Factual and Counterfactual Explanations for Black Box Decision Making
Open accessThe DOI matches its Crossref record; 12 of the 14 references it cites were found in a registry or an index.
Abstract
The rise of sophisticated machine learning models has brought accurate but obscure decision systems, which hide their logic, thus undermining transparency, trust, and the adoption of artificial intelligence (AI) in socially sensitive and safety-critical contexts. We introduce a local rule-based explanation method, providing faithful explanations of the decision made by a black box classifier on a specific instance. The proposed method first learns an interpretable, local classifier on a synthetic neighborhood of the instance under investigation, generated by a genetic algorithm. Then, it derives from the interpretable classifier an explanation consisting of a decision rule, explaining the factual reasons of the decision, and a set of counterfactuals, suggesting the changes in the instance features that would lead to a different outcome. Experimental results show that the proposed method outperforms existing approaches in terms of the quality of the explanations and of the accuracy in mimicking the black box.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
"Why Should I Trust You?": Explaining the Predictions of Any Classifier
Ribeiro, M. T., Singh, S., & Guestrin, C. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939778 · Cited by 17,555
Explanation in artificial intelligence: Insights from the social sciences
Miller, T. (2019) · Artificial Intelligence · doi:10.1016/j.artint.2018.07.007 · Cited by 5,302
Anchors: High-Precision Model-Agnostic Explanations
Ribeiro, M. T., Singh, S., & Guestrin, C. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11491 · Cited by 2,162 · cited 1×
“We also evaluate the rule coverage, namely the fraction of instances to explain covered by the rule: large values of coverage means better rule generalization.(4) Figure 8 also shows the average coverage of the decision rules, with LORE showing a consistently better coverage than ANCHOR.”
Counterfactual explanations without opening the black box: automated decisions and the GDPR
Wachter, S., Mittelstadt, B., & Russell, C. (2017) · SSRN Electronic Journal · doi:10.2139/ssrn.3063289 · Cited by 1,061
Evolutionary Computation 1: Basic Algorithms and Operators
(2018) · CRC Press eBooks · doi:10.1201/9781482268713 · Cited by 634
A Survey of Methods for Explaining Black Box Models
Guidotti, R., Monreale, A., Ruggieri, S., Turini, F., Giannotti, F., & Pedreschi, D. (2019) · ACM Computing Surveys · doi:10.1145/3236009 · Cited by 629
Self-citationEvolutionary Computation 2: Advanced Algorithms and Operators
(2000) · Taylor & Francis eBooks · doi:10.1201/9781420034349 · Cited by 451
A review of instance selection methods
Olvera-López, J. A., Carrasco-Ochoa, J. A., Martínez-Trinidad, J. F., & Kittler, J. (2010) · Artificial Intelligence Review · doi:10.1007/s10462-010-9165-y · Cited by 394 · cited 1×
“This task differs from approaches to instance selection based on genetic algorithms.(7) In our case, we cannot assume the availability of the training set of b, or not even that b is a supervised ML predictor for which a training set exists.”
MethodCounterfactuals in Explainable Artificial Intelligence (XAI): Evidence from Human Reasoning
Byrne, R. M. J. (2019) · Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2019/876 · Cited by 285 · cited 1×
“the cognitive psychology literature maintaining that counterfactuals help people to reason on explanations that identify cause-effect or reason-action relations between events.(5,6) Similarly, Grover et al.”
BackgroundA Survey on Evolutionary Instance Selection and Generation
Derrac, J., García, S., & Herrera, F. (2010) · International Journal of Applied Metaheuristic Computing · doi:10.4018/jamc.2010102604 · Cited by 75 · cited 1×
“Our task is instead similar to instance generation in the field of active learning, including evolutionary approaches.(8) We adopt an approach based on a genetic algorithm, which generates z 2 Z1⁄4 [ Z61⁄4 by maximizing the following fitness functions:”
Method
Cited by
Items 101–200 of 438
Boosting Synthetic Data Generation with Effective Nonlinear Causal Discovery
Cinquini, M., Giannotti, F., & Guidotti, R. (2021) · IEEE International Conference on Cognitive Machine Intelligence (CogMI) · doi:10.1109/cogmi52975.2021.00016 · Cited by 25
“LORE [30] is another local explanation method that exploits a synthetic neighborhood generation based on a genetic algorithm to create a more compact dataset around the explained instance.”
MethodSelf-citationOliveira, R., Sörensen, K., & Martens, D. (2023) · European Journal of Operational Research · doi:10.1016/j.ejor.2023.08.031 · Cited by 25
Factual and Counterfactual Explanations in Fuzzy Classification Trees
Fernández, G., Aledo, J. A., Gámez, J. A., & Puerta, J. M. (2022) · IEEE Transactions on Fuzzy Systems · doi:10.1109/tfuzz.2022.3179582 · Cited by 24
“There are different definitions for these kinds of explanations depending on the algorithm used: decision trees [10], [13], regression [8], probabilistic classifiers [14], [15], etc.”
