16 Towards a renewed operational ethics

Content

16.0.0.0.1 Tools and operational practices

16.0.0.0.2 Ethical verification table

The introduction of Artificial Intelligence systems into peer review processes requires the development of an operational ethics capable of translating abstract principles into concrete protocols, verifiable practices and standardised procedures.

Unlike traditional “ethical frameworks”, which are predominantly declarative and prescriptive in nature, often limited to the enunciation of abstract values, operational ethics takes on a “performative function”, as it binds the actors involved to daily practices that give effect to the principles of transparency, traceability and accountability.

This perspective allows us to recognise that ethics is not an external or accessory constraint, but a “constitutive prerequisite” for the epistemic and institutional legitimacy of peer review.

It is clear that the use of algorithmic tools in evaluation processes introduces elements of opacity, automation and potential disempowerment which, if unregulated, risk compromising the regulatory function of peer review.

Only an ethical framework translated into concrete practices can ensure that technological innovation does not undermine but rather strengthens science’s ability to subject its knowledge to intersubjective scrutiny based on shared criteria.

Operational ethics is inherently dynamic and multi-layered.
It is “dynamic” because it requires continuous adaptation to technological changes in AI tools and to transformations in the contexts in which they operate.
It is “multi-layered” because it does not end with the individual responsibility of reviewers, but extends to the collective dimension of institutional governance, which involves publishers, research institutions and academic communities.

Mandatory disclosure, supervisory mechanisms and audit systems are the first operational manifestations of this requirement.
The aim is not only to prevent abuse or methodological opacity, but also to promote a culture of shared responsibility, capable of maintaining and strengthening collective trust in peer review as a fundamental device for epistemic regulation.

16.1 1. Tools and operational practices

In order to define the framework for a renewed operational ethics, there are certain elements that are particularly important and deserve to be considered as methodological and applicative pillars.

Mandatory disclosure procedures: reviewers and publishers must clearly report the use of AI systems, specifying the stage of use and distinguishing between human and algorithmic contributions.
Ethical compliance checklist: before validating their judgement, reviewers must answer standardised questions that guide them in assessing bias, proportionality and responsibility.
Regular training for reviewers: editorial committees must establish ongoing training programmes to ensure understanding of the limitations of AI systems and critical supervision techniques.
Editorial codes of conduct: journals should include sections dedicated to the use of AI in their guidelines, defining limits, documentation requirements and accountability criteria.
Institutional audit and control systems: supervisory committees should randomly check decisions supported by AI to ensure reliability and effective correction.
Multi-level documentation: logs, methodological notes and metadata must be attached to reports to ensure the traceability of decisions and the replicability of processes.

16.2 2. Ethical verification table

Tables (see example below) can be used as an operational self-audit tool for reviewers and publishing institutions, translating ethical principles into practical questions and mandatory actions.

Each domain corresponds to a “critical point in the evaluation process”, from initial transparency to final institutional control.
👉 In this way, ethics does not remain an abstract reference, but becomes an integral part of daily review practice, promoting a balance between automation and human responsibility.

Domain	Verification Questions	Required Actions
Transparency	Has the use of AI been explicitly declared in the review report?	Include a methodological note distinguishing human contribution from algorithmic input.
Bias and Limitations	Have potential biases in the data or models used been considered?	Document any critical issues and specify their impact on the evaluation.
Proportionality	Has the intervention of AI been limited to support tasks rather than replacing human judgment?	Specify the role of AI and ensure that the final decision remains with the reviewer.
Accountability	Who assumes ultimate responsibility for the judgment (reviewer, editor, committee)?	Indicate in the report the figure responsible for the final opinion.
Documentation	Have logs, methodological notes, and metadata related to the use of AI been produced?	Attach supporting materials to ensure traceability and replicability.
Training	Does the reviewer possess the minimum competencies required to understand the functioning and limitations of the AI employed?	Participate in training courses or consult editorial guidelines.
Institutional Oversight	Is a random verification by the editorial board foreseen?	Include the review in a periodic compliance audit.

A renewed operating ethics should not be seen as an “additional constraint”, but as a safeguard that allows AI systems to be integrated into peer review in a sustainable manner.

Only the combination of disclosure practices, continuous training, institutional audits and multi-level documentation can transform AI from a potential risk into a resource for consolidating scientific quality.

16.2.1 Further Readings

See Ethics Guidelines for Trustworthy
See An Overview of Artificial Intelligence Ethics
See Artificial Intelligence, Humanistic Ethics