Artificial Intelligence in Research
Researchers should consider keeping records of consequential AI assistance, particularly when it influences research methods, evidence, analysis, or findings. Learn what documentation can establish, which interactions warrant preservation, and how AI-use records support transparency without requiring researchers to archive every conversation.
Read Guide →
Artificial Intelligence in Research
A research AI-use log can document which tools were used, what tasks they performed, when interactions occurred, and how consequential outputs were evaluated. Learn which records matter for transparency, reproducibility, and research integrity without turning every minor AI interaction into an administrative burden.
Read Guide →
Artificial Intelligence in Research
Researchers should consider preserving AI-generated responses that materially influence research methods, analysis, or findings. Learn why prompts alone may be insufficient, which outputs warrant retention, and how to maintain accurate, secure, and useful records of AI assistance.
Read Guide →
Artificial Intelligence in Research
Recording AI model versions can help researchers document which system contributed to their work, particularly when AI affects analysis or findings. Learn what information to preserve, how model versions differ from application names, and what to do when exact identifiers are unavailable.
Read Guide →
Artificial Intelligence in Research
AI tool settings may influence generated outputs and should be recorded when relevant to research methods or findings. Learn which configurations matter, how temperature and sampling parameters affect output variability, and what to document when commercial AI applications hide their settings.
Read Guide →
Artificial Intelligence in Research
AI-assisted research can be reproducible, but saving a prompt is rarely enough. Learn how model versions, settings, archived outputs, and independent validation affect the reproducibility of AI-supported research workflows.
Read Guide →
Artificial Intelligence in Research
The same AI prompt can produce different answers because language models may use probabilistic generation, changing model configurations, and variable execution environments. Understanding these differences helps researchers evaluate the reliability of AI-assisted methods.
Read Guide →
Artificial Intelligence in Research
AI model updates can make an earlier research workflow difficult or impossible to reproduce exactly, especially when the original model is no longer accessible. Learn what changes, which research results remain verifiable, and how to prepare for model retirement.
Read Guide →
Artificial Intelligence in Research
Researchers should record temperature and other relevant AI settings when those settings may affect research outputs. Learn which parameters matter, what they control, and how to document configurations that are not accessible.
Read Guide →
Artificial Intelligence in Research
AI prompts can function as analytical instructions when they determine how research data are classified, extracted, or interpreted. Learn when prompts should be documented, tested, and version-controlled like research code.
Read Guide →
Artificial Intelligence in Research
AI outputs that influence research data, analysis, or conclusions should generally be preserved as part of the research record. Learn what to archive, how to protect confidential information, and why saving outputs matters even when the original AI model remains available.
Read Guide →
Artificial Intelligence in Research
When the original AI model is no longer accessible, exact regeneration may be impossible, but important parts of an AI-assisted analysis can still be verified. Learn how to reconstruct the analytical record, reproduce downstream calculations, and evaluate replacement models without overstating reproducibility.
Read Guide →
Artificial Intelligence in Research
A nondeterministic AI model can be incorporated into a controlled research workflow without requiring every generated response to be identical. Learn how structured outputs, validation, archived records, and deterministic downstream procedures can improve reproducibility while preserving the distinction between computational consistency and scientific validity.
Read Guide →
Artificial Intelligence in Research
AI-assisted coding stability should be tested rather than assumed from a few consistent outputs. Learn how to design repeated-run evaluations, measure classification agreement, examine errors, and determine whether AI-generated codes are sufficiently dependable for research.
Read Guide →
Artificial Intelligence in Research
Important AI-assisted research tasks may need repeated execution to establish whether their outputs are sufficiently stable. Learn when rerunning is worthwhile, how to design meaningful tests, and why repeated agreement does not necessarily establish scientific accuracy.
Read Guide →
Artificial Intelligence in Research
Probabilistic AI outputs challenge the assumption that reproducibility always requires identical responses. Learn how researchers can evaluate distributions, measurement stability, and scientific conclusions while distinguishing reproducibility from accuracy and validity.
Read Guide →
Artificial Intelligence in Research
An AI-assisted method may be too unstable for scientific use when output variability undermines measurement validity or changes the conclusions. Learn how to recognize consequential instability, evaluate its effects, and decide when an AI method should be revised, restricted, or replaced.
Read Guide →
Artificial Intelligence in Research
Generative AI may help reviewers organize and refine their comments, but its use is subject to journal policies, confidentiality obligations, and independent scholarly judgment. Learn which forms of assistance may be appropriate and why reviewers remain responsible for their reports.
Read Guide →
Artificial Intelligence in Research
Peer reviewers generally should not upload confidential manuscripts to external AI tools without explicit journal authorization and appropriate safeguards. Learn why manuscript confidentiality applies to AI services, what permission actually means, and how to avoid unauthorized disclosure.
Read Guide →
Artificial Intelligence in Research
Using AI during peer review can violate confidentiality when unpublished manuscript information is disclosed to an unauthorized system. Learn why confidentiality extends beyond uploading PDFs, how publisher policies differ, and what safeguards reviewers must observe.
Read Guide →
Artificial Intelligence in Research
Peer reviewers should follow journal requirements for disclosing generative AI assistance, and disclosure may be expected even when AI is used only for language editing. Learn what should be reported, why permission must come before use, and how to distinguish transparent assistance from unauthorized delegation.
Read Guide →
Artificial Intelligence in Research
Journal editors may use AI to support selected manuscript-screening and editorial tasks, but automated assessments can introduce confidentiality, accuracy, and fairness concerns. Learn which applications may be appropriate, how editorial AI differs from reviewer AI use, and why consequential publication decisions require accountable human judgment.
Read Guide →
Artificial Intelligence in Research
Authors may have legitimate reasons to know when AI materially influences editorial assessment, particularly rejection or revision decisions. Learn how disclosure differs from permission, why routine automated checks require different consideration, and how journals can make AI-assisted editorial decisions more transparent and accountable.
Read Guide →
Artificial Intelligence in Research
AI-assisted peer review may introduce or reinforce biases affecting which manuscripts receive favorable assessments. Learn how methodological conventions, language patterns, research contexts, and automation bias can shape AI-generated evaluations, and how journals can test for unfair differences.
Read Guide →
Artificial Intelligence in Research
Some peer-review tasks may benefit from AI assistance, but evaluating scientific significance, methodological appropriateness, evidentiary strength, and the proportionality of criticism requires accountable human judgment. Learn where automation reaches its limits and why meaningful oversight is more than approving generated recommendations.
Read Guide →
