---
OA_place: repository
OA_type: green
_id: '22103'
abstract:
- lang: eng
  text: "Modern AI systems increasingly rely on opaque, highly complex models whose
    inner workings remain inaccessible even to experts. This opacity creates challenges
    for trust, accountability, and compliance with\r\nemerging regulatory expectations
    such as the “right to an explanation”. While traditional explainability methods—feature
    attributions, counterfactuals, surrogate models—and interpretable model classes
    provide valuable insights for engineers, they often fall short of delivering the
    contextual, conversational explanations that\r\nreal users expect. Large Language
    Models (LLMs) offer a promising new avenue for explanation due to their\r\nability
    to engage interactively, adapt to user needs, and translate technical outputs
    into more accessible reasoning. However, their tendencies toward hallucination,
    conflict avoidance, and oversimplification introduce\r\nserious risks when used
    as explanatory agents. This paper analyzes these opportunities and limitations,
    examines verification strategies for ensuring explanation fidelity, and situates
    LLM-generated explanations within\r\nbroader concerns about public trust. The
    paper concludes by outlining best practices and future research directions for
    building robust, verifiable, and human-aligned explanation systems."
acknowledgement: "This work has been supported by the European Research Council under
  Grant No.: ERC-2020-AdG\r\n101020093. LLM–based tools have been used as\r\nwriting
  assistance to help improve presentation.\r\n"
article_processing_charge: No
author:
- first_name: Filip
  full_name: Cano Cordoba, Filip
  id: 708cad98-e86a-11ef-8098-bdae2d7c6af1
  last_name: Cano Cordoba
  orcid: 0000-0002-0783-904X
citation:
  ama: 'Cano Cordoba F. Explaining decisions one conversation at a time: Opportunities
    and risks of LLMs as explainability assistants. In: <i>Proceedings of the 18th
    International Conference on Agents and Artificial Intelligence</i>. Vol 5. Science
    and Technology Publications; 2026:4689-4696. doi:<a href="https://doi.org/10.5220/0014483200004052">10.5220/0014483200004052</a>'
  apa: 'Cano Cordoba, F. (2026). Explaining decisions one conversation at a time:
    Opportunities and risks of LLMs as explainability assistants. In <i>Proceedings
    of the 18th International Conference on Agents and Artificial Intelligence</i>
    (Vol. 5, pp. 4689–4696). Marbella, Spain: Science and Technology Publications.
    <a href="https://doi.org/10.5220/0014483200004052">https://doi.org/10.5220/0014483200004052</a>'
  chicago: 'Cano Cordoba, Filip. “Explaining Decisions One Conversation at a Time:
    Opportunities and Risks of LLMs as Explainability Assistants.” In <i>Proceedings
    of the 18th International Conference on Agents and Artificial Intelligence</i>,
    5:4689–96. Science and Technology Publications, 2026. <a href="https://doi.org/10.5220/0014483200004052">https://doi.org/10.5220/0014483200004052</a>.'
  ieee: 'F. Cano Cordoba, “Explaining decisions one conversation at a time: Opportunities
    and risks of LLMs as explainability assistants,” in <i>Proceedings of the 18th
    International Conference on Agents and Artificial Intelligence</i>, Marbella,
    Spain, 2026, vol. 5, pp. 4689–4696.'
  ista: 'Cano Cordoba F. 2026. Explaining decisions one conversation at a time: Opportunities
    and risks of LLMs as explainability assistants. Proceedings of the 18th International
    Conference on Agents and Artificial Intelligence. ICAART: International Conference
    on Agents and Artificial Intelligence vol. 5, 4689–4696.'
  mla: 'Cano Cordoba, Filip. “Explaining Decisions One Conversation at a Time: Opportunities
    and Risks of LLMs as Explainability Assistants.” <i>Proceedings of the 18th International
    Conference on Agents and Artificial Intelligence</i>, vol. 5, Science and Technology
    Publications, 2026, pp. 4689–96, doi:<a href="https://doi.org/10.5220/0014483200004052">10.5220/0014483200004052</a>.'
  short: F. Cano Cordoba, in:, Proceedings of the 18th International Conference on
    Agents and Artificial Intelligence, Science and Technology Publications, 2026,
    pp. 4689–4696.
