure alone, such as the assumptions required by instrumental variable designs, regression discontinuity designs, and within-subjects designs. We present simulation-based identifiability (SBI), a fully automated identification test based on a particle optimization scheme with simulated observations. This approach expresses causal assumptions as priors over functions in a structural causal model, including flexible priors using Gaussian processes. We prove that SBI is asymptotically sound and complete, and produces practical finite-sample bounds. We also show empirically that SBI agrees with known results in graph-based identification as well as with widely-held intuitions for designs in which graph-based methods are inconclusive.},
keywords = {Causal Modeling},
pubstate = {published},
tppubtype = {misc}
}
This paper introduces a procedure for testing the identifiability of Bayesian models for causal inference. Although the do-calculus is sound and complete given a causal graph, many practical assumptions cannot be expressed in terms of graph structure alone, such as the assumptions required by instrumental variable designs, regression discontinuity designs, and within-subjects designs. We present simulation-based identifiability (SBI), a fully automated identification test based on a particle optimization scheme with simulated observations. This approach expresses causal assumptions as priors over functions in a structural causal model, including flexible priors using Gaussian processes. We prove that SBI is asymptotically sound and complete, and produces practical finite-sample bounds. We also show empirically that SBI agrees with known results in graph-based identification as well as with widely-held intuitions for designs in which graph-based methods are inconclusive.
Terrance E. Boult, Przemyslaw A. Grabowicz, D. S. Prijatelj, R. Stern, L. Holder, J. Alspector, M. Jafarzadeh, T. Ahmad, A. R. Dhamija, C. Li, S. Cruz, A. Shrivastava, C. Vondrick, W. J. Scheirer
A Unifying Framework for Formal Theories of Novelty:Framework, Examples and Discussion Proceedings Article
In: AAAI'21 SMPT, 2021, ISSN: 23318422.
Abstract | Links | BibTeX | Tags: Novelty
@inproceedings{Boult2020a,
title = {A Unifying Framework for Formal Theories of Novelty:Framework, Examples and Discussion},
author = {Terrance E. Boult and Przemyslaw A. Grabowicz and D. S. Prijatelj and R. Stern and L. Holder and J. Alspector and M. Jafarzadeh and T. Ahmad and A. R. Dhamija and C. Li and S. Cruz and A. Shrivastava and C. Vondrick and W. J. Scheirer},
url = {http://arxiv.org/abs/2012.04226},
issn = {23318422},
year = {2021},
date = {2021-12-01},
booktitle = {AAAI'21 SMPT},
abstract = {Managing inputs that are novel, unknown, or out-of-distribution is critical as an agent moves from the lab to the open world. Novelty-related problems include being tolerant to novel perturbations of the normal input, detecting when the input includes novel items, and adapting to novel inputs. While significant research has been undertaken in these areas, a noticeable gap exists in the lack of a formalized definition of novelty that transcends problem domains. As a team of researchers spanning multiple research groups and different domains, we have seen, first hand, the difficulties that arise from ill-specified novelty problems, as well as inconsistent definitions and terminology. Therefore, we present the first unified framework for formal theories of novelty and use the framework to formally define a family of novelty types. Our framework can be applied across a wide range of domains, from symbolic AI to reinforcement learning, and beyond to open world image recognition. Thus, it can be used to help kick-start new research efforts and accelerate ongoing work on these important novelty-related problems. This extended version of our AAAI 2021 paper included more details and examples in multiple domains.},
keywords = {Novelty},
pubstate = {published},
tppubtype = {inproceedings}
}
Managing inputs that are novel, unknown, or out-of-distribution is critical as an agent moves from the lab to the open world. Novelty-related problems include being tolerant to novel perturbations of the normal input, detecting when the input includes novel items, and adapting to novel inputs. While significant research has been undertaken in these areas, a noticeable gap exists in the lack of a formalized definition of novelty that transcends problem domains. As a team of researchers spanning multiple research groups and different domains, we have seen, first hand, the difficulties that arise from ill-specified novelty problems, as well as inconsistent definitions and terminology. Therefore, we present the first unified framework for formal theories of novelty and use the framework to formally define a family of novelty types. Our framework can be applied across a wide range of domains, from symbolic AI to reinforcement learning, and beyond to open world image recognition. Thus, it can be used to help kick-start new research efforts and accelerate ongoing work on these important novelty-related problems. This extended version of our AAAI 2021 paper included more details and examples in multiple domains.
Amanda M Gentzel, Purva Pruthi, David Jensen
How and Why to Use Experimental Data to Evaluate Methods for Observational Causal Inference Proceedings Article
In: International Conference on Machine Learning, pp. 3660–3671, PMLR 2021.