BackgroundMethodInfluentialLearning Paradigms for Hybrid Decision-Making Systems
Punzi, C., Pellungrini, R., Setzu, M., Giannotti, F., & Pedreschi, D. (2026) · ACM Computing Surveys · doi:10.1145/3802522 · Cited by 25
Self-citationGMM‐LIME explainable machine learning model for interpreting sensor‐based human gait
Mulwa, M. M., Mwangi, R. W., & Mindila, A. (2024) · Engineering Reports · doi:10.1002/eng2.12864 · Cited by 24
“22 Model agnostic approaches are popular because they are effective on any ML model.”
Extracting per Query Valid Explanations for Blackbox Learning-to-Rank Models
Singh, J., Khosla, M., Zhenye, W., & Anand, A. (2021) · ACM SIGIR International Conference on Theory of Information Retrieval (ICTIR) · doi:10.1145/3471158.3472241 · Cited by 24
“Model agnostic methods [3, 5, 10, 20] are more general in that they are agnostic to how the underlying model was trained or how the predictions are made.”
BackgroundMethodExplaining Image Classifiers Generating Exemplars and Counter-Exemplars from Latent Representations
Guidotti, R., Monreale, A., Matwin, S., & Pedreschi, D. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i09.7116 · Cited by 23
Self-citationBello, M., Bello, R., García, M.-M., Nowé, A., Sevillano-García, I., & Herrera, F. (2025) · Information Systems Frontiers · doi:10.1007/s10796-025-10668-1 · Cited by 23
“…in terms of input data from the model (tabular data, image data, text data, time series and graphs) and output representation of the explanation (Feature importance [30, 31], Rule-based explanation [32, 33], Prototype-based explanations [34, 35], and Counterfactual-based explanations [36, 37]).”
ExplainableFold: Understanding AlphaFold Prediction with Explainable AI
Tan, J., & Zhang, Y. (2023) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3580305.3599337 · Cited by 22
“From the counterfactual machine learning perspective, this can be considered from two complementary views (Guidotti et al., 2019; Tan et al., 2022): 1) Identify the minimal deletion that will alter the predicted structure and the deleted residues will be the necessary explanation; 2) Identify the maximal deletion that…”
BackgroundTorres-Tello, J. W., & Ko, S. (2021) · Journal of Ambient Intelligence and Humanized Computing · doi:10.1007/s12652-021-03470-9 · Cited by 22
Unifying VXAI: A Systematic Review and Framework for the Evaluation of Explainable AI
Dembinsky, D., Lucieri, A., Frolov, S., Najjar, H., Watanabe, K., & Dengel, A. (2026) · Transactions on machine learning research · doi:10.48550/arxiv.2506.15408 · Cited by 21
Eldrandaly, K. A., Abdel‐Basset, M., Ibrahim, M., & AbdelAziz, N. M. (2022) · Enterprise Information Systems · doi:10.1080/17517575.2022.2098537 · Cited by 21
“(Guidotti et al. 2019) introduce a local rule-based method of explanation that provides faithful explanations of decisions made by black-box classifier on a particular case.”
MethodCREATED: Generating Viable Counterfactual Sequences for Predictive Process Analytics
Hundogan, O. A., Lu, X., Du, Y., & Reijers, H. A. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-34560-9_32 · Cited by 21
A Framework to Learn with InterpretationPreprint
Parekh, J., Mozharovskyi, P., & d’Alché–Buc, F. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.09345 · Cited by 21
“Throughout this paper, we opt for interpretability as in [16] and leave the term explainability for the ability to provide logical explanations or causal reasoning, both requiring more sophisticated frameworks [18, 20, 48].”
BackgroundMethodOn Explaining Unfairness: An Overview
Fragkathoulas, C., Papanikou, V., Karidi, D. P., & Pitoura, E. (2024) · IEEE International Conference on Data Engineering Workshops (ICDEW) · doi:10.1109/icdew61823.2024.00035 · Cited by 20
“A key distinction is that counterfactual explanations focus on revealing the changes needed for different classification outcomes, while counterfactual fairness aims to assess fairness by considering predictions under different demographic scenarios using causal models [71].”
BackgroundEnsemble of Counterfactual Explainers
Guidotti, R., & Ruggieri, S. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-88942-5_28 · Cited by 20
Self-citationA New SVDD Approach to Reliable and Explainable AI
Carlevaro, A., & Mongelli, M. (2021) · IEEE Intelligent Systems · doi:10.1109/mis.2021.3123669 · Cited by 20
“The work on rule extraction is mainly focused on the LLM and the DT algorithms, other approaches may be of interest, such as BEEF 3 and Guidotti et al. ’ s work, 44 which could be the basis for further investigations in intelligible rule extraction.”
