Towards human-compatible XAI: Explaining data differentials with concept induction over background knowledge
Open accessThe claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Human-compatible, human-centric approaches, AI for social good, and human computation or interactive machine learning are computing paradigms aligned with the human agency and oversight requirement.
“For the preservation of human rights, humancompatible [20], human-centric approaches [14,21,22], AI for social good [7,23] and human computation or interactive machine learning [24] are computing paradigms aligned with this requirement.”
From Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation (2023) · cites this paper as [20]
In this revision of the AI Act, MEPs have also proposed tailored regulatory regimes for new and fast-evolving developments in the field of AI and GPAIS.
“In this revision of the AI Act, MEPs have also proposed tailored regulatory regimes for new and fast-evolving developments in the field of AI and GPAIS.”
From Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation (2023) · cites this paper as [20]
The Introduction (p. 1) discusses explainable AI, and the second-experiment section (p. 4) describes a study of explanations for classifier errors. Searches for the AI Act, MEPs, GPAIS, and tailored regulatory regimes found no passage addressing the claimed legislative proposal. (LLM, conf=0.78)
Experimental — accuracy not yet measuredAdditional transparency requirements have been demanded for generative foundation models such as GPT: they must inform that the content is generated by an AI model, the model must be designed to avoid generating illegal content and publishing summaries or copyrighted content used during training.
“Furthermore, additional transparency requirements have been demanded for generative foundation models such as GPT: they must inform that the content is generated by an AI model, the model must be designed to avoid generating illegal content and publishing summaries or copyrighted content used during training.”
From Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation (2023) · cites this paper as [20]
Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measuredThe revised classification of high-risk scenarios also considered AI systems used to influence voters in political campaigns, as well as recommender systems (with more than 45 million users) utilized by social media platforms.
“Such revised classification of high-risk scenarios also considered AI systems used to influence voters in political campaigns, as well as recommender systems (with more than 45 million users) utilized by social media platforms.”
From Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation (2023) · cites this paper as [20]
The Introduction (p. 1) describes concept induction for explaining machine-learning classifications, and the second-experiment section (p. 4) concerns identifying classification errors. Searches of our copy for political campaigns, voter influence, recommender systems, and the 45-million-user threshold found no passage addressing the claimed high-risk classification. (LLM, conf=0.78)
Experimental — accuracy not yet measuredMEPs expanded the classification of high-risk areas to include those that could compromise or harm people's health, safety, fundamental rights or the environment.
“MEPs expanded the classification of high-risk areas to include those that could compromise or harm people's health, safety, fundamental rights or the environment.”
From Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation (2023) · cites this paper as [20]
The Introduction (p. 1) discusses explainable AI, and the Experiment 2 discussion (p. 8) concerns detecting image-classification errors. Neither addresses MEPs or the classification of high-risk areas. Searches for the quoted claim and for European Parliament high-risk AI criteria found no relevant passage in our copy. (LLM, conf=0.78)
Experimental — accuracy not yet measured
In the case of the European AI Act, on 11th May 2023, MEPs endorsed new transparency and risk-management rules for AI systems.
“In the case of the European AI Act, on 11th May 2023, MEPs endorsed new transparency and risk-management rules for AI systems.”
From Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation (2023) · cites this paper as [20]
The Introduction (p. 1) discusses transparency in machine-learning models, but not the European AI Act. Searches for the 11 May 2023 MEP endorsement and for European Parliament AI rules found no passage addressing the claimed event or its transparency and risk-management rules. (LLM, conf=0.77)
Experimental — accuracy not yet measuredIntrusive and discriminatory uses of AI-based biometric systems have been identified as prohibited AI systems, such as real-time remote biometric identification, biometric categorization, post remote biometric identification, predictive policing, indiscriminate scraping of biometric data, and emotion recognition.
“Intrusive and discriminatory uses of AI-based biometric systems have been also identified as prohibited AI systems, such as:”
From Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation (2023) · cites this paper as [20]
The Introduction (p. 1) discusses explainable AI, and Experiment 2 (p. 4) concerns explanations of image-classification errors. Searches for prohibited biometric systems, remote biometric identification, predictive policing, biometric-data scraping, and emotion recognition found no passage addressing the claimed prohibitions. (LLM, conf=0.76)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Barredo Arrieta, A., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., Garcia, S., Gil-Lopez, S., Molina, D., Benjamins, R., et al. (2020) · Information Fusion · doi:10.1016/j.inffus.2019.12.012 · Cited by 10,423 · see citing works
Definitions, methods, and applications in interpretable machine learning
Murdoch, W. J., Singh, C., Kumbier, K., Abbasi-Asl, R., & Yu, B. (2019) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1900654116 · Cited by 2,167 · see citing works
Can we open the black box of AI?