Abstract | Links | BibTeX | Tags: Causal Modeling
@inproceedings{gentzel2021and,
title = {How and Why to Use Experimental Data to Evaluate Methods for Observational Causal Inference},
author = {Amanda M Gentzel and Purva Pruthi and David Jensen},
url = {http://proceedings.mlr.press/v139/gentzel21a/gentzel21a.pdf},
year = {2021},
date = {2021-01-01},
booktitle = {International Conference on Machine Learning},
pages = {3660--3671},
organization = {PMLR},
abstract = {Methods that infer causal dependence from observational data are central to many areas of science, including medicine, economics, and the social sciences. A variety of theoretical properties of these methods have been proven, but empirical evaluation remains a challenge, largely due to the lack of observational data sets for which treatment effect is known. We describe and analyze observational sampling from randomized controlled trials (OSRCT), a method for evaluating causal inference methods using data from randomized controlled trials (RCTs). This method can be used to create constructed observational data sets with corresponding unbiased estimates of treatment effect, substantially increasing the number of data sets available for evaluating causal inference methods. We show that, in expectation, OSRCT creates data sets that are equivalent to those produced by randomly sampling from empirical data sets in which all potential outcomes are available. We then perform a large-scale evaluation of seven causal inference methods over 37 data sets, drawn from RCTs, as well as simulators, real-world computational systems, and observational data sets augmented with a synthetic response variable. We find notable performance differences when comparing across data from different sources, demonstrating the importance of using data from a variety of sources when evaluating any causal inference method.},
keywords = {Causal Modeling},
pubstate = {published},
tppubtype = {inproceedings}
}
Methods that infer causal dependence from observational data are central to many areas of science, including medicine, economics, and the social sciences. A variety of theoretical properties of these methods have been proven, but empirical evaluation remains a challenge, largely due to the lack of observational data sets for which treatment effect is known. We describe and analyze observational sampling from randomized controlled trials (OSRCT), a method for evaluating causal inference methods using data from randomized controlled trials (RCTs). This method can be used to create constructed observational data sets with corresponding unbiased estimates of treatment effect, substantially increasing the number of data sets available for evaluating causal inference methods. We show that, in expectation, OSRCT creates data sets that are equivalent to those produced by randomly sampling from empirical data sets in which all potential outcomes are available. We then perform a large-scale evaluation of seven causal inference methods over 37 data sets, drawn from RCTs, as well as simulators, real-world computational systems, and observational data sets augmented with a synthetic response variable. We find notable performance differences when comparing across data from different sources, demonstrating the importance of using data from a variety of sources when evaluating any causal inference method.
David Jensen
Improving Causal Inference by Increasing Model Expressiveness Proceedings Article
In: Proceedings of the AAAI Conference on Artificial Intelligence, pp. 15053–15057, 2021.
Abstract | Links | BibTeX | Tags:
@inproceedings{jensen2021improving,
title = {Improving Causal Inference by Increasing Model Expressiveness},
author = {David Jensen},
url = {https://www.aaai.org/AAAI21Papers/SMT-427.JensenD.pdf},
year = {2021},
date = {2021-01-01},
booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
volume = {35},
number = {17},
pages = {15053--15057},
abstract = {The ability to learn and reason with causal knowledge is a key aspect of intelligent behavior. In contrast to mere statistical association, knowledge of causation enables reasoning about the effects of actions. Causal reasoning is vital for autonomous agents and for a range of applications in science, medicine, business, and government. However, current methods for causal inference are hobbled because they use relatively inexpressive models. Surprisingly, current causal models eschew nearly every major representational innovation common in a range of other fields both inside and outside of computer science, including representation of objects, relationships, time, space, and hierarchy. Even more surprisingly, a range of recent research provides strong evidence that more expressive representations make possible causal inferences that are otherwise impossible and remove key biases that would otherwise afflict more naive inferences. New research on causal inference should target increases in expressiveness to improve accuracy and effectiveness.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The ability to learn and reason with causal knowledge is a key aspect of intelligent behavior. In contrast to mere statistical association, knowledge of causation enables reasoning about the effects of actions. Causal reasoning is vital for autonomous agents and for a range of applications in science, medicine, business, and government. However, current methods for causal inference are hobbled because they use relatively inexpressive models. Surprisingly, current causal models eschew nearly every major representational innovation common in a range of other fields both inside and outside of computer science, including representation of objects, relationships, time, space, and hierarchy. Even more surprisingly, a range of recent research provides strong evidence that more expressive representations make possible causal inferences that are otherwise impossible and remove key biases that would otherwise afflict more naive inferences. New research on causal inference should target increases in expressiveness to improve accuracy and effectiveness.
Akanksha Atrey, Prashant J. Shenoy, David Jensen
Preserving Privacy in Personalized Models for Distributed Mobile Services Miscellaneous
2021.
Abstract | Links | BibTeX | Tags:
@misc{DBLP:journals/corr/abs-2101-05855,
title = {Preserving Privacy in Personalized Models for Distributed Mobile Services},
author = {Akanksha Atrey and Prashant J. Shenoy and David Jensen},
url = {https://arxiv.org/abs/2101.05855},
year = {2021},
date = {2021-01-01},
journal = {CoRR},
volume = {abs/2101.05855},
abstract = {The ubiquity of mobile devices has led to the proliferation of mobile services that provide personalized and context-aware content to their users. Modern mobile services are distributed between end-devices, such as smartphones, and remote servers that reside in the cloud. Such services thrive on their ability to predict future contexts to pre-fetch content or make context-specific recommendations. An increasingly common method to predict future contexts, such as location, is via machine learning (ML) models. Recent work in context prediction has focused on ML model personalization where a personalized model is learned for each individual user in order to tailor predictions or recommendations to a user's mobile behavior. While the use of personalized models increases efficacy of the mobile service, we argue that it increases privacy risk since a personalized model encodes contextual behavior unique to each user. To demonstrate these privacy risks, we present several attribute inference-based privacy attacks and show that such attacks can leak privacy with up to 78% efficacy for top-3 predictions. We present Pelican, a privacy-preserving personalization system for context-aware mobile services that leverages both device and cloud resources to personalize ML models while minimizing the risk of privacy leakage for users. We evaluate Pelican using real world traces for location-aware mobile services and show that Pelican can substantially reduce privacy leakage by up to 75%.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
The ubiquity of mobile devices has led to the proliferation of mobile services that provide personalized and context-aware content to their users. Modern mobile services are distributed between end-devices, such as smartphones, and remote servers that reside in the cloud. Such services thrive on their ability to predict future contexts to pre-fetch content or make context-specific recommendations. An increasingly common method to predict future contexts, such as location, is via machine learning (ML) models. Recent work in context prediction has focused on ML model personalization where a personalized model is learned for each individual user in order to tailor predictions or recommendations to a user's mobile behavior. While the use of personalized models increases efficacy of the mobile service, we argue that it increases privacy risk since a personalized model encodes contextual behavior unique to each user. To demonstrate these privacy risks, we present several attribute inference-based privacy attacks and show that such attacks can leak privacy with up to 78% efficacy for top-3 predictions. We present Pelican, a privacy-preserving personalization system for context-aware mobile services that leverages both device and cloud resources to personalize ML models while minimizing the risk of privacy leakage for users. We evaluate Pelican using real world traces for location-aware mobile services and show that Pelican can substantially reduce privacy leakage by up to 75%.