MethodYulianti, H. (2021) · MULTINETICS · doi:10.32722/multinetics.v7i1.3769 · Cited by 20
Counterfactual Building and Evaluation via eXplainable Support Vector Data Description
Carlevaro, A., Lenatti, M., Paglialonga, A., & Mongelli, M. (2022) · IEEE Access · doi:10.1109/access.2022.3180026 · Cited by 19
“Extending the validity (covering) of a local rule over neighbour points is not a straightforward matter [23]; for this reason, the LLM is preferred to facilitate the knowledge extraction from the SVDD counterfactuals, by following the approach in [15], [16].”
BackgroundMethodZhang, Z., Cui, H., Xu, R., Xie, Y., Ho, J. C., & Yang, C. (2024) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3637528.3671594 · Cited by 19
“Approaches such as factual counterfactual reasoning [21, 72] and time-aware mechanisms [62, 78] can extract interpretable subsets for EHR data, but they require supervised learning.”
BackgroundA Combinatorial Approach to Explaining Image Classifiers
Chandrasekaran, J., Lei, Y., Kacker, R. N., & Kuhn, D. R. (2021) · IEEE International Conference on Software Testing, Verification and Validation Workshops (ICSTW) · doi:10.1109/icstw52544.2021.00019 · Cited by 19
“A counterfactual approach tries to identify a minimum set of features that, if removed, would cause a different decision to be made [9].”
BackgroundExemplars and Counterexemplars Explanations for Image Classifiers, Targeting Skin Lesion Labeling
Metta, C., Guidotti, R., Yin, Y., Gallinari, P., & Rinzivillo, S. (2021) · IEEE Symposium on Computers and Communications (ISCC) · doi:10.1109/iscc53001.2021.9631485 · Cited by 19
“2) may be accomplished using different strategies ranging from pure random strategy using a given distribution to a genetic approach maximizing a fitness function [18].”
BackgroundInfluentialSelf-citationWarren, G., Keane, M. T., Guéret, C., & Delaney, E. D. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2303.09297 · Cited by 19
“Many different methods have been proposed in XAI to explain single instance-prediction from a single AI model using a single counterfactual: methods adopting optimisation approaches [8, 27, 48, 62], instance-based approaches [4, 9, 31, 44, 49], distributional approaches [10, 25, 52] and other approaches [21, 22] (for…”
TripleTree: A Versatile Interpretable Representation of Black Box Agents and their Environments
Bewley, T., & Jonathan Lawry, B. (2021) · AAAI Publications (The Association for the Advancement of Artificial Intelligence (AAAI)) · doi:10.1609/aaai.v35i13.17360 · Cited by 18
Accurate Explanation Model for Image Classifiers using Class Association Embedding
Xie, R., Chen, J., Jiang, L., Xiao, R., Pan, Y., & Cai, Y. (2024) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde60146.2024.00180 · Cited by 18
“LIME [15], a well-known perturbation method, generates new data by perturbing the input and trains a simple and interpretable model to approximate the output of a complex model.”
MethodEvaluating the Privacy Exposure of Interpretable Global Explainers
Naretto, F., Monreale, A., & Giannotti, F. (2022) · IEEE International Conference on Cognitive Machine Intelligence (CogMI) · doi:10.1109/cogmi56440.2022.00012 · Cited by 18
“As pointed out in [1], there exists two families of explainers in the eXplainable Artificial Intelligence (XAI) literature: local explainers, which explain the reason for a specific instance classification [2]–[4] and global explainers, which explain the logic of the machine learning (ML) model as a whole [5]–[7].”
BackgroundSelf-citationSpArX: Sparse Argumentative Explanations for Neural Networks
Ayoobi, H., Potyka, N., & Toni, F. (2023) · Frontiers in artificial intelligence and applications · doi:10.3233/faia230265 · Cited by 18
An Empirical Comparison of Interpretable Models to Post-Hoc Explanations
Mahya, P., & Fürnkranz, J. (2023) · AI · doi:10.3390/ai4020023 · Cited by 17
“LOcal Rule-based Explanations (LORE) is a variant of LIME, which is specifically tailored to rules as local surrogate models [14].”
BackgroundMethodAlJalaud, E., & Hosny, M. (2024) · IEEE Access · doi:10.1109/access.2024.3404043 · Cited by 17
“We used the default hyperparameters as recommended in [2] and [41].”
