Explaining Multi-Criteria Decision Aiding Models with an Extended Shapley Value

Labreuche, Christophe; Fossier, Simon

doi:10.24963/ijcai.2018/46

Cited by 30 publications

(16 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The third approach falls within the theory of causality and explanation in databases (Livshits and Kimelfeld 2020;Livshits, Ilyas, et al 2019;Labreuche and Fossier 2018;Meliou, Roy, and Suciu 2014). It consists of understand the underlying causes of a particular observation by determining the relative contribution of features in machine-learning predictions (Labreuche and Fossier 2018), the responsibility of tuples to database queries (Bertossi and Geerts 2020;Livshits, Bertossi, et al 2019), or the reliability of data sources (Cholvy, Perrussel, and Thévenin 2017). In (Livshits and Kimelfeld 2020), a Shapley value is used to quantify the extent to which the database violates a set of integrity constraints.…”

Section: Related Workmentioning

confidence: 99%

Data Obsolescence Detection in the Light of Newly Acquired Valid Observations

Chaieb,

Hnich,

Mrad

2021

Preprint

View full text Add to dashboard Cite

The information describing the conditions of a system or a person is constantly evolving and may become obsolete and contradict other information. A database, therefore, must be consistently updated upon the acquisition of new valid observations that contradict obsolete ones contained in the database. In this paper, we propose a novel approach for dealing with the information obsolescence problem. Our approach aims to detect, in real-time, contradictions between observations and then identify the obsolete ones, given a representation model. Since we work within an uncertain environment characterized by the lack of information, we choose to use a Bayesian network as our representation model and propose a new approximate concept, ǫ-Contradiction. The new concept is parameterised by a confidence level of having a contradiction in a set of observations. We propose a polynomial-time algorithm for detecting obsolete information. We show that the resulting obsolete information is better represented by an AND-OR tree than a simple set of observations. Finally, we demonstrate the effectiveness of our approach on a real elderly fall-prevention database and showcase how this tree can be used to give reliable recommendations to doctors. Our experiments give systematically and substantially very good results.

show abstract

Section: Related Workmentioning

confidence: 99%

Data Obsolescence Detection in the Light of Newly Acquired Valid Observations

Chaieb,

Hnich,

Mrad

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…A more general, multi-layered form was introduced in (Ovchinnikov 2002), and later refined in (Grabisch et al 2009;Angilella et al 2013) in order to support a tree-like hierarchical structure. This hierarchical extension, referred to as HCI, enables to represent some given structure on the criteria, with better interpretability than flat CIs when the number of critera increases (Labreuche and Fossier 2018). HCIs can also be fitted with marginal attribute rescalings, called marginal utilities, for a better expressivity (Bresson et al 2020b).…”

Section: Related Workmentioning

confidence: 99%

“…A hierarchical version of the Shapley value was also established for analyzing such models (Labreuche and Fossier 2018). Finally, the normalization and monotonicity constraints are still valid, as a composition of CIs.…”

Section: Hci Modelmentioning

confidence: 99%

On the Identifiability of Hierarchical Decision Models

Bresson

Cohen

Hüllermeier

et al. 2021

Proceedings of the Eighteenth International Conference on Principles of Knowledge Representation and Reasoning

Self Cite

View full text Add to dashboard Cite

Interpretability is a desirable property for machine learning and decision models, particularly in the context of safety-critical applications. Another most desirable property of the sought model is to be unique or {\em identifiable} in the considered class of models: the fact that the same functional dependency can be represented by a number of syntactically different models adversely affects the model interpretability, and prevents the expert from easily checking their validity. This paper focuses on the Choquet integral (CI) models and their hierarchical extensions (HCI). HCIs aim to support expert decision making, by gradually aggregating preferences based on criteria; they are widely used in multi-criteria decision aiding {and are receiving interest from the} Machine Learning {community}, as they preserve the high readability of CIs while efficiently scaling up w.r.t. the number of criteria. The main contribution is to establish the identifiability property of HCI under mild conditions: two HCIs implementing the same aggregation function on the criteria space necessarily have the same hierarchical structure and aggregation parameters. The identifiability property holds even when the marginal utility functions are learned from the data. This makes the class of HCI models a most appropriate choice in domains where the model interpretability and reliability are of primary concern.

show abstract

“…Since SHAP belongs to the same class of methods as LIME [Lundberg and Lee, 2017], the same phenomena could be at play. Labreuche and Fossier [2018] leverage Shapley values to explain the result of aggregation models for Multi-Criteria Decision Aiding. However, their solution requires full knowledge of the models involved, whereas we want to be agnostic about individual models.…”

Section: Related Workmentioning

confidence: 99%

SMACE: A New Method for the Interpretability of Composite Decision Systems

Lopardo¹,

Garreau²,

Precioso³

et al. 2021

Preprint

View full text Add to dashboard Cite

Interpretability is a pressing issue for decision systems. Many post hoc methods have been proposed to explain the predictions of any machine learning model. However, business processes and decision systems are rarely centered around a single, standalone model. These systems combine multiple models that produce key predictions, and then apply decision rules to generate the final decision. To explain such decision, we present SMACE, Semi-Model-Agnostic Contextual Explainer, a novel interpretability method that combines a geometric approach for decision rules with existing post hoc solutions for machine learning models to generate an intuitive feature ranking tailored to the end user. We show that established model-agnostic approaches produce poor results in this framework.

show abstract

Explaining Multi-Criteria Decision Aiding Models with an Extended Shapley Value

Cited by 30 publications

References 10 publications

Data Obsolescence Detection in the Light of Newly Acquired Valid Observations

Data Obsolescence Detection in the Light of Newly Acquired Valid Observations

On the Identifiability of Hierarchical Decision Models

SMACE: A New Method for the Interpretability of Composite Decision Systems

Contact Info

Product

Resources

About