Aarshee Mishra, Przemyslaw A. Grabowicz, Nicholas Perello
Towards Fair and Explainable Supervised Learning Proceedings Article
In: ICML Workshop on Socially Responsible Machine Learning, 2021.
Abstract | Links | BibTeX | Tags: Fairness
@inproceedings{Mishra2021,
title = {Towards Fair and Explainable Supervised Learning},
author = {Aarshee Mishra and Przemyslaw A. Grabowicz and Nicholas Perello},
url = {https://drive.google.com/file/d/1z24hITF0Xrlc_IX_rOZVZ2aigOj1hxhD/view?usp=sharing},
year = {2021},
date = {2021-01-01},
booktitle = {ICML Workshop on Socially Responsible Machine Learning},
abstract = {Algorithms that aid human decision-making may inadvertently discriminate against certain protected groups. We formalize direct discrimination as a direct causal effect of the protected attributes on the decisions, while textitinduced indirect discrimination as a change in the influence of non-protected features associated with the protected attributes. The measurements of average treatment effect (ATE) and SHapley Additive exPlanations (SHAP) reveal that state-of-the-art fair learning methods can inadvertently induce indirect discrimination in synthetic and real-world datasets. To inhibit discrimination in algorithmic systems, we propose to nullify the influence of the protected attribute on the output of the system, while preserving the influence of remaining features. To achieve this objective, we introduce a risk minimization method which optimizes for the proposed fairness objective. We show that the method leverages model accuracy and disparity measures.},
keywords = {Fairness},
pubstate = {published},
tppubtype = {inproceedings}
}
Algorithms that aid human decision-making may inadvertently discriminate against certain protected groups. We formalize direct discrimination as a direct causal effect of the protected attributes on the decisions, while textitinduced indirect discrimination as a change in the influence of non-protected features associated with the protected attributes. The measurements of average treatment effect (ATE) and SHapley Additive exPlanations (SHAP) reveal that state-of-the-art fair learning methods can inadvertently induce indirect discrimination in synthetic and real-world datasets. To inhibit discrimination in algorithmic systems, we propose to nullify the influence of the protected attribute on the output of the system, while preserving the influence of remaining features. To achieve this objective, we introduce a risk minimization method which optimizes for the proposed fairness objective. We show that the method leverages model accuracy and disparity measures.
Sam Witty, Kenta Takatsu, David Jensen, Vikash Mansinghka
Causal Inference using Gaussian Processes with Structured Latent Confounders Proceedings Article
In: Proceedings of the 37th International Conference on Machine Learning, ICML 2020, 13-18 July 2020, Virtual Event, pp. 10313–10323, PMLR, 2020.
Abstract | Links | BibTeX | Tags: Causal Modeling
@inproceedings{DBLP:conf/icml/WittyTJM20,
title = {Causal Inference using Gaussian Processes with Structured Latent Confounders},
author = {Sam Witty and Kenta Takatsu and David Jensen and Vikash Mansinghka},
url = {http://proceedings.mlr.press/v119/witty20a.html},
year = {2020},
date = {2020-01-01},
booktitle = {Proceedings of the 37th International Conference on Machine Learning,
ICML 2020, 13-18 July 2020, Virtual Event},
volume = {119},
pages = {10313--10323},
publisher = {PMLR},
series = {Proceedings of Machine Learning Research},
abstract = {Latent confounders---unobserved variables that influence both treatment and outcome---can bias estimates of causal effects. In some cases, these confounders are shared across observations, e.g. all students taking a course are influenced by the course's difficulty in addition to any educational interventions they receive individually. This paper shows how to semiparametrically model latent confounders that have this structure and thereby improve estimates of causal effects. The key innovations are a hierarchical Bayesian model, Gaussian processes with structured latent confounders (GP-SLC), and a Monte Carlo inference algorithm for this model based on elliptical slice sampling. GP-SLC provides principled Bayesian uncertainty estimates of individual treatment effect with minimal assumptions about the functional forms relating confounders, covariates, treatment, and outcome. Finally, this paper shows GP-SLC is competitive with or more accurate than widely used causal inference techniques on three benchmark datasets, including the Infant Health and Development Program and a dataset showing the effect of changing temperatures on state-wide energy consumption across New England.},
keywords = {Causal Modeling},
pubstate = {published},
tppubtype = {inproceedings}
}
Latent confounders---unobserved variables that influence both treatment and outcome---can bias estimates of causal effects. In some cases, these confounders are shared across observations, e.g. all students taking a course are influenced by the course's difficulty in addition to any educational interventions they receive individually. This paper shows how to semiparametrically model latent confounders that have this structure and thereby improve estimates of causal effects. The key innovations are a hierarchical Bayesian model, Gaussian processes with structured latent confounders (GP-SLC), and a Monte Carlo inference algorithm for this model based on elliptical slice sampling. GP-SLC provides principled Bayesian uncertainty estimates of individual treatment effect with minimal assumptions about the functional forms relating confounders, covariates, treatment, and outcome. Finally, this paper shows GP-SLC is competitive with or more accurate than widely used causal inference techniques on three benchmark datasets, including the Infant Health and Development Program and a dataset showing the effect of changing temperatures on state-wide energy consumption across New England.
