Research Data Availability & Reproducibility Policy

Purpose, Scope and Policy Principles

ARCITECH: Journal of Computer Science and Artificial Intelligence promotes credible, transparent, verifiable, and ethically responsible research in computer science, artificial intelligence, data science, information systems, computer vision, cybersecurity, and related computational fields. This policy describes how authors disclose the availability of research data and document the computational materials needed to evaluate their findings.

This policy applies to manuscripts reporting original empirical, experimental, simulation, computational, or secondary-data research. For articles that do not generate or analyze research data, authors must explain why a data-availability statement is not applicable. Scientific evaluation is based on research quality and integrity, not solely on whether all research assets are publicly downloadable.

• Transparency and accountability for results and research objects.
• As open as reasonably possible; as restricted as necessary.
• FAIR data stewardship and meaningful reproducibility without violating privacy, security, intellectual-property rights or other legitimate restrictions.

Research Data and Supporting Materials

Research data include observations, measurements, survey or interview records, annotations, derived features, benchmark datasets, experiment logs, tabulated results, images, audio, video, system telemetry, network traces, and synthetic or simulated outputs that support the reported conclusions.

Supporting computational materials include source code, analysis scripts, pseudocode, configuration files, workflow descriptions, model architectures, hyperparameters, trained-model identifiers or checkpoints where distributable, prompts when material, software dependencies, execution environments, random seeds, and protocols.

• The journal does not require unlawful redistribution of external datasets, proprietary code, or security-sensitive material.
• Supplementary files submitted to OJS are not by themselves proof of preservation or a substitute for reliable repository hosting when long-term public availability is claimed.

Mandatory Data Availability Statement

Every manuscript reporting research that collects, generates, transforms, simulates, or analyzes data must contain a clearly labelled “Data Availability Statement”. For papers without underlying empirical or computational research data, an explicit “Not applicable” statement with a brief reason is permitted.

Authors shall place this statement with the other declarations before the References section or in the designated journal template location. The statement must accurately describe what data or materials exist, whether and how they can be accessed, and any lawful or ethical restrictions. It must not falsely claim public access or imply that a private dataset has been independently verified.

• Identify public repositories and stable identifiers (DOI or persistent URL), including dataset versions, when available.
• Distinguish open data from controlled-access, third-party, proprietary, unavailable, or not-applicable materials.
• Identify the data custodian or application procedure for legitimate restrictions, without publishing personal contact information of participants.
• Update broken or altered links during peer review and before publication.

Data Sharing and Repository Deposit

ARCITECH strongly encourages authors to deposit lawfully shareable research data, accompanying documentation, and supporting research objects in trustworthy institutional, subject-specific, or general-purpose repositories that provide stable access, documented versions, and appropriate preservation practices.

Public sharing is encouraged but is not an automatic condition of acceptance for research with reasonable restrictions. Where journal or funder requirements apply, authors must satisfy the more specific obligations or explain conflicts before editorial acceptance. Authors should deposit the precise dataset version underlying the accepted article whenever feasible.

• Prefer stable identifiers, version history, descriptions, license terms, and a clear citation.
• Provide a README, data dictionary, file formats and processing notes adequate to interpret shared materials.
• Avoid unstable download links, unversioned personal storage, or private account links as the only access route.

FAIR Data Documentation and Provenance

Where practical, deposited data should follow the FAIR Guiding Principles. FAIR does not mean every dataset must be unrestricted or publicly downloadable. Restricted data can still be discoverable through metadata and accessed under documented authorization rules.

Authors should describe how data were collected, obtained, filtered, labelled, cleaned, transformed, balanced, split, and analyzed; include original sources, licenses, dates, units, encoding, schema, and known quality limitations when relevant.

• Make dataset identifiers and access conditions discoverable.
• Use machine-readable, non-proprietary formats when feasible, without compromising fidelity or legal obligations.
• Explain transformations and provenance, including labelling or annotation quality and any synthetic augmentation.

Source Code and Computational Reproducibility

Computational research must report enough methodological and technical detail for a qualified researcher to assess, repeat, or meaningfully interrogate the reported procedure. Authors are encouraged to release the analysis code and scripts used for figures, metrics, preprocessing, simulations, and statistical tests when permitted.