conference:
  end_date: 2026-03-08
  location: Marbella, Spain
  name: 'ICAART: International Conference on Agents and Artificial Intelligence'
  start_date: 2026-03-05
corr_author: '1'
das_tickbox: '0'
date_created: 2026-06-21T22:03:00Z
date_published: 2026-04-01T00:00:00Z
date_updated: 2026-06-24T08:37:00Z
day: '01'
department:
- _id: ToHe
doi: 10.5220/0014483200004052
ec_funded: 1
intvolume: '         5'
keyword:
- Explainable AI
- Large Language Models
- Trust in AI
language:
- iso: eng
main_file_link:
- open_access: '1'
  url: https://filipcano.org/files/icaart26llm.pdf
month: '04'
oa: 1
oa_version: Accepted Version
page: 4689-4696
project:
- _id: 62781420-2b32-11ec-9570-8d9b63373d4d
  call_identifier: H2020
  grant_number: '101020093'
  name: Vigilant Algorithmic Monitoring of Software
publication: Proceedings of the 18th International Conference on Agents and Artificial
  Intelligence
publication_identifier:
  eissn:
  - 2184-433X
  isbn:
  - '9789897587962'
  issn:
  - 2184-3589
publication_status: published
publisher: Science and Technology Publications
quality_controlled: '1'
researchdata_availability: no
scopus_import: '1'
status: public
supplementarymaterial: no
title: 'Explaining decisions one conversation at a time: Opportunities and risks of
  LLMs as explainability assistants'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 5
year: '2026'
...
---
OA_place: repository
OA_type: green
_id: '21090'
abstract:
- lang: eng
  text: Fairness in AI is traditionally studied as a static property evaluated once,
    over a fixed dataset. However, real-world AI systems operate sequentially, with
    outcomes and environments evolving over time. This paper proposes a framework
    for analysing fairness as a runtime property. Using a minimal yet expressive model
    based on sequences of coin tosses with possibly evolving biases, we study the
    problems of monitoring and enforcing fairness expressed in either toss outcomes
    or coin biases. Since there is no one-size-fits-all solution for either problem,
    we provide a summary of monitoring and enforcement strategies, parametrised by
    environment dynamics, prediction horizon, and confidence thresholds. For both
    problems, we present general results under simple or minimal assumptions. We survey
    existing solutions for the monitoring problem for Markovian and additive dynamics,
    and existing solutions for the enforcement problem in static settings with known
    dynamics.
acknowledgement: 'This work is supported by the European Research Council under Grant
  No.: ERC-2020-AdG 101020093.'
alternative_title:
- LNCS
article_processing_charge: No
arxiv: 1
author:
- first_name: Filip
  full_name: Cano Cordoba, Filip
  id: 708cad98-e86a-11ef-8098-bdae2d7c6af1
  last_name: Cano Cordoba
  orcid: 0000-0002-0783-904X
- first_name: Thomas A
  full_name: Henzinger, Thomas A
  id: 40876CD8-F248-11E8-B48F-1D18A9856A87
  last_name: Henzinger
  orcid: 0000-0002-2985-7724
- first_name: Konstantin
  full_name: Kueffner, Konstantin
  id: 8121a2d0-dc85-11ea-9058-af578f3b4515
  last_name: Kueffner
  orcid: 0000-0001-8974-2542
citation:
  ama: 'Cano Cordoba F, Henzinger TA, Kueffner K. Algorithmic fairness: A runtime
    perspective. In: <i>25th International Conference on Runtime Verification</i>.
    Vol 16087. Springer Nature; 2025:1-21. doi:<a href="https://doi.org/10.1007/978-3-032-05435-7_1">10.1007/978-3-032-05435-7_1</a>'
  apa: 'Cano Cordoba, F., Henzinger, T. A., &#38; Kueffner, K. (2025). Algorithmic
    fairness: A runtime perspective. In <i>25th International Conference on Runtime
    Verification</i> (Vol. 16087, pp. 1–21). Graz, Austria: Springer Nature. <a href="https://doi.org/10.1007/978-3-032-05435-7_1">https://doi.org/10.1007/978-3-032-05435-7_1</a>'
  chicago: 'Cano Cordoba, Filip, Thomas A Henzinger, and Konstantin Kueffner. “Algorithmic
    Fairness: A Runtime Perspective.” In <i>25th International Conference on Runtime
    Verification</i>, 16087:1–21. Springer Nature, 2025. <a href="https://doi.org/10.1007/978-3-032-05435-7_1">https://doi.org/10.1007/978-3-032-05435-7_1</a>.'
  ieee: 'F. Cano Cordoba, T. A. Henzinger, and K. Kueffner, “Algorithmic fairness:
    A runtime perspective,” in <i>25th International Conference on Runtime Verification</i>,
    Graz, Austria, 2025, vol. 16087, pp. 1–21.'
  ista: 'Cano Cordoba F, Henzinger TA, Kueffner K. 2025. Algorithmic fairness: A runtime
    perspective. 25th International Conference on Runtime Verification. RV: Runtime
    Verification, LNCS, vol. 16087, 1–21.'
  mla: 'Cano Cordoba, Filip, et al. “Algorithmic Fairness: A Runtime Perspective.”