David Ifeoluwa Adelani, Ryota Kobayashi, Ingmar Weber, Przemyslaw A. Grabowicz
Estimating community feedback effect on topic choice in social media with predictive modeling Journal Article
In: EPJ Data Science, vol. 9, no. 1, pp. 25, 2020, ISSN: 2193-1127.
Abstract | Links | BibTeX | Tags: Computational Social Science, Social feedback, Social influence, User behavior modeling
@article{Adelani2020,
title = {Estimating community feedback effect on topic choice in social media with predictive modeling},
author = {David Ifeoluwa Adelani and Ryota Kobayashi and Ingmar Weber and Przemyslaw A. Grabowicz},
url = {http://dx.doi.org/10.1140/epjds/s13688-020-00243-w https://epjdatascience.springeropen.com/articles/10.1140/epjds/s13688-020-00243-w},
doi = {10.1140/epjds/s13688-020-00243-w},
issn = {2193-1127},
year = {2020},
date = {2020-12-01},
journal = {EPJ Data Science},
volume = {9},
number = {1},
pages = {25},
publisher = {The Author(s)},
abstract = {Social media users post content on various topics. A defining feature of social media is that other users can provide feedback—called community feedback—to their content in the form of comments, replies, and retweets. We hypothesize that the amount of received feedback influences the choice of topics on which a social media user posts. However, it is challenging to test this hypothesis as user heterogeneity and external confounders complicate measuring the feedback effect. Here, we investigate this hypothesis with a predictive approach based on an interpretable model of an author's decision to continue the topic of their previous post. We explore the confounding factors, including author's topic preferences and unobserved external factors such as news and social events, by optimizing the predictive accuracy. This approach enables us to identify which users are susceptible to community feedback. Overall, we find that 33% and 14% of active users in Reddit and Twitter, respectively, are influenced by community feedback. The model suggests that this feedback alters the probability of topic continuation up to 14%, depending on the user and the amount of feedback.},
keywords = {Computational Social Science, Social feedback, Social influence, User behavior modeling},
pubstate = {published},
tppubtype = {article}
}
Social media users post content on various topics. A defining feature of social media is that other users can provide feedback—called community feedback—to their content in the form of comments, replies, and retweets. We hypothesize that the amount of received feedback influences the choice of topics on which a social media user posts. However, it is challenging to test this hypothesis as user heterogeneity and external confounders complicate measuring the feedback effect. Here, we investigate this hypothesis with a predictive approach based on an interpretable model of an author's decision to continue the topic of their previous post. We explore the confounding factors, including author's topic preferences and unobserved external factors such as news and social events, by optimizing the predictive accuracy. This approach enables us to identify which users are susceptible to community feedback. Overall, we find that 33% and 14% of active users in Reddit and Twitter, respectively, are influenced by community feedback. The model suggests that this feedback alters the probability of topic continuation up to 14%, depending on the user and the amount of feedback.
Akanksha Atrey, Kaleigh Clary, David Jensen
Exploratory Not Explanatory: Counterfactual Analysis of Saliency Maps for Deep Reinforcement Learning Proceedings Article
In: International Conference on Learning Representations, 2020.
Abstract | Links | BibTeX | Tags: Explainable AI
@inproceedings{atrey2020exploratory,
title = {Exploratory Not Explanatory: Counterfactual Analysis of Saliency Maps for Deep Reinforcement Learning},
author = {Akanksha Atrey and Kaleigh Clary and David Jensen},
url = {https://openreview.net/pdf?id=rkl3m1BFDB},
year = {2020},
date = {2020-01-01},
booktitle = {International Conference on Learning Representations},
abstract = {Saliency maps are frequently used to support explanations of the behavior of deep reinforcement learning (RL) agents. However, a review of how saliency maps are used in practice indicates that the derived explanations are often unfalsifiable and can be highly subjective. We introduce an empirical approach grounded in counterfactual reasoning to test the hypotheses generated from saliency maps and assess the degree to which they correspond to the semantics of RL environments. We use Atari games, a common benchmark for deep RL, to evaluate three types of saliency maps. Our results show the extent to which existing claims about Atari games can be evaluated and suggest that saliency maps are best viewed as an exploratory tool rather than an explanatory tool.},
keywords = {Explainable AI},
pubstate = {published},
tppubtype = {inproceedings}
}
Saliency maps are frequently used to support explanations of the behavior of deep reinforcement learning (RL) agents. However, a review of how saliency maps are used in practice indicates that the derived explanations are often unfalsifiable and can be highly subjective. We introduce an empirical approach grounded in counterfactual reasoning to test the hypotheses generated from saliency maps and assess the degree to which they correspond to the semantics of RL environments. We use Atari games, a common benchmark for deep RL, to evaluate three types of saliency maps. Our results show the extent to which existing claims about Atari games can be evaluated and suggest that saliency maps are best viewed as an exploratory tool rather than an explanatory tool.
David Jensen, Javier Burroni, Matthew Rattigan
Object conditioning for causal inference Proceedings Article
In: Uncertainty in Artificial Intelligence, pp. 1072–1082, PMLR 2020.