When code cannot be shared, authors must identify the reason and document the algorithm, relevant pseudocode, implementation choices, software components, and execution conditions to the maximum permissible extent.

• Document software/library versions, operating environment, hardware or accelerator where material, runtime constraints and dependency information.
• Report configuration files, random seeds and computational resources needed to interpret results.
• Link to a tagged release or immutable archive when possible; a changing default branch alone is not an adequate reference to a particular experiment.
• Do not publish tokens, API keys, passwords, authentication secrets or exploitable infrastructure configuration.

Machine Learning, Deep Learning and Generative AI Experiments

Research using AI/ML should document datasets, preprocessing, inclusion/exclusion rules, train/validation/test partitions, evaluation design, baselines, metrics and uncertainty where applicable. Authors must disclose controls against data leakage, contamination, improper test reuse and selective reporting.

When generative models materially affect the scientific method or reported outputs, describe relevant model identity/version, prompt design, inference settings, seeds and validation steps, when lawful and feasible. AI-assisted manuscript preparation is governed separately by ARCITECH’s AI & Generative AI Policy.

• Identify training data origin, annotation procedures, sampling and validation design.
• Report hyperparameters, initialization, checkpoints or comparable technical detail where sharing is permitted.
• Distinguish synthetic, augmented and generated data from independently observed data.
• Describe variation due to nondeterministic hardware, closed services, costly training, changed APIs, and other reproducibility limitations.

Experimental Validity and Reporting

Manuscripts should state how results were computed, how many independent runs were performed when relevant, and whether outcome variability was assessed. Authors should justify baseline selection and clearly identify the comparison conditions.

For computer science experiments, explain task-specific evaluation metrics, data partitioning, statistical summaries, potential confounders, and known threats to internal or external validity. Claims of superiority must not rely on hidden data filtering or unfair comparison settings.

• Include enough information to reproduce tables and plots or verify their derivation.
• Report negative or inconclusive results when they are central to the stated findings.
• Avoid overstating exact reproducibility for inherently stochastic, hardware-sensitive, time-varying or proprietary systems.

Human, Personal, Proprietary and Restricted Data

Authors must protect personal data and respect consent, human-research ethics approvals, confidentiality obligations, contract terms, and applicable law. Publication or repository deposit must not disclose raw records containing identifiable individuals, private communications, protected records, or restricted institutional data without a lawful basis and appropriate safeguards.

For legitimately restricted data, authors should state the specific category of restriction and describe a feasible access request, secure analysis environment, aggregated statistics, de-identified extracts, documented procedures, or substitute materials where these options are safe and authorized. Anonymization must not be claimed unless re-identification risk has been appropriately addressed.

• Ethics approval or consent information must be reported where applicable.
• Data access through another institution must comply with that institution’s written permissions.
• Restrictions must not be used to conceal nonexistent data, invalid analyses or otherwise unsubstantiated conclusions.

Cybersecurity Data and Dual-Use Research

Cybersecurity and digital-forensics studies may contain raw security logs, IP addresses, credentials, exploit traces, malware samples, vulnerabilities, attack paths or other dual-use artifacts. Such materials should not be publicly deposited when release could facilitate abuse, disclose personal information, or expose unpatched systems.

Authors should document authorization, collection scope, provenance, sanitization and risk controls. A defensible research record may consist of redacted telemetry, aggregated features, synthetic test cases, non-weaponizing pseudocode, responsible disclosure documentation, and controlled-access verification instead of a raw public dump.

• Never disclose active credentials, exploit-ready details targeting identifiable systems, or private security telemetry without authorization.
• Separate real incident evidence from laboratory simulations and generated attack examples.
• Explain the security reason for non-disclosure and provide safe supporting information where practical.

Third-Party, Licensed and Commercial Datasets

Authors using datasets or software created by others must identify the original source, access date or version, applicable license, terms of use, permission requirements and appropriate citation. A manuscript’s publication under the journal’s license does not grant permission to redistribute third-party data or code.

If a dataset is commercially licensed, access-controlled or prohibited from redistribution, authors must link to the lawful provider and give sufficient study details to allow independent evaluation without misrepresenting availability.

