Manuel B. Garcia

Manuel B. Garcia serves as the Senior Director for Educational Technology and Digital Learning at FEU Institute of Technology, Manila, Philippines. Read More

Contact Info

1607, FEU Tech Building,
P. Paredes St, Sampaloc,
Manila, Philippines
mbgarcia@feutech.edu.ph

Follow Me

The Complete Library

All Research Guides

Browse every published guide in the Researcher's Field Guide, including the complete research journey and supplemental collections, listed alphabetically for quick reference.

Complete Library

Every published guide

Showing 51–100 of 5,065 guides

5,065 guides
PART 0D · SUBPART 89 · GUIDE 1146

AI Outputs That Cannot Be Independently Verified

Should Researchers Avoid AI When They Cannot Independently Verify the Output?

Researchers should not rely on unverifiable AI outputs for consequential scientific claims or decisions. The appropriate response depends on the purpose of the output, the consequences of error, and whether a credible independent validation method is available.

PART 0D · SUBPART 89 · GUIDE 1145

When Not to Use AI in Research

When Should Researchers Not Use AI?

AI is not appropriate for every research task. Researchers should avoid or suspend its use when they cannot verify consequential outputs, protect confidential information, satisfy applicable requirements, or defend the scientific decisions involved.

PART 0D · SUBPART 88 · GUIDE 1144

AI Policy Misunderstanding vs. Deliberate Concealment

How Do You Distinguish an Honest Misunderstanding of a Changing AI Policy From Deliberate Concealment?

A missing AI disclosure does not establish whether an author misunderstood a changing policy or intentionally concealed AI use. A defensible assessment examines what the policy required, what the researcher knew, what they did, and whether the evidence supports an inference of deliberate misrepresentation.

PART 0D · SUBPART 88 · GUIDE 1143

When Undisclosed AI Use Raises Integrity Concerns

When Does Undisclosed AI Use Become a Research-Integrity Concern?

Undisclosed AI use becomes a research-integrity concern when it compromises the honesty, transparency, accountability, or reliability of research. A missing declaration alone does not necessarily establish misconduct, but certain omissions warrant investigation beyond a simple disclosure correction.

PART 0D · SUBPART 88 · GUIDE 1142

Correcting a Missing AI Disclosure

Can a Missing AI Disclosure Be Corrected After Publication?

A missing AI disclosure can often be addressed after publication through the journal's editorial procedures. Learn how to prepare an accurate disclosure, request a correction, and distinguish transparency amendments from problems affecting the research itself.

PART 0D · SUBPART 88 · GUIDE 1141

Assessing the Seriousness of Undisclosed AI Use

How Does the Seriousness of Undisclosed AI Use Depend on What the AI Actually Did?

The seriousness of undisclosed AI use depends on more than whether a researcher used an AI tool. Its scientific significance, potential consequences, applicable disclosure requirements, and circumstances of omission all influence how the problem should be assessed.

PART 0D · SUBPART 88 · GUIDE 1140

When to Notify a Journal About AI Use

Does Every Undisclosed Use of AI Require Notifying the Journal?

Not every undisclosed AI interaction requires notifying a journal. The decision depends on the applicable disclosure policy, what AI contributed, and whether the omission affects the accuracy, transparency, or reliability of the research.

PART 0D · SUBPART 88 · GUIDE 1138

Undisclosed AI Use Discovered During Peer Review

What Should Researchers Do if Undisclosed AI Use Is Discovered During Peer Review?

When undisclosed AI use is identified during peer review, authors may face questions about transparency, methodological validity, and compliance with journal policies. Learn how to respond to editorial inquiries, establish what AI contributed, verify the affected research, and distinguish disclosure problems from more serious integrity concerns.

PART 0D · SUBPART 88 · GUIDE 1137

Undisclosed AI Use After Manuscript Submission

What Should Researchers Do if Undisclosed AI Use Is Discovered After Manuscript Submission?

Discovering undisclosed AI assistance after manuscript submission requires more than adding a statement. Learn how to establish what AI contributed, verify affected research, assess applicable policies, and communicate with the journal without unnecessarily disrupting peer review.