Abstract | Links | BibTeX | Tags:
@inproceedings{jensen2020object,
title = {Object conditioning for causal inference},
author = {David Jensen and Javier Burroni and Matthew Rattigan},
url = {http://proceedings.mlr.press/v115/jensen20a/jensen20a.pdf},
year = {2020},
date = {2020-01-01},
booktitle = {Uncertainty in Artificial Intelligence},
pages = {1072--1082},
organization = {PMLR},
abstract = {We describe and analyze a form of conditioning that is widely applied within social science and applied statistics but that is virtually unknown within causal graphical models. This approach, which we term object conditioning, can adjust for the effects of latent confounders and yet avoid the pitfall of conditioning on colliders. We describe object conditioning using plate models and show how its probabilistic implications can be explained using the property of exchangeability. We show that several seemingly obvious interpretations of object conditioning are insufficient to describe its probabilistic implications. Finally, we use object conditioning to describe and unify key aspects of a diverse set of techniques for causal inference, including within-subjects designs, difference-in-differences designs, and interrupted time-series designs.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We describe and analyze a form of conditioning that is widely applied within social science and applied statistics but that is virtually unknown within causal graphical models. This approach, which we term object conditioning, can adjust for the effects of latent confounders and yet avoid the pitfall of conditioning on colliders. We describe object conditioning using plate models and show how its probabilistic implications can be explained using the property of exchangeability. We show that several seemingly obvious interpretations of object conditioning are insufficient to describe its probabilistic implications. Finally, we use object conditioning to describe and unify key aspects of a diverse set of techniques for causal inference, including within-subjects designs, difference-in-differences designs, and interrupted time-series designs.
Katherine A. Keith, David Jensen, Brendan O'Connor
Text and Causal Inference: A Review of Using Text to Remove Confounding from Causal Estimates Proceedings Article
In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020, Online, July 5-10, 2020, pp. 5332–5344, Association for Computational Linguistics, 2020.
Abstract | Links | BibTeX | Tags:
@inproceedings{DBLP:conf/acl/KeithJO20,
title = {Text and Causal Inference: A Review of Using Text to Remove Confounding
from Causal Estimates},
author = {Katherine A. Keith and David Jensen and Brendan O'Connor},
url = {https://doi.org/10.18653/v1/2020.acl-main.474},
year = {2020},
date = {2020-01-01},
booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational
Linguistics, ACL 2020, Online, July 5-10, 2020},
pages = {5332--5344},
publisher = {Association for Computational Linguistics},
abstract = {Many applications of computational social science aim to infer causal conclusions from non-experimental data. Such observational data often contains confounders, variables that influence both potential causes and potential effects. Unmeasured or latent confounders can bias causal estimates, and this has motivated interest in measuring potential confounders from observed text. For example, an individual's entire history of social media posts or the content of a news article could provide a rich measurement of multiple confounders. Yet, methods and applications for this problem are scattered across different communities and evaluation practices are inconsistent. This review is the first to gather and categorize these examples and provide a guide to data-processing and evaluation decisions. Despite increased attention on adjusting for confounding using text, there are still many open problems, which we highlight in this paper.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Many applications of computational social science aim to infer causal conclusions from non-experimental data. Such observational data often contains confounders, variables that influence both potential causes and potential effects. Unmeasured or latent confounders can bias causal estimates, and this has motivated interest in measuring potential confounders from observed text. For example, an individual's entire history of social media posts or the content of a news article could provide a rich measurement of multiple confounders. Yet, methods and applications for this problem are scattered across different communities and evaluation practices are inconsistent. This review is the first to gather and categorize these examples and provide a guide to data-processing and evaluation decisions. Despite increased attention on adjusting for confounding using text, there are still many open problems, which we highlight in this paper.
Amanda Gentzel, Justin Clarke, David Jensen
Using Experimental Data to Evaluate Methods for Observational Causal Inference Miscellaneous
2020.
Abstract | Links | BibTeX | Tags:
@misc{DBLP:journals/corr/abs-2010-03051,
title = {Using Experimental Data to Evaluate Methods for Observational Causal
Inference},
author = {Amanda Gentzel and Justin Clarke and David Jensen},
url = {https://arxiv.org/abs/2010.03051},
year = {2020},
date = {2020-01-01},
journal = {CoRR},
volume = {abs/2010.03051},
abstract = {Methods that infer causal dependence from observational data are central to many areas of science, including medicine, economics, and the social sciences. A variety of theoretical properties of these methods have been proven, but empirical evaluation remains a challenge, largely due to the lack of observational data sets for which treatment effect is known. We propose and analyze observational sampling from randomized controlled trials (OSRCT), a method for evaluating causal inference methods using data from randomized controlled trials (RCTs). This method can be used to create constructed observational data sets with corresponding unbiased estimates of treatment effect, substantially increasing the number of data sets available for evaluating causal inference methods. We show that, in expectation, OSRCT creates data sets that are equivalent to those produced by randomly sampling from empirical data sets in which all potential outcomes are available. We analyze several properties of OSRCT theoretically and empirically, and we demonstrate its use by comparing the performance of four causal inference methods using data from eleven RCTs.},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Methods that infer causal dependence from observational data are central to many areas of science, including medicine, economics, and the social sciences. A variety of theoretical properties of these methods have been proven, but empirical evaluation remains a challenge, largely due to the lack of observational data sets for which treatment effect is known. We propose and analyze observational sampling from randomized controlled trials (OSRCT), a method for evaluating causal inference methods using data from randomized controlled trials (RCTs). This method can be used to create constructed observational data sets with corresponding unbiased estimates of treatment effect, substantially increasing the number of data sets available for evaluating causal inference methods. We show that, in expectation, OSRCT creates data sets that are equivalent to those produced by randomly sampling from empirical data sets in which all potential outcomes are available. We analyze several properties of OSRCT theoretically and empirically, and we demonstrate its use by comparing the performance of four causal inference methods using data from eleven RCTs.