    <i>25th International Conference on Runtime Verification</i>, vol. 16087, Springer
    Nature, 2025, pp. 1–21, doi:<a href="https://doi.org/10.1007/978-3-032-05435-7_1">10.1007/978-3-032-05435-7_1</a>.'
  short: F. Cano Cordoba, T.A. Henzinger, K. Kueffner, in:, 25th International Conference
    on Runtime Verification, Springer Nature, 2025, pp. 1–21.
conference:
  end_date: 2025-09-19
  location: Graz, Austria
  name: 'RV: Runtime Verification'
  start_date: 2025-09-15
corr_author: '1'
date_created: 2026-01-29T16:01:41Z
date_published: 2025-09-13T00:00:00Z
date_updated: 2026-02-16T11:57:00Z
day: '13'
department:
- _id: ToHe
doi: 10.1007/978-3-032-05435-7_1
ec_funded: 1
external_id:
  arxiv:
  - '2507.20711'
intvolume: '     16087'
language:
- iso: eng
main_file_link:
- open_access: '1'
  url: https://doi.org/10.48550/arXiv.2507.20711
month: '09'
oa: 1
oa_version: Preprint
page: 1-21
project:
- _id: 62781420-2b32-11ec-9570-8d9b63373d4d
  call_identifier: H2020
  grant_number: '101020093'
  name: Vigilant Algorithmic Monitoring of Software
publication: 25th International Conference on Runtime Verification
publication_identifier:
  eisbn:
  - '9783032054357'
  eissn:
  - 1611-3349
  issn:
  - 0302-9743
publication_status: published
publisher: Springer Nature
quality_controlled: '1'
status: public
title: 'Algorithmic fairness: A runtime perspective'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 16087
year: '2025'
...
---
OA_place: repository
OA_type: green
_id: '19665'
abstract:
- lang: eng
  text: As AI-based decision-makers increasingly influence human lives, it is a growing
    concern that their decisions may be unfair or biased with respect to people's
    protected attributes, such as gender and race. Most existing bias prevention measures
    provide probabilistic fairness guarantees in the long run, and it is possible
    that the decisions are biased on any decision sequence of fixed length. We introduce
    *fairness shielding*, where a symbolic decision-maker---the fairness shield---continuously
    monitors the sequence of decisions of another deployed black-box decision-maker,
    and makes interventions so that a given fairness criterion is met while the total
    intervention costs are minimized. We present four different algorithms for computing
    fairness shields, among which one guarantees fairness over fixed horizons, and
    three guarantee fairness periodically after fixed intervals. Given a distribution
    over future decisions and their intervention costs, our algorithms solve different
    instances of bounded-horizon optimal control problems with different levels of
    computational costs and optimality guarantees. Our empirical evaluation demonstrates
    the effectiveness of these shields in ensuring fairness while maintaining cost
    efficiency across various scenarios.
acknowledgement: 'This work is partly supported by the European Research Council under
  Grant No.: ERC-2020-AdG 101020093. It is also partially supported by the State Government
  of Styria, Austria – Department Zukunftsfonds Steiermark.'
article_processing_charge: No
arxiv: 1
author:
- first_name: Filip
  full_name: Cano Cordoba, Filip
  id: 708cad98-e86a-11ef-8098-bdae2d7c6af1
  last_name: Cano Cordoba
  orcid: 0000-0002-0783-904X
- first_name: Thomas A
  full_name: Henzinger, Thomas A
  id: 40876CD8-F248-11E8-B48F-1D18A9856A87
  last_name: Henzinger
  orcid: 0000-0002-2985-7724
- first_name: Bettina
  full_name: Könighofer, Bettina
  last_name: Könighofer
- first_name: Konstantin
  full_name: Kueffner, Konstantin
  id: 8121a2d0-dc85-11ea-9058-af578f3b4515
  last_name: Kueffner
  orcid: 0000-0001-8974-2542
- first_name: Kaushik
  full_name: Mallik, Kaushik
  id: 0834ff3c-6d72-11ec-94e0-b5b0a4fb8598
  last_name: Mallik
  orcid: 0000-0001-9864-7475
citation:
  ama: 'Cano Cordoba F, Henzinger TA, Könighofer B, Kueffner K, Mallik K. Fairness
    shields: Safeguarding against biased decision makers. In: <i>Proceedings of the
    39th AAAI Conference on Artificial Intelligence</i>. Vol 39. Association for the
    Advancement of Artificial Intelligence; 2025:15659-15668. doi:<a href="https://doi.org/10.1609/aaai.v39i15.33719">10.1609/aaai.v39i15.33719</a>'
  apa: 'Cano Cordoba, F., Henzinger, T. A., Könighofer, B., Kueffner, K., &#38; Mallik,
    K. (2025). Fairness shields: Safeguarding against biased decision makers. In <i>Proceedings
    of the 39th AAAI Conference on Artificial Intelligence</i> (Vol. 39, pp. 15659–15668).
