Artificial Intelligence Tools Policy
Artificial intelligence tools may support research and manuscript preparation when their use is transparent and checked by a human. This policy defines permitted uses, disclosure requirements, restrictions in peer review, and safeguards for reliability and confidentiality.
Contents
- Scope and responsibility
- Permitted assistance for authors
- What, where and when to disclose
- Checking text, sources and code
- AI as a research method and reproducibility
- Images, diagrams, plots and maps
- Confidentiality and rights in materials
- AI use by reviewers
- Editorial work and production
- Concerns, action and policy review
01. Scope and responsibility
This policy of Irrigation and Melioration applies to authors, reviewers, editors, guest editors and publisher staff. It covers the main text, translations, abstracts, tables, images, code, supplementary materials and editorial correspondence. Requirements depend on what a tool does, rather than its brand name.
AI means artificial intelligence tools, including generative models, language models, image generators, coding assistants and machine learning models. Using AI as an object of study or research method differs from assistance in writing an article and requires a methodological description.
AI cannot be an author or co-author and cannot accept responsibility for publication. Authors are responsible for the final material they have checked, editors for decisions, and reviewers for their own assessments. Attributing an error to a tool does not remove human responsibility. Authorship criteria are set out in the Authorship Principles.
03. What, where and when to disclose
At initial submission, authors inform the editorial office of AI use and include a separate statement, “Use of artificial intelligence tools”, before the references. It remains in the published article. The statement identifies the tool and provider, model or version where available, period of use, specific tasks, affected parts of the work and verification performed by the authors. If the service does not disclose its version, this is stated; details must not be invented.
AI used in data collection or analysis, modelling and production of results is also described in the methods; assistance with diagrams is additionally identified in their captions. The general declaration may refer to the relevant subsections without repeating the methodology. Information is updated after revision and checked before publication. In double-anonymous review, the technical description remains available while author-identifying information is supplied separately to the editorial office.
Example to complete according to actual use: “For [task and part of the work], we used [tool, provider, model/version, period]. The authors checked [sources, calculations, code or other outputs] by [verification method], made the necessary changes and accept responsibility for the final content. AI used in the research methodology is described in [section, if applicable].”
If there was no use requiring disclosure, state: “No AI tools requiring disclosure under the journal’s policy were used.” Authors are not required to disclose a diagnosis or other personal information because they use accessibility aids.
04. Checking text, sources and code
Authors check factual claims, quotations, definitions, formulae, dimensions and numerical values against reliable original sources and their own materials. For every AI-suggested reference, they verify that the publication exists, that its bibliographic details are correct and that it supports the claim. A chatbot response does not replace evidence or an original source; when AI itself is studied, its responses may be research objects and must be identified as such.
AI-generated or AI-modified code is checked for suitability to the task and for correct inputs, units, algorithms and outputs. Authors perform the necessary control calculations and retain the code version used. A program running without errors does not by itself establish that its water balance, seepage or irrigation requirement calculations are correct.
Invented sources, measurements, consent, ethics approvals, reviews or author details, and undisclosed changes to results to support a desired conclusion, are prohibited. AI must not disguise unattributed borrowing or evade checks. Attribution and material reuse are governed by the Plagiarism Policy; financial relationships with a tool developer are disclosed under the Conflicts of Interest Policy.
05. AI as a research method and reproducibility
Machine learning, neural networks and generative models may be research methods in hydrological forecasting, soil analysis, remote sensing or irrigation management. The methods describe the model’s task, data provenance and preparation, available model and version details, material settings, training and validation, performance criteria and limitations. For a ready-made external service, authors report the access date and the available conditions for reproduction.
Where they affect results, report prompt templates, output selection rules, randomness settings and subsequent processing. Authors explain how training and validation data were separated; spatial and temporal series require attention to information leakage across sites, seasons and datasets. Baseline comparisons, uncertainty and limits of applicability must be appropriate to the conclusions claimed.
Synthetic data, model estimates and imputed missing values are clearly distinguished from direct observations. Their generation may be part of research, but they must not be presented as actual field measurements. Original data and material processing records are retained; access to them, code and necessary prompts follows the Data and Reproducibility Policy. Limitations of a closed model, changing service or restricted data are disclosed rather than concealed by a claim of full reproducibility.
06. Images, diagrams, plots and maps
Generative explanatory diagrams and graphical abstract elements are permitted if they accurately reflect concepts verified by the authors and are not presented as observational data. Captions identify AI use, the tool and the authors’ involvement. Visual plausibility does not replace checking relationships, labels and technical accuracy.