• Public availability is not equivalent to permission for arbitrary redistribution or model training.
• Disclose material contractual or dataset access restrictions that limit independent validation.

Licensing of Data, Software and Research Outputs

ARCITECH articles normally use CC BY-NC-SA 4.0 under the separate Copyright & Licensing Policy. That article license does not automatically establish the ownership, permissions or licensing terms for an independently hosted dataset, source-code repository, pre-trained model or third-party content.

Authors must state or link to the actual license or usage restrictions attached to each shared dataset, software release and model artifact. Software code should normally use an appropriate software license rather than assume the journal’s Creative Commons article license applies.

Peer Review, Anonymity and Editorial Verification

During double-blind review, authors should provide data or code links that do not unnecessarily disclose their identities, where practical. Public repository metadata may incidentally reveal identity; the editor should manage this transparently under the journal’s Peer Review Policy rather than requiring deceptive anonymization or removal of provenance.

Reviewers and editors may request reasonable clarification or confidential inspection of underlying data, analysis code, or computational documentation to assess disputed or central claims. Access must be lawful and proportionate, with restricted records protected. Such requests do not mean the journal promises independent replication of all reported results.

• Reviewers must not upload confidential manuscripts, data or code to unauthorized AI systems or third-party services.
• Any reviewer access to restricted materials must follow prior approval, confidentiality and any relevant data-access agreement.

Data Retention, Corrections and Integrity

Authors are responsible for maintaining appropriate source records and computational evidence in accordance with their institution’s policies, funding obligations, legal duties, and good scientific practice. ARCITECH does not prescribe a universal retention period for all research contexts.

If an author discovers an error, corrupted repository object, inaccessible citation, manipulated dataset, or material discrepancy affecting a published claim, the author must inform the journal promptly. Corrections or further editorial action will follow the journal’s Publication Ethics and Corrections, Retractions & Expressions of Concern Policy.

Assessment, Exceptions and Editorial Decisions

Editors may request clarification, stronger provenance, more complete methodological information, or a revised data-availability statement when essential to evaluating a manuscript. The handling editor considers the scientific importance of the data, the reasonableness of restrictions, and whether the findings remain verifiable.

Unresolved inconsistencies, knowingly false data-availability claims, fabricated observations, concealed data manipulation, or refusal to provide adequate verification without a legitimate reason may lead to revision requests, rejection, or post-publication review under the journal’s established ethics procedures.

The journal does not guarantee that third-party repositories will maintain access permanently, and it does not certify that every accepted analysis has been independently reproduced. Authors and repository custodians retain responsibility for the accurate statements they make about availability and preservation.

Data Availability Statement: Sample Wording

The examples below must be customized truthfully. Replace bracketed text with accurate manuscript-specific information; do not leave placeholders in a submitted article. If both public and restricted datasets are used, state the availability of each relevant category separately.

Implementation in ARCITECH OJS

The corresponding author is responsible for including an accurate Data Availability Statement in the manuscript and submission materials. Data citations and software references should be part of the article’s references or linked supporting resources when appropriate. Editors assess compliance through the normal screening and peer-review workflow.

Where stable data or software DOIs exist, authors should supply those identifiers for inclusion in the article and, where supported by the journal’s metadata workflow, linkage through Crossref. This policy does not claim that automatic data DOI deposit, repository hosting, or independent computational verification is already enabled by OJS.

• For technical questions, use the official ARCITECH editorial contact listed on the journal website.
• Read this policy together with the Author Guidelines, Peer Review, Publication Ethics, AI Policy and Copyright & Licensing Policy.

Standards, Cross-References and Interpretation

This policy adopts transparency and reproducibility practices informed by recognized publishing and open-science frameworks. References are guidance, not endorsements, accreditation decisions, or claims that an external organization has certified ARCITECH.

The policy operates alongside ARCITECH’s established rights and editorial policies. Where a particular dataset or software license imposes a stricter lawful restriction, the specific rights and permissions attached to that research object remain applicable.

Related journal policies: Author Guidelines, AI Policy, Publication Ethics, Peer Review Policy, Copyright & Licensing, Open Access & Repository.