PART 0D · SUBPART 88 · GUIDE 1136

When a Coauthor Uses AI Without Disclosure

What Should a Research Team Do if a Coauthor Used AI Without Telling the Team?

Discovering that a coauthor used AI without informing the research team does not automatically establish misconduct, but it may expose important gaps in verification, transparency, and collaborative oversight. Learn how to establish what AI contributed, assess potential risks, resolve disclosure obligations, and strengthen research team procedures.

PART 0D · SUBPART 88 · GUIDE 1135

Correcting a Missing AI Disclosure in Research

What Should Researchers Do if They Realize AI Use Should Have Been Disclosed but Wasn’t?

Realizing that AI assistance should have been disclosed can create uncertainty, particularly after manuscript submission or publication. Learn how to establish which disclosure requirements apply, document the omitted contribution, correct the record, and distinguish an honest reporting mistake from more serious research-integrity concerns.

PART 0D · SUBPART 87 · GUIDE 1134

When AI Errors Require Research Record Corrections

When Does an AI-Caused Error Require a Correction to the Research Record?

Not every AI-generated mistake requires a formal correction notice, but errors that materially affect published evidence or conclusions must be addressed. Learn how to assess the seriousness of AI-related inaccuracies, distinguish internal revisions from editorial corrections, and determine when more substantial action may be necessary.

PART 0D · SUBPART 87 · GUIDE 1133

Blaming AI and Researcher Responsibility

Does Blaming AI Reduce a Researcher’s Responsibility for an Error?

When AI introduces an error, identifying the system as its source may explain what happened, but it does not automatically reduce a researcher's professional responsibility. Learn how accountability differs from blame, why intent and verification matter, and what researchers must do to correct AI-assisted work.

PART 0D · SUBPART 87 · GUIDE 1132

AI Errors Discovered After Publication

What Should You Do if an AI Error Is Discovered After Research Publication?

Discovering an AI-generated error after publication requires careful assessment of its consequences for the scholarly record. Learn how to verify the mistake, determine whether findings remain reliable, communicate with the journal, and distinguish corrections from retractions and other editorial actions.

PART 0D · SUBPART 87 · GUIDE 1131

AI Errors Discovered After Manuscript Submission

What Should You Do if an AI Error Is Discovered After Manuscript Submission?

Discovering an AI-generated error after manuscript submission creates responsibilities that depend on the error's seriousness and editorial stage. Learn how to verify the problem, assess its impact, communicate with coauthors and editors, and correct the manuscript before publication.

PART 0D · SUBPART 87 · GUIDE 1130

Correcting AI-Generated Inaccuracies in Research Manuscripts

What Should You Do if AI Introduced an Inaccurate Statement Into a Research Manuscript?

AI-generated academic writing can introduce inaccurate statements that appear credible and well supported. Learn how to identify the problem, verify the underlying evidence, correct affected passages, audit related claims, and determine whether further research-integrity or editorial action is necessary.

PART 0D · SUBPART 87 · GUIDE 1129

Correcting AI-Altered Participant Meaning

What Should You Do if AI Altered the Meaning of Participant Text?

AI-assisted transcription, translation, paraphrasing, and qualitative analysis can unintentionally change what participants meant. Learn how to identify semantic distortions, restore original meaning, reassess affected interpretations, and protect the integrity of participant accounts.

PART 0D · SUBPART 87 · GUIDE 1128

Correcting AI-Misclassified Research Data

What Should You Do if AI Misclassified Research Data?

AI misclassification can distort research categories, statistical estimates, qualitative interpretations, and scientific conclusions. Learn how to detect classification errors, assess their extent, validate corrected labels, and establish appropriate human oversight before using AI-classified data.

PART 0D · SUBPART 87 · GUIDE 1127

Correcting Incorrect Results From AI-Generated Code

What Should You Do if AI-Generated Code Produced Incorrect Research Results?

AI-generated code can execute successfully while producing scientifically incorrect results. Learn how to contain the error, trace affected analyses, validate corrected code, reassess conclusions, and determine whether collaborators or journals must be informed.

PART 0D · SUBPART 87 · GUIDE 1126

Fixing an AI-Fabricated Citation in a Research Draft

What Should You Do if AI Invented a Citation That Is Already in a Research Draft?