MethodVelmurugan, M., Ouyang, C., Xu, Y., Sindhgatta, R., Wickramanayake, B., & Moreira, C. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.109772 · Cited by 18
Explaining short text classification with diverse synthetic exemplars and counter-exemplars
Lampridis, O., State, L., Guidotti, R., & Ruggieri, S. (2022) · Machine Learning · doi:10.1007/s10994-022-06150-7 · Cited by 17
“…Z and Y = b ( ̃ Z ) , xspells builds a latent decision tree ldt (line 5) acting as a local surrogate of the black box, i.e., being able to locally mime the behavior of b . xspells adopts decision trees because decision rules can be naturally derived from a root-to-leaf path (Guidotti et al. 2019a).”
BackgroundMethodInfluentialSelf-citationFörster, M., Hühn, P., Klier, M., & Kluge, K. (2022) · Journal of Decision System · doi:10.1080/12460125.2022.2119707 · Cited by 17
“The first class locally approximates the AI system with a simpler model from which a foil is extracted (Guidotti et al., 2019).”
BackgroundInterpretable Regional Descriptors: Hyperbox-Based Local Explanations
Dandl, S., Casalicchio, G., Bischl, B., & Bothmann, L. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43418-1_29 · Cited by 17
Opacity, Machine Learning and Explainable AI
Fernández, A. (2023) · The International library of ethics, law and technology · doi:10.1007/978-3-031-48135-2_3 · Cited by 16
Understanding User Preferences in Explainable Artificial Intelligence: A Mapping Function Proposal
Hashemi, M., Darejeh, A., & Cruz, F. (2025) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3733837 · Cited by 16
“Saliency maps offer a straightforward and intuitive visualization of influential parts of input data, primarily used for images, aiding in understanding the model’s behavior.”
Decoding AI: Transparent Models forUnderstandable Decision-Making
Kanungo, S. (2023) · Journal of Propulsion Technology · Cited by 16
Martínez, M. A. M., Nadj, M., Langner, M., Toreini, P., & Maedche, A. (2023) · ACM Transactions on Interactive Intelligent Systems · doi:10.1145/3607145 · Cited by 16
“Additionally, it has been argued that the perturbed samples can contain many incorrectly classiied instances, as the sampling process does not consider the density of the data [37], which can produce data points with high prediction uncertainty.”
BackgroundAIMEE: An Exploratory Study of How Rules Support AI Developers to Explain and Edit Models
Piorkowski, D., Vejsbjerg, I., Cornec, O., Daly, E., & Alkan, Ö. (2023) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3610046 · Cited by 16
“There are two primary types of interpretable ML models: (1) models that are interpretable by design [12], and (2) black box models that can be explained using either a global explainer [13, 23, 25, 27, 52, 84] or a local explainer [37, 74–76].”
MethodOn Quantitative Evaluations of CounterfactualsPreprint
Hvilshøj, F., Iosifidis, A., & Assent, I. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.00177 · Cited by 16
“To name but a few, some are based on heatmaps that identify salient input features [22, 3, 18, 4], others rely on transparent surrogate models [8, 1], and some produce counterfactual examples [30, 6, 25].”
BackgroundXAI-based Data Visualization in Multimodal Medical DataPreprint
Sharma, S., Singh, M., Mcdaid, L. J., & Bhattacharyya, S. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.07.11.664302 · Cited by 16
Kozielski, M., Sikora, M., & Wawrowski, Ł. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.113092 · Cited by 15
“In this way, both local and global explanations can be generated, and the resulting explanations can take different forms e.g. linear regression coefficients [12], decision tree based rules determining coun-terfactual explanations [13] or rule-based explantions [14].”
BackgroundGenerating Counterfactual Explanations Under Temporal Constraints
Buliga, A., Francescomarino, C. D., Ghidini, C., Montali, M., & Ronzani, M. (2025) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v39i15.33715 · Cited by 15
“…Local Rule-Based Explanations (LORE) framework (Guidotti et al. 2019), which generates counterfactual explanations leveraging a surrogate decision tree model us-ing a genetically generated neighbourhood of artificial data instances to be trained (Huang, Metzger, and Pohl 2022; Guidotti et al. 2019).”
Buliga, A., Francescomarino, C. D., Ghidini, C., Donadello, I., & Maggi, F. M. (2025) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-025-01117-3 · Cited by 15
“While single-objective solutions may converge faster due to a lower complexity [22], multi-objective GAs have the advantage of providing a Pareto Front of optimal solutions, where trade-offs are made between different optimisation objectives [12].”
