Omission and commission errors underlying AI failures
Open accessThe DOI matches its Crossref record, with a small difference in title or year; 7 of the 7 references it cites were found in a registry or an index.
Abstract
In this article we investigate origins of several cases of failure of Artificial Intelligence (AI) systems employing machine learning and deep learning. We focus on omission and commission errors in (a) the inputs to the AI system, (b) the processing logic, and (c) the outputs from the AI system. Our framework yields a set of 28 factors that can be used for reconstructing the path of AI failures and for determining corrective action. Our research helps identify emerging themes of inquiry necessary for developing more robust AI-ML systems. We are hopeful that our work will help strengthen the use of machine-learning AI by enhancing the rates of true positive and true negative judgements from AI systems, and by lowering the probabilities of false positive and false negative judgements.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
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?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Hence, the AI platforms can commit errors due to extracting details from unauthenticated sources”
From Artificial intelligence and medical education: application in classroom instruction and student assessment using a pharmacology & therapeutics case study (2024) · cites this paper as [22]
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.
Citations
Cites
Judgment under Uncertainty: Heuristics and Biases
Tversky, A., & Kahneman, D. (1975) · Utility, Probability, and Human Decision Making · doi:10.1007/978-94-010-1834-0_8 · Cited by 22,650
The Unanticipated Consequences of Purposive Social Action
Merton, R. K. (1936) · American Sociological Review · doi:10.2307/2084615 · Cited by 2,520
The accuracy, fairness, and limits of predicting recidivism
Dressel, J., & Farid, H. (2018) · Science Advances · doi:10.1126/sciadv.aao5580 · Cited by 1,071 · cited 1×
“Source: Dressel & Farid (2018: 1).”
SCHIER, F. (1984) · Philosophical Books · doi:10.1111/j.1468-0149.1984.tb01924.x · Cited by 266
Predicting future AI failures from historic examples
Yampolskiy, R. V. (2019) · foresight · doi:10.1108/fs-04-2018-0034 · Cited by 75
Ackoff, R. L. (1994) · Systemic Practice and Action Research · doi:10.1007/bf02169161 · Cited by 10
The two biggest NQT challenges
Stanley, J. C. (2017) · SecEd · doi:10.12968/sece.2017.33.10a · Cited by 8
Research works printed but not linked: 3
- Babic B, Cohen IG, Evgeniou T et al (2021) When machine learning goes off the rails. Harv Bus Rev 99(1):76–84. https://psnet.ahrq.gov/issue/when-machine-learning-goes-rails-guide-managing-risks
- Cai F, Yuan Y (2021) 2020 in review: 10 AI failures. Synced Review, (January 1) Editor: Michael Sarazen https://syncedreview.com/2021/01/01/2020-in-review-10-ai-failures/. Accessed 24 Mar 2021
- Luca M, Kleinberg J, Mullainathan S (2016) Algorithms need managers, too. Harv Bus Rev 96(1):96–101. https://sendhil.org/algorithms-need-managers-too/
Cited by
Sridharan, K., & Sequeira, R. P. (2024) · BMC Medical Education · doi:10.1186/s12909-024-05365-7 · Cited by 114
“Hence, the AI platforms can commit errors due to extracting details from unauthenticated sources [22] created a framework identifying 28 factors for reconstructing the path of AI failures and for determining corrective actions.”
BackgroundNot My Voice! A Taxonomy of Ethical and Safety Harms of Speech Generators
Hutiri, W. T., Papakyriakopoulos, O., & Xiang, A. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3658911 · Cited by 113
Metacognitive sensitivity: The key to calibrating trust and optimal decision making with AI
Lee, D., Pruitt, J., Zhou, T., Du, J., & Odegaard, B. (2025) · PNAS Nexus · doi:10.1093/pnasnexus/pgaf133 · Cited by 44
Product liability for defective AI
Buiten, M. C. (2024) · European Journal of Law and Economics · doi:10.1007/s10657-024-09794-z · Cited by 41
“AI systems can fail if there are problems with its inputs, such as the data or the sensor hardware, if the processing logic is deficient in some way, or if the repertoire of actions available to the AI system is inadequate, resulting in an inappropriate output (Chanda & Banerjee, 2022).”
Innovative deep learning techniques for monitoring aggressive behavior in social media posts
Han, H., Asif, M., Awwad, E. M., Sarhan, N. M., Ghadi, Y. Y., & Xu, B. (2024) · Journal of Cloud Computing Advances Systems and Applications · doi:10.1186/s13677-023-00577-6 · Cited by 38
Metadata
- Authors
- Chanda, S. S., & Banerjee, D. N.
- Affiliations
- Indian Institute of Management Indore (IN)
- Corresponding author
- Chanda, Sasanka Sekhar
- Venue
- AI & SOCIETY
- Published
- 2022-11-17
- Volume
- 39
- Issue
- 3
- Pages
- 937-960
- ISSN
- 0951-5666 (ISSN-L)
- 1435-5655
- PMC
- PMC9669536
- OpenAlex
- W4309637331
- Cited by
- 45
- References
- 7
- Citations by year
- 2026: 11 · 2025: 21 · 2024: 13
- FWCI
- 9.86
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- public-domain
- www.springernature.com/gp/researchers/text-and-data-mining (tdm)
- www.springernature.com/gp/researchers/text-and-data-mining (vor)
- Locations
- AI & Society (publishedVersion, public-domain)
- PubMed Central (publishedVersion)
- PubMed (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Physical Sciences › Engineering › Safety, Risk, Reliability and Quality › Safety Systems Engineering in Autonomy
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence Applications
- Keywords
- AI failure, omission errors, commission errors, machine learning, deep learning, error analysis, AI robustness, false positives, false-negative results, corrective actions
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.