Plots, charts and result maps must be produced reproducibly from the specified data. Generating a plausible plot without corresponding values, or adding a field, structure or experimental effect that did not exist to an image, is prohibited. Original photographs must not be replaced with synthetic images presented as documentation of the research.
AI processing of images as an explicitly described method, such as land-use classification, segmentation or image reconstruction, is assessed under the methodological requirements. Authors retain original material, identify processed and model-derived elements, and demonstrate the necessary validation. Image attribution and usage rights are separate obligations; labelling an image “AI-generated” does not establish those rights.
07. Confidentiality and rights in materials
Before providing materials to an AI service, users check its conditions for retention, reuse, model training, third-party access and deletion. A paid subscription, a locally hosted model or a “private” label does not by itself establish confidentiality. Authors may process their own materials only with the necessary authority, co-author agreements and data rights; materials belonging to other people or organisations require a separate basis.
Personal data, restricted research materials, protected technical information or another person’s unpublished text must not be supplied to a service whose terms do not provide the necessary protection. Removing a name may not eliminate identifiability. Prompts, activity logs and AI responses may contain information as sensitive as the original files and must be protected accordingly.
Tool conditions must allow the intended publication without conflicting with rights in the included materials. Licensing and confidentiality details are governed by the respective policies. An author’s permission for a particular use does not override the restrictions on reviewers and editors below.
08. AI use by reviewers
Reviewers independently read the manuscript, examine its methods and formulate their comments. Delegating scientific assessment, substantive review writing or an acceptance or rejection recommendation to AI is prohibited. An AI report is not an independent review and does not replace the required number of expert reports.
Manuscripts, excerpts, unpublished results, supplements and review content containing their information must not be uploaded to external generative services, including services described as private. This restriction also covers prompts and descriptions from which confidential information can be reconstructed. Searches of public literature must not disclose the submitted work’s content.
Only local basic spelling and grammar checks that do not transmit materials externally or generate scientific comments are permitted without specific approval. Other supporting AI use, such as language editing of independently written generic phrases, requires prior written permission from the handling editor, verified data protection and compliance with the prohibitions above. Permitted use is recorded in the review, identifying the tool, task and checks performed; this information is provided to the editor, and the relevant declaration is shared with authors without revealing the reviewer’s identity.
09. Editorial work and production
Editorial decisions, assessment of significance, reviewer selection and checks of their independence remain the responsibility of an authorised person. Generative production of scientific assessments, reasons for decisions or substantive revision requirements is not permitted. Automatic reviewer invitations or decisions based on AI recommendations without an editor’s verification are prohibited.
The restrictions on transferring confidential material for reviewers also apply to editors and publisher staff. Supporting AI use beyond local basic checks requires documented permission from the editor-in-chief or an appointed independent editor, assessment of data protection and assignment of responsibility. This creates no exception for scientific assessment and does not imply that the journal has already implemented such a system.
Before introducing AI for technical checks, language processing or production, the editorial office publishes information on the tool, tasks, data supplied and human oversight. Substantive text changes are agreed with authors; AI assistance in preparing published material is disclosed in an appropriate note. AI-assisted translations and public-facing summaries are checked by a responsible editor and labelled. Ordinary automation of formatting or required-field checks is not in itself generative assessment of content.
10. Concerns, action and policy review
The journal sets no acceptable “AI-text percentage” and does not impose sanctions solely on a detector score or writing style. If obtained through a permitted process, such a signal requires examination of the specific circumstances: sources, calculations, versions, the declaration and authors’ explanations. Language proficiency, nationality and use of translation do not establish a breach. Requests for explanations identify concrete grounds rather than requiring authors to prove innocence against a score alone.
Incomplete disclosure that does not affect reliability may lead to a request to clarify the declaration. Substantial fabrication, confidentiality breaches, substitution of a review or other serious concerns are handled under the Allegations of Misconduct Policy; appropriate correction, expression of concern or retraction procedures apply after publication. A missing declaration does not automatically require retraction. Related policies provide the right to respond, independent handling and appeal.
Questions and reports may be sent through OJS or to i_m_jurnal@tiiame.uz. The editorial office records the actual approval and posting dates of each policy version. New disclosure-format requirements are not applied retrospectively as evidence of misconduct; concerns about the reliability of previously published work may be examined regardless of date. Changes in tools and practice provide grounds for updating the rules.
Guidance informing this policy
This policy draws on DOAJ transparency principles, ICMJE guidance on AI use, and Elsevier publishing practice. The specific permission, disclosure and verification requirements below are journal rules; they do not imply that the editorial office has any particular software product.