    Philadelphia, PA, United States: Association for the Advancement of Artificial
    Intelligence. <a href="https://doi.org/10.1609/aaai.v39i15.33719">https://doi.org/10.1609/aaai.v39i15.33719</a>'
  chicago: 'Cano Cordoba, Filip, Thomas A Henzinger, Bettina Könighofer, Konstantin
    Kueffner, and Kaushik Mallik. “Fairness Shields: Safeguarding against Biased Decision
    Makers.” In <i>Proceedings of the 39th AAAI Conference on Artificial Intelligence</i>,
    39:15659–68. Association for the Advancement of Artificial Intelligence, 2025.
    <a href="https://doi.org/10.1609/aaai.v39i15.33719">https://doi.org/10.1609/aaai.v39i15.33719</a>.'
  ieee: 'F. Cano Cordoba, T. A. Henzinger, B. Könighofer, K. Kueffner, and K. Mallik,
    “Fairness shields: Safeguarding against biased decision makers,” in <i>Proceedings
    of the 39th AAAI Conference on Artificial Intelligence</i>, Philadelphia, PA,
    United States, 2025, vol. 39, no. 15, pp. 15659–15668.'
  ista: 'Cano Cordoba F, Henzinger TA, Könighofer B, Kueffner K, Mallik K. 2025. Fairness
    shields: Safeguarding against biased decision makers. Proceedings of the 39th
    AAAI Conference on Artificial Intelligence. AAAI: Conference on Artificial Intelligence
    vol. 39, 15659–15668.'
  mla: 'Cano Cordoba, Filip, et al. “Fairness Shields: Safeguarding against Biased
    Decision Makers.” <i>Proceedings of the 39th AAAI Conference on Artificial Intelligence</i>,
    vol. 39, no. 15, Association for the Advancement of Artificial Intelligence, 2025,
    pp. 15659–68, doi:<a href="https://doi.org/10.1609/aaai.v39i15.33719">10.1609/aaai.v39i15.33719</a>.'
  short: F. Cano Cordoba, T.A. Henzinger, B. Könighofer, K. Kueffner, K. Mallik, in:,
    Proceedings of the 39th AAAI Conference on Artificial Intelligence, Association
    for the Advancement of Artificial Intelligence, 2025, pp. 15659–15668.
conference:
  end_date: 2025-03-04
  location: Philadelphia, PA, United States
  name: 'AAAI: Conference on Artificial Intelligence'
  start_date: 2025-02-25
corr_author: '1'
date_created: 2025-05-11T22:02:39Z
date_published: 2025-04-11T00:00:00Z
date_updated: 2026-02-16T12:24:30Z
day: '11'
department:
- _id: ToHe
doi: 10.1609/aaai.v39i15.33719
ec_funded: 1
external_id:
  arxiv:
  - '2412.11994'
intvolume: '        39'
issue: '15'
language:
- iso: eng
main_file_link:
- open_access: '1'
  url: https://doi.org/10.48550/arXiv.2412.11994
month: '04'
oa: 1
oa_version: Preprint
page: 15659-15668
project:
- _id: 62781420-2b32-11ec-9570-8d9b63373d4d
  call_identifier: H2020
  grant_number: '101020093'
  name: Vigilant Algorithmic Monitoring of Software
publication: Proceedings of the 39th AAAI Conference on Artificial Intelligence
publication_identifier:
  eissn:
  - 2374-3468
  issn:
  - 2159-5399
publication_status: published
publisher: Association for the Advancement of Artificial Intelligence
quality_controlled: '1'
scopus_import: '1'
status: public
title: 'Fairness shields: Safeguarding against biased decision makers'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 39
year: '2025'
...