Kwubeghari, A., & Ezeji, N. G. (2025) · ABUAD Journal of Engineering Research and Development (AJERD) · doi:10.53982/ajerd.2025.0802.32-j · Cited by 15
Crash Prediction and Risk Assessment with Individual Mobility Networks
Guidotti, R., & Nanni, M. (2020) · IEEE International Conference on Mobile Data Management (MDM) · doi:10.1109/mdm48529.2020.00030 · Cited by 15
“using rule-based methods [30]) and infer actionable changes in driving habits that might help the user exit the risk zone.”
MethodSelf-citationFairShades: Fairness Auditing via Explainability in Abusive Language Detection Systems
Manerba, M. M., & Guidotti, R. (2021) · IEEE International Conference on Cognitive Machine Intelligence (CogMI) · doi:10.1109/cogmi52975.2021.00014 · Cited by 15
Self-citationGenerative Model for Decision Trees
Guidotti, R., Monreale, A., Setzu, M., & Volpi, G. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i19.30104 · Cited by 15
“Our proposal follows the latter direction by focusing on decision trees (DTs) (Breiman et al. 2017) that have a structure directly providing the learned decision logic in human-comprehensible terms (Craven et al. 1995; Guidotti et al. 2019a).”
BackgroundSelf-citationCausality-Aware Local Interpretable Model-Agnostic Explanations
Cinquini, M., & Guidotti, R. (2024) · Communications in computer and information science · doi:10.1007/978-3-031-63800-8_6 · Cited by 15
“We highlight that our proposal of the causality-aware explanation method can be easily adapted to extend and improve other model-agnostic explainers such as lore [13], or shap [14].”
MethodSelf-citationAlonso, J. M., Toja-Alamancos, J., & Bugarin, A. (2020) · IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) · doi:10.1109/fuzz48607.2020.9177770 · Cited by 15
“• Local Rule-based Explanation (LORE) [17] provides users with a factual explanation of reasons why a decision is made for a given test data instance, along with one or more counterfactual explanations which hypothesize what may have happened in case some features in the given data instance were…”
BackgroundRyan, C., Guéret, C., Berry, D. P., Corcoran, M., Keane, M. T., & Namee, B. M. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2101.02188 · Cited by 15
“…methods generate counterfactuals by perturbing features of the original instance (Wachter, Mittelstadt, and Russell 2017), or nearest unlike neighbor (Keane and Smyth 2020); which can be seen as a follow on from the popular LIME methods (Guidotti et al. 2019; Ribeiro, Singh, and Guestrin 2018).”
MethodYasui, D., & Sato, H. (2025) · IEEE Access · doi:10.1109/access.2025.3553505 · Cited by 14
“Anchor [7] creates rules for input-output relationships from samples, and LORE [8] constructs decision trees from samples.”
FACEGroup: Feasible and Actionable Counterfactual Explanations for Group Fairness
Fragkathoulas, C., Papanikou, V., Pitoura, E., & Terzi, E. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-06078-5_3 · Cited by 14
“Subsequent work [15,25,18,12,34,31,27] refined CF generation, emphasizing properties such as feasibility, actionability, sparsity [34], and robustness[16].”
Understanding machine learning classifier decisions in automated radiotherapy quality assurance
Chen, Y., Aleman, D. M., Purdie, T. G., & McIntosh, C. (2021) · Physics in Medicine and Biology · doi:10.1088/1361-6560/ac3e0e · Cited by 14
Local Multi-label Explanations for Random Forest
Mylonas, N., Mollas, I., Bassiliades, N., & Tsoumakas, G. (2023) · Communications in computer and information science · doi:10.1007/978-3-031-23618-1_25 · Cited by 14
“Another interesting technique, LORE [22], uses a genetic algorithm to generate neighbors, which are then used to train a decision tree that produces rule-based explanations.”
MethodContextualizing Support Vector Machine Predictions
Loor, M., & Tré, G. D. (2020) · International Journal of Computational Intelligence Systems · doi:10.2991/ijcis.d.200910.002 · Cited by 14
“…code of LIME (which is available in https://github.com/marcotcr/lime) with the same configuration of Regarding ABELE, it is an extension of LORE [22] that, in a similar way to LIME, tries to explain a prediction by building an interpretable local classifier with a synthetic neighborhood of the…”
MethodMatrix Profile-Based Interpretable Time Series Classifier
Guidotti, R., & D’Onofrio, M. (2021) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2021.699448 · Cited by 14
“…intrinsic interpretability (Breiman et al., 1984; Craven and Shavlik, 1995) that allows us to reason both “ factually, ” deriving the reasons why a certain decision is taken, and “ counterfactually, ” observing what should have been dissimilar to reach a different outcome (Guidotti et al., 2019).”
BackgroundSelf-citationBeyond Trivial Counterfactual Explanations with Diverse Valuable ExplanationsPreprint
Rodríguez, P., Caccia, M., Lacoste, A., Zamparo, L., Laradji, I., Charlin, L., & Vázquez, D. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.10226 · Cited by 14
“[14] first construct a local neighbourhood around test instances, finding both positive and negative exemplars within the neighbourhood.”
