Local Explanations via Necessity and Sufficiency: Unifying Theory and Practice

Watson, David; Gultchin, Limor; Taly, Ankur; Floridi, Luciano

doi:10.48550/arxiv.2103.14651

Cited by 6 publications

(10 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…certa builds perturbed copies in a data-driven way, using sequences of tokens that come from the training set distribution and hence are more likely to be correctly classified by the ER system. Perturbed copies are used to compute saliency and counterfactual explanations according to the probabilistic framework developed in [36], which associates the former to the probability of necessity, and the latter to the probability of sufficiency.…”

Section: Foundations and Problem Statementmentioning

confidence: 99%

“…It has been observed that saliency and counterfactual explanation methods are different but complimentary methods to be used to best evaluate causality aspects of a classifier prediction [14]. Saliency methods align well with the notion of necessity, while counterfactual explanation methods align with the notion of sufficiency [36].…”

Section: Introductionmentioning

confidence: 99%

“…This paper presents certa, an original method that provides both saliency and counterfactual explanations for ER systems. certa considers specific characteristics of the ER task, and builds on the sound theoretical framework developed by Watson et al [36], which frames the concepts of probability of necessity and probability of sufficiency in the context of explanations.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Effective Explanations for Entity Resolution Models

Teofili¹,

Firmani²,

Koudas³

et al. 2022

Preprint

View full text Add to dashboard Cite

Entity resolution (ER) aims at matching records that refer to the same real-world entity. Although widely studied for the last 50 years, ER still represents a challenging data management problem, and several recent works have started to investigate the opportunity of applying deep learning (DL) techniques to solve this problem. In this paper, we study the fundamental problem of explainability of the DL solution for ER. Understanding the matching predictions of an ER solution is indeed crucial to assess the trustworthiness of the DL model and to discover its biases. We treat the DL model as a black box classifier and -while previous approaches to provide explanations for DL predictions are agnostic to the classification task -we propose the certa approach that is aware of the semantics of the ER problem. Our approach produces both saliency explanations, which associate each attribute with a saliency score, and counterfactual explanations, which provide examples of values that can flip the prediction. certa builds on a probabilistic framework that aims at computing the explanations evaluating the outcomes produced by using perturbed copies of the input records.We experimentally evaluate certa's explanations of state-of-the-art ER solutions based on DL models using publicly available datasets, and demonstrate the effectiveness of certa over recently proposed methods for this problem.

show abstract

Section: Foundations and Problem Statementmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Effective Explanations for Entity Resolution Models

Teofili¹,

Firmani²,

Koudas³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Reverse causal inference. Our work is also related to a body of work on reverse causal inference, a task that aims to nd "causes of e ects" (Chalupka et al, 2017(Chalupka et al, , 2015Galhotra et al, 2021;Gelman & Imbens, 2013;Janzing & Schölkopf, 2015;Kilbertus et al, 2018;Kommiya Mothilal et al, 2021;Paul, 2017;Schölkopf et al, 2012Schölkopf et al, , 2013Wang & Culotta, 2020Watson et al, 2021;Weichwald et al, 2014). Existing approaches formulate the search for causes as causal hypotheses generation and testing.…”

Section: Unsupervisedmentioning

confidence: 99%

Desiderata for Representation Learning: A Causal Perspective

Wang

Jordan

2021

Preprint

View full text Add to dashboard Cite

Representation learning constructs low-dimensional representations to summarize essential features of high-dimensional data. This learning problem is often approached by describing various desiderata associated with learned representations; e.g., that they be non-spurious, e cient, or disentangled. It can be challenging, however, to turn these intuitive desiderata into formal criteria that can be measured and enhanced based on observed data. In this paper, we take a causal perspective on representation learning, formalizing non-spuriousness and e ciency (in supervised representation learning) and disentanglement (in unsupervised representation learning) using counterfactual quantities and observable consequences of causal assertions. This yields computable metrics that can be used to assess the degree to which representations satisfy the desiderata of interest and learn non-spurious and disentangled representations from single observational datasets.

show abstract

Section: Introductionmentioning

confidence: 99%

Effective Explanations for Entity Resolution Models

Teofili

Firmani

Koudas

et al. 2022

2022 IEEE 38th International Conference on Data Engineering (ICDE)

View full text Add to dashboard Cite

Entity resolution (ER) aims at matching records that refer to the same real-world entity. Although widely studied for the last 50 years, ER still represents a challenging data management problem, and several recent works have started to investigate the opportunity of applying deep learning (DL) techniques to solve this problem. In this paper, we study the fundamental problem of explainability of the DL solution for ER. Understanding the matching predictions of an ER solution is indeed crucial to assess the trustworthiness of the DL model and to discover its biases. We treat the DL model as a black box classifier and -while previous approaches to provide explanations for DL predictions are agnostic to the classification task -we propose the CERTA approach that is aware of the semantics of the ER problem. Our approach produces both saliency explanations, which associate each attribute with a saliency score, and counterfactual explanations, which provide examples of values that can flip the prediction. CERTA builds on a probabilistic framework that aims at computing the explanations evaluating the outcomes produced by using perturbed copies of the input records. We experimentally evaluate CERTA's explanations of state-of-the-art ER solutions based on DL models using publicly available datasets, and demonstrate the effectiveness of CERTA over recently proposed methods for this problem.

show abstract

Local Explanations via Necessity and Sufficiency: Unifying Theory and Practice

Cited by 6 publications

References 18 publications

Effective Explanations for Entity Resolution Models

Effective Explanations for Entity Resolution Models

Desiderata for Representation Learning: A Causal Perspective

Effective Explanations for Entity Resolution Models

Contact Info

Product

Resources

About