Discovering an AI-fabricated citation in a research draft requires more than deleting a reference. Learn how to establish whether the source exists, check the claim it was supposed to support, examine other AI-generated references, and correct the manuscript before submission.

PART 0D · SUBPART 87 · GUIDE 1125

What to Do When AI Introduces a Research Error

What Should Researchers Do When AI Introduces an Error Into Their Research?

An AI-generated error can affect research data, analyses, citations, interpretations, or conclusions. Learn how to contain the problem, determine its scope, correct affected work, and decide whether collaborators, institutions, or journals need to be informed.

PART 0D · SUBPART 86 · GUIDE 1124

Deciding Whether AI Suggestions Should Change Research

What Should Ultimately Determine Whether an AI Suggestion Changes a Research Study?

The decision to change a research study because of AI should depend on scientific justification, not the model's confidence or apparent sophistication. Learn how to evaluate evidence, methodological suitability, potential consequences, and approval requirements before accepting consequential AI recommendations.

PART 0D · SUBPART 86 · GUIDE 1123

Productive AI Challenges vs. Unreliable Suggestions

How Do Researchers Distinguish a Productive AI Challenge From an Unreliable Suggestion?

Not every AI-generated criticism deserves a change to a research study. Learn how to distinguish productive challenges that reveal meaningful questions from unreliable suggestions based on incorrect assumptions, unsupported claims, or inappropriate methodological advice.

PART 0D · SUBPART 86 · GUIDE 1122

AI Challenging Research Assumptions

Can AI Be Scientifically Useful Because It Challenges a Researcher’s Assumptions?

Generative AI may contribute to scientific inquiry by questioning assumptions that researchers have overlooked or taken for granted. Learn when AI-generated challenges can improve research reasoning, how to distinguish productive criticism from unsupported objections, and why independent verification remains essential.

PART 0D · SUBPART 86 · GUIDE 1121

Independent Evidence Before AI-Inspired Methodological Changes

Should Researchers Seek Independent Evidence Before Changing a Major Methodological Decision Because of AI?

Major methodological changes prompted by AI should be evaluated against independent evidence before implementation. Learn how to verify an AI-generated recommendation, compare defensible alternatives, assess the consequences of changing methods, and preserve accountability throughout the research process.

PART 0D · SUBPART 86 · GUIDE 1120

Documenting Research Decisions Changed by AI

Should Researchers Record Major Research Decisions That Changed After Consulting AI?

When AI consultation materially changes a research decision, documenting the change can strengthen transparency, reproducibility, and accountability. Learn which decisions warrant a record, what information to preserve, and how to distinguish internal documentation from formal AI disclosure requirements.

PART 0D · SUBPART 86 · GUIDE 1119

AI Anchoring Bias in Research Interpretation

Can AI Anchor a Researcher on an Interpretation They Would Not Otherwise Have Considered?

An AI-generated interpretation may become a reference point that influences how researchers examine evidence and evaluate competing explanations. Learn how anchoring can arise during AI-assisted research, why an initially plausible suggestion may be misleading, and how to protect scientific judgment through independent interpretation and structured verification.

PART 0D · SUBPART 86 · GUIDE 1118

AI Reasoning Weakness vs. Persuasive Output

How Do You Know Whether AI Exposed a Real Weakness in Your Reasoning or Merely Sounded Persuasive?

A persuasive AI explanation does not necessarily reveal a genuine weakness in scientific reasoning. Learn how to distinguish valid methodological criticism from fluent but unsupported AI output by testing assumptions, checking evidence, evaluating counterarguments, and examining whether the proposed correction survives independent scrutiny.

PART 0D · SUBPART 86 · GUIDE 1117

Can AI Change Scientific Judgment?

Should AI Be Allowed to Change a Researcher’s Scientific Judgment?

AI may legitimately influence a researcher's scientific judgment when its suggestions expose relevant evidence, overlooked assumptions, or defensible alternatives. The crucial distinction is between reconsidering a judgment after independent evaluation and allowing AI-generated advice to determine a scientific decision without adequate scrutiny.

PART 0D · SUBPART 85 · GUIDE 1116

Independent Evidence in AI and Expert Disagreements

What Role Should Independent Evidence Play When AI and Human Experts Disagree?