Sam Witty, Alexander Lew, David Jensen, Vikash Mansinghka
Bayesian causal inference via probabilistic program synthesis Miscellaneous
2019.
Abstract | Links | BibTeX | Tags: Probabilistic Programming
@misc{witty2019bayesian,
title = {Bayesian causal inference via probabilistic program synthesis},
author = {Sam Witty and Alexander Lew and David Jensen and Vikash Mansinghka},
url = {https://arxiv.org/pdf/1910.14124.pdf},
year = {2019},
date = {2019-01-01},
journal = {arXiv preprint arXiv:1910.14124},
abstract = {Causal inference can be formalized as Bayesian inference that combines a prior distribution over causal models and likelihoods that account for both observations and interventions. We show that it is possible to implement this approach using a sufficiently expressive probabilistic programming language. Priors are represented using probabilistic programs that generate source code in a domain specific language. Interventions are represented using probabilistic programs that edit this source code to modify the original generative process. This approach makes it straightforward to incorporate data from atomic interventions, as well as shift interventions, variance-scaling interventions, and other interventions that modify causal structure. This approach also enables the use of general-purpose inference machinery for probabilistic programs to infer probable causal structures and parameters from data. This abstract describes a prototype of this approach in the Gen probabilistic programming language.},
keywords = {Probabilistic Programming},
pubstate = {published},
tppubtype = {misc}
}
Causal inference can be formalized as Bayesian inference that combines a prior distribution over causal models and likelihoods that account for both observations and interventions. We show that it is possible to implement this approach using a sufficiently expressive probabilistic programming language. Priors are represented using probabilistic programs that generate source code in a domain specific language. Interventions are represented using probabilistic programs that edit this source code to modify the original generative process. This approach makes it straightforward to incorporate data from atomic interventions, as well as shift interventions, variance-scaling interventions, and other interventions that modify causal structure. This approach also enables the use of general-purpose inference machinery for probabilistic programs to infer probable causal structures and parameters from data. This abstract describes a prototype of this approach in the Gen probabilistic programming language.
David Jensen, others
Comment: Strengthening empirical evaluation of causal inference methods Journal Article
In: Statistical Science, vol. 34, no. 1, pp. 77–81, 2019.
Abstract | Links | BibTeX | Tags:
@article{jensen2019comment,
title = {Comment: Strengthening empirical evaluation of causal inference methods},
author = {David Jensen and others},
url = {https://projecteuclid.org/journals/statistical-science/volume-34/issue-1/Comment-Strengthening-Empirical-Evaluation-of-Causal-Inference-Methods/10.1214/18-STS690.short},
year = {2019},
date = {2019-01-01},
journal = {Statistical Science},
volume = {34},
number = {1},
pages = {77--81},
publisher = {Institute of Mathematical Statistics},
abstract = {This is a contribution to the discussion of the paper by Dorie et al. (Statist. Sci. 34 (2019) 43–68), which reports the lessons learned from 2016 Atlantic Causal Inference Conference Competition. My comments strongly support the authors’ focus on empirical evaluation, using examples and experience from machine learning research, particularly focusing on the problem of algorithmic complexity. I argue that even broader and deeper empirical evaluation should be undertaken by the researchers who study causal inference. Finally, I highlight a few key conclusions that suggest where future research should focus.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This is a contribution to the discussion of the paper by Dorie et al. (Statist. Sci. 34 (2019) 43–68), which reports the lessons learned from 2016 Atlantic Causal Inference Conference Competition. My comments strongly support the authors’ focus on empirical evaluation, using examples and experience from machine learning research, particularly focusing on the problem of algorithmic complexity. I argue that even broader and deeper empirical evaluation should be undertaken by the researchers who study causal inference. Finally, I highlight a few key conclusions that suggest where future research should focus.
Huseyin Oktay, Akanksha Atrey, David Jensen
Identifying when effect restoration will improve estimates of causal effect Proceedings Article
In: Proceedings of the 2019 SIAM International Conference on Data Mining, pp. 190–198, Society for Industrial and Applied Mathematics 2019.
Abstract | Links | BibTeX | Tags:
@inproceedings{oktay2019identifying,
title = {Identifying when effect restoration will improve estimates of causal effect},
author = {Huseyin Oktay and Akanksha Atrey and David Jensen},
url = {https://doi.org/10.1137/1.9781611975673.22},
year = {2019},
date = {2019-01-01},
booktitle = {Proceedings of the 2019 SIAM International Conference on Data Mining},
pages = {190--198},
organization = {Society for Industrial and Applied Mathematics},
abstract = {Several methods have been developed that combine multiple models learned on different data sets and then use that combination to reach conclusions that would not have been possible with any one of the models alone. We examine one such method—effect restoration—which was originally developed to mitigate the effects of poorly measured confounding variables in a causal model. We show how effect restoration can be used to combine results from different machine learning models and how the combined model can be used to estimate causal effects that are not identifiable from either of the original studies alone. We characterize the performance of effect restoration by using both theoretical analysis and simulation studies. Specifically, we show how conditional independence tests and common assumptions can help distinguish when effect restoration should and should not be applied, and we use empirical analysis to show the limited range of conditions under which effect restoration should be applied in practical situations.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Several methods have been developed that combine multiple models learned on different data sets and then use that combination to reach conclusions that would not have been possible with any one of the models alone. We examine one such method—effect restoration—which was originally developed to mitigate the effects of poorly measured confounding variables in a causal model. We show how effect restoration can be used to combine results from different machine learning models and how the combined model can be used to estimate causal effects that are not identifiable from either of the original studies alone. We characterize the performance of effect restoration by using both theoretical analysis and simulation studies. Specifically, we show how conditional independence tests and common assumptions can help distinguish when effect restoration should and should not be applied, and we use empirical analysis to show the limited range of conditions under which effect restoration should be applied in practical situations.