BackgroundHuman-centred explanations for artificial intelligence systems
Baber, C., Kandola, P., Apperly, I. A., & McCormick, E. (2024) · Ergonomics · doi:10.1080/00140139.2024.2334427 · Cited by 13
A Two‐Stage Interpretable Model to Explain Classifier in Credit Risk Prediction
Wang, L., Yu, Z., Ma, J., Chen, X., & Wu, C. (2025) · Journal of Forecasting · doi:10.1002/for.3288 · Cited by 13
Stevens, A., Ouyang, C., Smedt, J. D., & Moreira, C. (2025) · IEEE Transactions on Services Computing · doi:10.1109/tsc.2025.3609837 · Cited by 13
“On the other hand, LORELEY, an extension of the LORE technique [6] introduced by Huang et al. [58], addresses control flow attribute representation challenges.”
Mutahar, G., & Miller, T. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2211.10807 · Cited by 13
Aryal, S., & Keane, M. T. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-63646-2_3 · Cited by 12
“Hence, there are now possibly > 150 distinct XAI methods that compute counterfactual explanations using either optimisation techniques [33,22,13], instance-based approaches [15,28,5], genetic algorithms [9,27], deep neural networks [10,30] and so on.”
CARMA: A practical framework to generate recommendations for causal algorithmic recourse at scale
Majumdar, A., & Valera, I. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3659003 · Cited by 12
Uncovering Patterns for Local Explanations in Outcome-Based Predictive Process Monitoring
Buliga, A., Vazifehdoostirani, M., Genga, L., Lu, X., Dijkman, R., Francescomarino, C. D., Ghidini, C., & Reijers, H. A. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-70396-6_21 · Cited by 12
“To uncover truly predictive patterns, we propose to extend the combined strategy leveraging both factual and counterfactual explanations proposed by previous studies [3,6] to account for control-flow alterations.”
BackgroundMethodTransparent Latent Space Counterfactual Explanations for Tabular Data
Bodria, F., Guidotti, R., Giannotti, F., & Pedreschi, D. (2022) · IEEE International Conference on Data Science and Advanced Analytics (DSAA) · doi:10.1109/dsaa54385.2022.10032407 · Cited by 12
“Other approaches involves genetic algorithms [19], [20], Shapley values [21], or regression searches [22].”
MethodSelf-citationTowards Responsible AI: Developing Explanations to Increase Human-AI Collaboration
Duarte, R. D. B. (2023) · Frontiers in artificial intelligence and applications · doi:10.3233/faia230126 · Cited by 12
Real-Time Counterfactual Explanations For Robotic Systems With Multiple Continuous Outputs*
Gjærum, V. B., Strümke, I., Lekkas, A. M., & Miller, T. (2023) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2023.10.1328 · Cited by 12
Explainability and Interpretability Concepts for Edge AI Systems
Vermesan, O., Piuri, V., Scotti, F., Genovese, A., Labati, R. D., & Coscia, P. (2024) · River Publishers eBooks · doi:10.1201/9781003478713-9 · Cited by 12
Predicting and Explaining Privacy Risk Exposure in Mobility Data
Naretto, F., Pellungrini, R., Monreale, A., Nardini, F. M., & Musolesi, M. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-61527-7_27 · Cited by 12
“Regarding the explanation task in our experiments, we employed Lore [11] and Shap [13].”
BackgroundMethodInfluentialSelf-citationAlterfactual Explanations -- The Relevance of Irrelevance for Explaining AI SystemsPreprint
Mertes, S., Karle, C., Huber, T., Weitz, K., Schlagowski, R., & André, E. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2207.09374 · Cited by 12
Tempo: Helping Data Scientists and Domain Experts Collaboratively Specify Predictive Modeling Tasks
Sivaraman, V., Vaishampayan, A., Li, X., Buck, B. R., Ma, Z., Boyce, R. D., & Perer, A. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713664 · Cited by 11
“However, by allowing us to gain insight into the predictability of readmission given the available data, we were able to quickly pivot away from less promising specifications and spend more time exploring alternatives (addressing 3.1).”
InfluentialXAI-Eval: A framework for comparative evaluation of explanation methods in healthcare
Agrawal, K., Shawi, R. E., & Ahmed, N. M. O. S. (2025) · Digital Health · doi:10.1177/20552076251368045 · Cited by 11
“34 In this work, for methods entailing the development of a surrogate model g to emulate b , fi delity is assessed by comparing the predictions of b and g using the instances employed in training g .”