Castelvecchi, D. (2016) · Nature · doi:10.1038/538020a · Cited by 1,773 · see citing works
Interpretation of Neural Networks Is Fragile
Ghorbani, A., Abid, A., & Zou, J. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33013681 · Cited by 765 · see citing works
Manipulating and Measuring Model Interpretability
Poursabzi-Sangdeh, F., Goldstein, D. G., Hofman, J. M., Wortman Vaughan, J. W., & Wallach, H. (2021) · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445315 · Cited by 728 · see citing works
Flagged first, then by citations
Research works printed but not linked: 27
- Lehmann, 2010, Concept learning, vol. 18, 71
- Rajabi, 2022, Knowledge-graph based explainable AI: A systematic review, J. Inf. Sci., 538
- Sarker, 2020, Wikipedia knowledge graph for explainable AI, vol. 1232, 72
- Lehmann, 2010, Concept learning in description logics using refinement operators, Mach. Learn., 78, 203, 10.1007/s10994-009-5146-2 · DOI
- Bühmann, 2016, DL-Learner – a framework for inductive learning on the semantic web, J. Web Sem., 39, 15, 10.1016/j.websem.2016.06.001 · DOI
- Muggleton, 1991, Inductive logic programming, New Gener. Comput., 8, 295, 10.1007/BF03037089 · DOI · Paper page
- Lehmann, 2010, ORE – a tool for repairing and enriching knowledge bases, vol. 6497, 177
- Paulheim, 2017, Knowledge graph refinement: A survey of approaches and evaluation methods, Semantic Web, 8, 489
- Sarker, 2017, Explaining trained neural networks with semantic web technologies: First steps, vol. 2003
- Lehmann, 2015, DBpedia – A large-scale, multilingual knowledge base extracted from Wikipedia, Semantic Web, 6, 167
- Bradley, 1952, Rank analysis of incomplete block designs: I. The method of paired comparisons, Biometrika, 39, 324
- Zhou, 2019, Semantic understanding of scenes through the ADE20K dataset, Int. J. Comput. Vis., 127, 302, 10.1007/s11263-018-1140-0 · DOI · Paper page
- CArptenter, 2017, Stan: A probabilistic programming language, J. Stat. Softw., 76, 1
- Bürkner, 2021, Bayesian item response modeling in R with brms and stan, J. Stat. Softw., 100, 10.18637/jss.v100.i05 · DOI
- Wagenmakers, 2010, Bayesian hypothesis testing for psychologists: A tutorial on the savage-dickey method, Cogn. Psychol., 60, 158, 10.1016/j.cogpsych.2009.12.001 · DOI · Paper page
- Rouder, 2014, Optional stopping: No problem for Bayesians, Psychon. Bull. & Rev., 21, 301, 10.3758/s13423-014-0595-4 · DOI · Paper page
- Brysbaert, 2014, Concreteness ratings for 40 thousand generally known English word lemmas, Behav. Res. Methods, 46, 904, 10.3758/s13428-013-0403-5 · DOI · Paper page
- Lundberg, 2017, A unified approach to interpreting model predictions, vol. 30, 4765
- Koh, 2017, Understanding black-box predictions via influence functions, vol. 70, 1885
- Nguyen, 2019, Understanding neural networks via feature visualization: A survey, vol. 11700, 55
- Goyal, 2019, Counterfactual visual explanations, vol. 97, 2376
- Kim, 2018, Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (TCAV), vol. 80, 2673
- Frosst, 2017, Distilling a neural network into a soft decision tree, vol. 2071
- Kindermans, 2019, The (un)reliability of saliency methods, vol. 11700, 267
- Confalonieri, 2021, Using ontologies to enhance human understandability of global post-hoc explanations of black-box models, Artificial Intelligence, 296, 10.1016/j.artint.2021.103471 · DOI
- Gacto, 2011, Interpretability of linguistic fuzzy rule-based systems: An overview of interpretability measures, Inform. Sci., 181, 4340, 10.1016/j.ins.2011.02.021 · DOI
- Carvalho, 2019, Machine learning interpretability: A survey on methods and metrics, Electronics, 8, 832, 10.3390/electronics8080832 · DOI · Paper page
Cited by
Sorted by citations
Ali, S., Abuhmed, T., El-Sappagh, S., Muhammad, K., Alonso-Moral, J. M., Confalonieri, R., Guidotti, R., Del Ser, J., Díaz-Rodríguez, N., & Herrera, F. (2023) · Information Fusion · doi:10.1016/j.inffus.2023.101805 · Cited by 1,762 · see citing works
Díaz-Rodríguez, N., Del Ser, J., Coeckelbergh, M., López de Prado, M., Herrera-Viedma, E., & Herrera, F. (2023) · Information Fusion · doi:10.1016/j.inffus.2023.101896 · Cited by 800 · see citing works
- Authors
- Widmer, C. L., Sarker, M. K., Nadella, S., Fiechter, J., Juvina, I., Minnery, B., Hitzler, P., Schwartz, J., & Raymer, M.
- Affiliations
- Kairos (United States) (US)
- Bowie State University (US)
- Wright State University (US)
- Kansas State University (US)
- Corresponding author
- Hitzler, Pascal
- ORCID
- Venue
- Journal of Web Semantics
- Publisher
- Elsevier BV
- Published
- 2023-12
- Volume
- 79
- Pages
- 100807
- ISSN
- 1570-8268 (ISSN-L)
- OpenAlex
- W4387045351
- Cited by
- 10
- References
- 5
- Citations by year
- 2026: 2 · 2025: 2 · 2024: 4 · 2023: 2
- FWCI
- 0.89
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Licence
- www.elsevier.com/tdm/userlicense/1.0/ (tdm)
- www.elsevier.com/legal/tdmrep-license (tdm)
- www.elsevier.com/open-access/userlicense/1.0/ (am, 307 d)
- doi.org/10.15223/policy-017 (stm-asf)
- doi.org/10.15223/policy-037 (stm-asf)
- doi.org/10.15223/policy-012 (stm-asf)
- doi.org/10.15223/policy-029 (stm-asf)
- doi.org/10.15223/policy-004 (stm-asf)
- Locations
- Journal of Web Semantics (acceptedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- DOD (FA8649-20-9-9045)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Physical Sciences › Computer Science › Artificial Intelligence › Semantic Web and Ontologies
- Physical Sciences › Computer Science › Artificial Intelligence › Logic, Reasoning, and Knowledge
- Keywords
- explainable AI, concept induction, description logics, prior knowledge, ontology engineering
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