Emma Tosch, Eytan Bakshy, Emery D Berger, David Jensen, J Eliot B Moss
PlanAlyzer: assessing threats to the validity of online experiments Journal Article
In: Proceedings of the ACM on Programming Languages, vol. 3, no. OOPSLA, pp. 1–30, 2019.
Abstract | Links | BibTeX | Tags:
@article{tosch2019planalyzer,
title = {PlanAlyzer: assessing threats to the validity of online experiments},
author = {Emma Tosch and Eytan Bakshy and Emery D Berger and David Jensen and J Eliot B Moss},
url = {https://dl.acm.org/doi/pdf/10.1145/3360608},
year = {2019},
date = {2019-01-01},
journal = {Proceedings of the ACM on Programming Languages},
volume = {3},
number = {OOPSLA},
pages = {1--30},
publisher = {ACM New York, NY, USA},
abstract = {Online experiments are ubiquitous. As the scale of experiments has grown, so has the complexity of their design and implementation. In response, firms have developed software frameworks for designing and deploying online experiments. Ensuring that experiments in these frameworks are correctly designed and that their results are trustworthy---referred to as *internal validity*---can be difficult. Currently, verifying internal validity requires manual inspection by someone with substantial expertise in experimental design. We present the first approach for statically checking the internal validity of online experiments. Our checks are based on well-known problems that arise in experimental design and causal inference. Our analyses target PlanOut, a widely deployed, open-source experimentation framework that uses a domain-specific language to specify and run complex experiments. We have built a tool, PlanAlyzer, that checks PlanOut programs for a variety of threats to internal validity, including failures of randomization, treatment assignment, and causal sufficiency. PlanAlyzer uses its analyses to automatically generate *contrasts*, a key type of information required to perform valid statistical analyses over experimental results. We demonstrate PlanAlyzer's utility on a corpus of PlanOut scripts deployed in production at Facebook, and we evaluate its ability to identify threats to validity on a mutated subset of this corpus. PlanAlyzer has both precision and recall of 92% on the mutated corpus, and 82% of the contrasts it automatically generates match hand-specified data.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Online experiments are ubiquitous. As the scale of experiments has grown, so has the complexity of their design and implementation. In response, firms have developed software frameworks for designing and deploying online experiments. Ensuring that experiments in these frameworks are correctly designed and that their results are trustworthy---referred to as *internal validity*---can be difficult. Currently, verifying internal validity requires manual inspection by someone with substantial expertise in experimental design. We present the first approach for statically checking the internal validity of online experiments. Our checks are based on well-known problems that arise in experimental design and causal inference. Our analyses target PlanOut, a widely deployed, open-source experimentation framework that uses a domain-specific language to specify and run complex experiments. We have built a tool, PlanAlyzer, that checks PlanOut programs for a variety of threats to internal validity, including failures of randomization, treatment assignment, and causal sufficiency. PlanAlyzer uses its analyses to automatically generate *contrasts*, a key type of information required to perform valid statistical analyses over experimental results. We demonstrate PlanAlyzer's utility on a corpus of PlanOut scripts deployed in production at Facebook, and we evaluate its ability to identify threats to validity on a mutated subset of this corpus. PlanAlyzer has both precision and recall of 92% on the mutated corpus, and 82% of the contrasts it automatically generates match hand-specified data.
Przemyslaw A. Grabowicz, Nicholas Perello, Kenta Takatsu
Resilience of Supervised Learning Algorithms to Discriminatory Data Perturbations Journal Article
In: 2019.
Abstract | Links | BibTeX | Tags: Fairness
@article{Grabowicz2019c,
title = {Resilience of Supervised Learning Algorithms to Discriminatory Data Perturbations},
author = {Przemyslaw A. Grabowicz and Nicholas Perello and Kenta Takatsu},
url = {http://arxiv.org/abs/1912.08189},
year = {2019},
date = {2019-12-01},
abstract = {Discrimination is a focal concern in supervised learning algorithms augmenting human decision-making. These systems are trained using historical data, which may have been tainted by discrimination, and may learn biases against the protected groups. An important question is how to train models without propagating discrimination. In this study, we i) define and model discrimination as perturbations of a data-generating process and show how discrimination can be induced via attributes correlated with the protected attributes; ii) introduce a measure of resilience of a supervised learning algorithm to potentially discriminatory data perturbations, iii) propose a novel supervised learning algorithm that inhibits discrimination, and iv) show that it is more resilient to discriminatory perturbations in synthetic and real-world datasets than state-of-the-art learning algorithms. The proposed method can be used with general supervised learning algorithms and avoids inducement of discrimination, while maximizing model accuracy.},
keywords = {Fairness},
pubstate = {published},
tppubtype = {article}
}
Discrimination is a focal concern in supervised learning algorithms augmenting human decision-making. These systems are trained using historical data, which may have been tainted by discrimination, and may learn biases against the protected groups. An important question is how to train models without propagating discrimination. In this study, we i) define and model discrimination as perturbations of a data-generating process and show how discrimination can be induced via attributes correlated with the protected attributes; ii) introduce a measure of resilience of a supervised learning algorithm to potentially discriminatory data perturbations, iii) propose a novel supervised learning algorithm that inhibits discrimination, and iv) show that it is more resilient to discriminatory perturbations in synthetic and real-world datasets than state-of-the-art learning algorithms. The proposed method can be used with general supervised learning algorithms and avoids inducement of discrimination, while maximizing model accuracy.
