CoChat

A Review of Trustworthy and Explainable Artificial Intelligence (XAI)

Open access

The DOI matches its Crossref record; 113 of the 122 references it cites were found in a registry or an index; 52 of the 152 sources its claims cite were checked for support.

Abstract

The advancement of Artificial Intelligence (AI) technology has accelerated the development of several systems that are elicited from it. This boom has made the systems vulnerable to security attacks and allows considerable bias in order to handle errors in the system. This puts humans at risk and leaves machines, robots, and data defenseless. Trustworthy AI (TAI) guarantees human value and the environment. In this paper, we present a comprehensive review of the state-of-the-art on how to build a Trustworthy and eXplainable AI, taking into account that AI is a black box with little insight into its underlying structure. The paper also discusses various TAI components, their corresponding bias, and inclinations that make the system unreliable. The study also discusses the necessity for TAI in many verticals, including banking, healthcare, autonomous system, and IoT. We unite the ways of building trust in all fragmented areas of data protection, pricing, expense, reliability, assurance, and decision-making processes utilizing TAI in several diverse industries and to differing degrees. It also emphasizes the importance of transparent and post hoc explanation models in the construction of an eXplainable AI and lists the potential drawbacks and pitfalls of building eXplainable AI. Finally, the policies for developing TAI in the autonomous vehicle construction sectors are thoroughly examined and eclectic ways of building a reliable, interpretable, eXplainable, and Trustworthy AI sys…

Overview

CoChat Summary

TL;DR

This review surveys trustworthy AI (TAI) and explainable AI (XAI), covering the European Commission’s seven TAI criteria, XAI methods, and applications in healthcare, banking/commerce, IoT, and autonomous systems. It argues that trustworthiness and explainability are increasingly necessary as AI systems become more autonomous and opaque, while identifying unresolved technical and governance challenges.

Key findings

  • The paper reports that more than 70% of individuals in a cited University of Queensland study had neutral or no confidence in AI systems, leaving the total proportion of trustworthiness in AI systems at just a quarter; the public was more trusting and supportive of AI use in healthcare.
  • It cites AI investment growth from $37.5 billion in 2019 to $97.9 billion in 2023, and a worldwide AI software market value of over $126 billion by 2025; in healthcare, AI is estimated to potentially address 20% of all clinical needs.
  • For autonomous vehicles, the paper cites an automotive intelligent hardware/technology/operations market expanding from $1.25 billion in 2017 to $28.5 billion by 2025, and Intel’s analysis that automated cars could save 250 million hours of commuters’ annual travel time and more than 500,000 lives in the USA between 2035–2045.
  • The review organizes TAI around seven EC criteria—human agency and oversight, robustness and safety, privacy and data governance, transparency, diversity and fairness, societal and environmental well-being, and accountability—and divides XAI work into transparency design and post-hoc explanation, with global and local methods.

Methods

  • Design: a comprehensive narrative review/survey of TAI and XAI, not an empirical study; it synthesizes prior literature, regulations, taxonomies, and sector-specific examples.
  • Data/sample: no primary dataset or experimental sample is reported; the paper draws on cited studies, EC/GDPR materials, and application examples in healthcare, banking/commerce, IoT, and autonomous systems.
  • Main analysis: conceptual and taxonomic synthesis, including the seven TAI criteria, XAI categories such as transparent design versus post-hoc explanation, global versus local methods, and tables summarizing XAI research for autonomous systems.

Limitations

  • The paper states that significant constraints remain: replicability of post-hoc explainability methodologies, absence of a unified understanding/criteria/metrics for interpretability, friction between efficiency and predictability, and limitations in explaining deep neural networks.
  • Evident from the text: as a review, it does not report a systematic search strategy, inclusion/exclusion criteria, or quantitative meta-analysis, and it presents no primary empirical validation of the reviewed TAI/XAI approaches.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/access.2023.3294569

Cites

113 of 113 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

29 Valid · 13 Not valid · 59 Not checked · 38 Own results and interpretations

28 more claims · 12 with their cited sources checked

Create a free account to see all 139 claims and to verify the 22 references still open.

See all claims — free account
  • Probably not a substantial claim. 101
  • It may state two separate claims as one. 36

Create a free account to see these claims.

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?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

5 of 7 · the rest aren’t loaded

2 more findings

Create a free account to see every finding of this check.

See all findings — free account

Cited works that were retracted

Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.

1 of 1

Citations

Cites

1 retracted

Research works printed but not linked: 5

  1. hoos, 2018, Automated artificial intelligence (AutoAI)
  2. 2019, What is GDPR the EU’s new data protection law?
  3. 2023, GDPR Articles and Recitals
  4. dastani, 2021, Responsibility Research for Trustworthy Autonomous Systems (Blue Sky Ideas Track)
  5. stierle, 2021, Bringing light into the darkness—A systematic literature review on explainable predictive business process monitoring techniques

Cited by

See all 352

Metadata

Authors
Chamola, V., Hassija, V., Sulthana, A. R., Ghosh, D., Dhingra, D., & Sikdar, B.
Affiliations
  • Birla Institute of Technology and Science, Pilani (IN)
  • National University of Singapore (SG)
  • University of Greenwich (GB)
  • Jaypee Institute of Information Technology (IN)
Venue
IEEE Access
Published
2023
Volume
11
Pages
78994-79015
ISSN
2169-3536 (ISSN-L)
OpenAlex
W4385059984
Cited by
352
References
122
Citations by year
2026: 127 · 2025: 143 · 2024: 76 · 2023: 6
FWCI
33.09
Type
journal article
Language
en
Open access
Locations
  • IEEE Access (publishedVersion, cc-by-nc-nd)
  • Greenwich Academic Literature Archive (University of Greenwich) (acceptedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Ministry of Education, Singapore (A-0009040-00-00, A-0009040-01-00)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • 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
Keywords
trustworthy AI, explainable AI, algorithmic bias, black-box models, post-hoc explanation, interpretability, autonomous vehicles, AI security, artificial intelligence in banking, artificial intelligence in healthcare, Internet of Things, AI policy

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

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.