Integrating human knowledge for explainable AI
Cappuccio, E., Kathirgamanathan, B., Rinzivillo, S., Andrienko, G., & Andrienko, N. V. (2025) · Machine Learning · doi:10.1007/s10994-025-06879-x · Cited by 11
“Similarly to LORE , LIME could also benefit from a custom neighbourhood generation process that integrates domain knowledge.”
InfluentialExplaining 3D Semantic Segmentation Through Generative AI ‐Based Counterfactuals
Rozajac, D., Lukač, N., Schweng, S., Gollob, C., Nothdurft, A., Stampfer, K., Ser, J. D., & Holzinger, A. (2025) · Expert Systems · doi:10.1111/exsy.70163 · Cited by 11
“They were first formalised in causal inference (Galles and Pearl 1998) and later extended to interpretable ML (Guidotti et al. 2019; Keane and Smyth 2020).”
Lin, T., Lau, R. Y. K., & Hu, S. (2026) · Artificial Intelligence Review · doi:10.1007/s10462-026-11565-y · Cited by 11
“In contrast, the coun - terfactual interpretation reveals how minimal and targeted changes to the input would alter the model’s decision, using techniques such as gradient-based perturbations (Wachter et al. 2017 ) or rule-based neighborhood search (e.g., LORE Guidotti et al. 2019 ).”
Truthful meta-explanations for local interpretability of machine learning models
Mollas, I., Bassiliades, N., & Tsoumakas, G. (2023) · Applied Intelligence · doi:10.1007/s10489-023-04944-3 · Cited by 11
“Numerous measures have been proposed to assess the quality of an interpretation, including fidelity [26], coverage [27], and stability [28].”
BackgroundIntroducing User Feedback-Based Counterfactual Explanations (UFCE)
Suffian, M., Alonso, J. M., & Bogliolo, A. (2024) · International Journal of Computational Intelligence Systems · doi:10.1007/s44196-024-00508-6 · Cited by 11
“As an alternative, the rule-based search uses a set of predefined rules to guide the search for perturbations, but this approach requires comprehensive domain knowledge to define the rules [30].”
Kočić, V., Lukač, N., Rozajac, D., Schweng, S., Gollob, C., Nothdurft, A., Stampfer, K., Ser, J. D., & Holzinger, A. (2025) · IEEE Access · doi:10.1109/access.2025.3568052 · Cited by 11
“Researchers are increasingly turning towards counterfactual methods to address these issues, and counterfactual explanations (CFEs) have emerged as a promising solution [17], [18], [19], [20].”
RACCER: Towards Reachable and Certain Counterfactual Explanations for Reinforcement Learning
Gajcin, J., & Dusparić, I. (2024) · Adaptive Agents and Multi-Agents Systems · doi:10.65109/lxnh7393 · Cited by 11
Mouakher, A., & Kononov, R. (2024) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-024-02421-y · Cited by 10
Interpreting Machine and Deep Learning Models for PDF Malware Detection using XAI and SHAP Framework
Rahman, T., Ahmed, N., Monjur, S., Haque, F. M., & Hossain, M. I. (2023) · International Conference for Innovation in Technology (INOCON) · doi:10.1109/inocon57975.2023.10101116 · Cited by 10
“such as LIME, GRAD-CAM, and LORE,[11] , [6], [13] have”
MethodHandling Missing Values in Local Post-hoc Explainability
Cinquini, M., Giannotti, F., Guidotti, R., & Mattei, A. (2023) · Communications in computer and information science · doi:10.1007/978-3-031-44067-0_14 · Cited by 10
Self-citationFASTER-CE: Fast, Sparse, Transparent, and Robust Counterfactual Explanations
Sharma, S. S. K., Gee, A. H., Henderson, J., & Ghosh, J. (2024) · IFIP advances in information and communication technology · doi:10.1007/978-3-031-63223-5_14 · Cited by 10
How robust are ensemble machine learning explanations?
Calzarossa, M. C., Giudici, P., & Zieni, R. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.129686 · Cited by 10
Balbino, M. D. S., Zárate, L. E., & Nobre, C. N. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.120373 · Cited by 10
MAIRE - A Model-Agnostic Interpretable Rule Extraction Procedure for Explaining Classifiers
Sharma, R., Reddy, N., Kamakshi, V., Krishnan, N. C., & Jain, S. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-84060-0_21 · Cited by 10
Pradeep, P., Caro-Martínez, M., & Wijekoon, A. (2025) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2025.3609825 · Cited by 9
“Factual explanation [38] clarify why a specific prediction was made by offeringsupportingevidence,suchasanapprovedloanduetothe applicant’sincomeandcreditscoremeetingstandards.”