Amanda Gentzel, Dan Garant, David Jensen
The Case for Evaluating Causal Models Using Interventional Measures and Empirical Data Proceedings Article
In: Advances in Neural Information Processing Systems, Curran Associates, Inc., 2019.
Abstract | Links | BibTeX | Tags: Causal Modeling
@inproceedings{gentzel2019case,
title = {The Case for Evaluating Causal Models Using Interventional Measures and Empirical Data},
author = {Amanda Gentzel and Dan Garant and David Jensen},
url = {https://proceedings.neurips.cc/paper/2019/file/a87c11b9100c608b7f8e98cfa316ff7b-Paper.pdf},
year = {2019},
date = {2019-01-01},
booktitle = {Advances in Neural Information Processing Systems},
volume = {32},
publisher = {Curran Associates, Inc.},
abstract = {Causal inference is central to many areas of artificial intelligence, including complex reasoning, planning, knowledge-base construction, robotics, explanation, and fairness. An active community of researchers develops and enhances algorithms that learn causal models from data, and this work has produced a series of impressive technical advances. However, evaluation techniques for causal modeling algorithms have remained somewhat primitive, limiting what we can learn from experimental studies of algorithm performance, constraining the types of algorithms and model representations that researchers consider, and creating a gap between theory and practice. We argue for more frequent use of evaluation techniques that examine interventional measures rather than structural or observational measures, and that evaluate those measures on empirical data rather than synthetic data. We survey the current practice in evaluation and show that the techniques we recommend are rarely used in practice. We show that such techniques are feasible and that data sets are available to conduct such evaluations. We also show that these techniques produce substantially different results than using structural measures and synthetic data.},
keywords = {Causal Modeling},
pubstate = {published},
tppubtype = {inproceedings}
}
Causal inference is central to many areas of artificial intelligence, including complex reasoning, planning, knowledge-base construction, robotics, explanation, and fairness. An active community of researchers develops and enhances algorithms that learn causal models from data, and this work has produced a series of impressive technical advances. However, evaluation techniques for causal modeling algorithms have remained somewhat primitive, limiting what we can learn from experimental studies of algorithm performance, constraining the types of algorithms and model representations that researchers consider, and creating a gap between theory and practice. We argue for more frequent use of evaluation techniques that examine interventional measures rather than structural or observational measures, and that evaluate those measures on empirical data rather than synthetic data. We survey the current practice in evaluation and show that the techniques we recommend are rarely used in practice. We show that such techniques are feasible and that data sets are available to conduct such evaluations. We also show that these techniques produce substantially different results than using structural measures and synthetic data.
Emma Tosch, Kaleigh Clary, John Foley, David Jensen
Toybox: A Suite of Environments for Experimental Evaluation of Deep Reinforcement Learning Miscellaneous
2019.
Abstract | Links | BibTeX | Tags: Explainable AI
@misc{DBLP:journals/corr/abs-1905-02825,
title = {Toybox: A Suite of Environments for Experimental Evaluation of Deep
Reinforcement Learning},
author = {Emma Tosch and Kaleigh Clary and John Foley and David Jensen},
url = {http://arxiv.org/abs/1905.02825},
year = {2019},
date = {2019-01-01},
journal = {CoRR},
volume = {abs/1905.02825},
abstract = {Evaluation of deep reinforcement learning (RL) is inherently challenging. In particular, learned policies are largely opaque, and hypotheses about the behavior of deep RL agents are difficult to test in black-box environments. Considerable effort has gone into addressing opacity, but almost no effort has been devoted to producing high quality environments for experimental evaluation of agent behavior. We present TOYBOX, a new high-performance, open-source* subset of Atari environments re-designed for the experimental evaluation of deep RL. We show that TOYBOX enables a wide range of experiments and analyses that are impossible in other environments.},
keywords = {Explainable AI},
pubstate = {published},
tppubtype = {misc}
}
Evaluation of deep reinforcement learning (RL) is inherently challenging. In particular, learned policies are largely opaque, and hypotheses about the behavior of deep RL agents are difficult to test in black-box environments. Considerable effort has gone into addressing opacity, but almost no effort has been devoted to producing high quality environments for experimental evaluation of agent behavior. We present TOYBOX, a new high-performance, open-source* subset of Atari environments re-designed for the experimental evaluation of deep RL. We show that TOYBOX enables a wide range of experiments and analyses that are impossible in other environments.
Sam Witty, David Jensen
Causal Graphs vs. Causal Programs: The Case of Conditional Branching Proceedings Article
In: First Conference on Probabilistic Programming (ProbProg), 2018.
Abstract | Links | BibTeX | Tags: Probabilistic Programming
@inproceedings{witty2018causal,
title = {Causal Graphs vs. Causal Programs: The Case of Conditional Branching},
author = {Sam Witty and David Jensen},
url = {https://arxiv.org/pdf/2007.07127.pdf},
year = {2018},
date = {2018-01-01},
booktitle = {First Conference on Probabilistic Programming (ProbProg)},
abstract = {We evaluate the performance of graph-based causal discovery algorithms when the generative process is a probabilistic program with conditional branching. Using synthetic experiments, we demonstrate empirically that graph-based causal discovery algorithms are able to learn accurate associational distributions for probabilistic programs with contextsensitive structure, but that those graphs fail to accurately model the effects of interventions on the programs},
keywords = {Probabilistic Programming},
pubstate = {published},
tppubtype = {inproceedings}
}
We evaluate the performance of graph-based causal discovery algorithms when the generative process is a probabilistic program with conditional branching. Using synthetic experiments, we demonstrate empirically that graph-based causal discovery algorithms are able to learn accurate associational distributions for probabilistic programs with contextsensitive structure, but that those graphs fail to accurately model the effects of interventions on the programs