Explainable for Trustworthy AI
Giannotti, F., Naretto, F., & Bodria, F. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-24349-3_10 · Cited by 9
Self-citationVery Fast, Approximate Counterfactual Explanations for Decision Forests
Carreira-Perpiñán, M. Á., & Hada, S. S. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i6.25848 · Cited by 9
Post-hoc Rule Based Explanations for Black Box Bayesian Optimization
Chakraborty, T., Wirth, C., & Seifert, C. (2024) · Communications in computer and information science · doi:10.1007/978-3-031-50396-2_18 · Cited by 9
An eXplainable Artificial Intelligence Methodology on Big Data Architecture
Gatta, V. L., Moscato, V., Postiglione, M., & Sperlí, G. (2024) · Cognitive Computation · doi:10.1007/s12559-024-10272-6 · Cited by 9
“Additionally, other approaches like those in Guidotti et al. [35] and Grover et al. [36] employ global information to generate local explanations, including both supporting and counterfactualrules.”
Explainable paper classification system using topic modeling and SHAP
Shin, N., Lee, Y., Moon, H.-S., Kim, J., & Jung, H. (2024) · Intelligent Data Analysis · doi:10.3233/ida-240075 · Cited by 9
“A lower fidelity error indicates a higher accuracy in the explanations provided by SHAP values in approximating the model’s predictions [93,94].”
Fasan, M. (2022) · DPCE Online · Cited by 9
Counterfactual rule generation for fuzzy rule-based classification systems
Zhang, T., Wagner, C., & Garibaldi, J. M. (2022) · IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) · doi:10.1109/fuzz-ieee55066.2022.9882705 · Cited by 9
“Although CF explanations have a long history in some areas such as philosophy and social sciences [19], they are a fairly recent topic in the XAI area, attracting much attention only in recent years [20], [21].”
BackgroundArtificial Intelligence in Visual Analytics
Nazemi, K. (2023) · International Conference on Information Visualisation (IV) · doi:10.1109/iv60283.2023.00048 · Cited by 9
“Therefore, an interpretable model can be utilized to capture this boundary effectively [31, 32].”
BackgroundMethodModelling Agent Policies with Interpretable Imitation Learning
Bewley, T., Jonathan Lawry, B., & Richards, A. G. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-73959-1_16 · Cited by 9
Local Cascade Ensemble for Multivariate Data Classification.Preprint
Fauvel, K., Fromont, É., Masson, V., Faverdin, P., & Termier, A. (2020) · arXiv (Cornell University) · Cited by 9
Lee, M. H., Choo, S., & Thilarajah, S. D. O. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2405.16424 · Cited by 9
“A counterfactual explanation describes how input features should be changed to update an AI output [25, 45, 66].”
BackgroundArgumentation-Based Explainability for Legal AI: Comparative and Regulatory PerspectivesPreprint
Prajescu, A. I., & Confalonieri, R. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.11079 · Cited by 9
“It generates explanations consisting of a decision rule, which clarifies the reasoning behind a decision, and a set of counterfactual rules, which suggest modifications to the input features that would lead to a different outcome [37].”
Rodrigo, E. G., Alfaro, J. C., Aledo, J. A., & Gámez, J. A. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.111882 · Cited by 8
Metadata
- Authors
- Guidotti, R., Monreale, A., Giannotti, F., Pedreschi, D., Ruggieri, S., & Turini, F.
- Affiliations
- Istituto di Scienza e Tecnologie dell'Informazione "Alessandro Faedo" (IT)
- University of Pisa (IT)
- ORCID
- Venue
- IEEE Intelligent Systems
- Published
- 2019-11-01
- Volume
- 34
- Issue
- 6
- Pages
- 14-23
- ISSN
- 1541-1672 (ISSN-L)
- 1941-1294
- OpenAlex
- W2994120362
- Cited by
- 349
- References
- 14
- Citations by year
- 2026: 38 · 2025: 67 · 2024: 62 · 2023: 60 · 2022: 47 · 2021: 48 · 2020: 26 · 2019: 1
- FWCI
- 13.92
- Type
- journal article
- Language
- en
- Open access
- green
- ieeexplore.ieee.org
- Licence
- Locations
- IEEE Intelligent Systems (publishedVersion)
- CINECA IRIS Institutial research information system (University of Pisa) (submittedVersion, other-oa)
- ISTI Open Portal (submittedVersion, other-oa)
- Scuola Normale Superiore di Pisa (submittedVersion)
- Funders
- European Commission (INFRAIA-1-2014-2015)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Physical Sciences › Computer Science › Software › Software Testing and Debugging Techniques
- Physical Sciences › Computer Science › Information Systems › Data Mining Algorithms and Applications
- Keywords
- explainable AI, black-box classifiers, counterfactual explanations, genetic algorithm, interpretable classifiers, model transparency, trust in AI
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.