When AI and human experts disagree, independent evidence provides a stronger basis for evaluating scientific claims than confidence, credentials, or model agreement. Learn what qualifies as independent evidence, how to assess its reliability, and how to resolve disputes when the evidence remains uncertain.

PART 0D · SUBPART 85 · GUIDE 1115

Human Expert Overconfidence vs. AI

Can Human Experts Be Overconfident Just Like AI?

Human experts and generative AI systems can both express greater certainty than the evidence warrants, although their mechanisms differ. Understand how overconfidence affects research advice, why expertise does not guarantee accuracy, and how to evaluate recommendations without mistaking confidence for scientific validity.

PART 0D · SUBPART 85 · GUIDE 1114

Can AI Find Research Errors Experts Miss?

Can Generative AI Identify a Research Error That Experienced Human Researchers Missed?

Generative AI can sometimes flag inconsistencies, coding mistakes, or methodological concerns that experienced researchers overlooked. Understand which errors AI may help identify, why its findings are not automatically reliable, and how to verify suspected problems before changing a study.

PART 0D · SUBPART 85 · GUIDE 1113

Evidence Behind AI and Expert Advice

Should Researchers Ask for the Evidence Behind Both AI and Human Expert Recommendations?

Researchers should examine the evidence behind consequential recommendations, regardless of whether they come from AI or human experts. Learn what evidence to request, how to verify supporting claims, and why persuasive explanations and professional credentials cannot substitute for scientific justification.

PART 0D · SUBPART 85 · GUIDE 1112

Determining Whether AI or an Expert Is Wrong

How Do You Determine Whether AI or a Human Expert Made the Mistake?

When AI and a human expert disagree, determining who made the mistake requires more than comparing confidence or credentials. Learn how to isolate disputed claims, test assumptions, examine independent evidence, reproduce analyses, and distinguish genuine errors from defensible methodological differences.

PART 0D · SUBPART 85 · GUIDE 1111

Should Researchers Ignore AI When Experts Disagree?

Should Researchers Automatically Ignore AI When a Human Expert Disagrees?

Disagreement from a human expert is not sufficient reason to dismiss an AI recommendation without examination. Learn when an AI suggestion warrants investigation, when it should be rejected, and how to evaluate conflicting advice without compromising scientific judgment or research accountability.

PART 0D · SUBPART 85 · GUIDE 1110

Should Human Expertise Override AI?

Should Human Expertise Automatically Override AI Advice in Research?

Human expertise remains essential to responsible research, but does that mean expert recommendations should always prevail over AI advice? Understand why accountability and scientific correctness are different questions, when expert judgment deserves greater weight, and how to evaluate AI challenges without surrendering human oversight.

PART 0D · SUBPART 85 · GUIDE 1109

When AI Contradicts a Human Expert

What Should Researchers Do When AI Advice Contradicts a Supervisor or Human Expert?

When AI advice contradicts a supervisor or human expert, neither recommendation should be accepted solely because of its source. Learn how to investigate the disagreement, assess the evidence, respect supervisory responsibilities, and reach a defensible research decision.

PART 0D · SUBPART 84 · GUIDE 1108

AI Disagreement and Poorly Specified Research Tasks

When Does Disagreement Between AI Models Simply Mean the Research Task Is Poorly Specified?

Conflicting AI answers may result from a research task that lacks essential definitions, assumptions, or decision rules. Learn how to identify underspecification, distinguish it from genuine scientific uncertainty, and design clearer AI-assisted research procedures without assuming that better prompts guarantee correct results.

PART 0D · SUBPART 84 · GUIDE 1107

AI Model Disagreement and Scientific Uncertainty

When Does Disagreement Between AI Models Reflect Genuine Scientific Uncertainty?

Disagreement between AI models may reveal genuine uncertainty in scientific evidence, but it can also result from errors, missing information, or incompatible assumptions. Learn how researchers can distinguish these possibilities and represent uncertainty without mistaking model disagreement for scholarly controversy.

PART 0D · SUBPART 84 · GUIDE 1106

Resolving AI Disagreement With External Evidence

How Do Researchers Resolve Disagreement Between AI Models Using External Evidence?

When AI models disagree, researchers need evidence rather than another unsupported opinion. Learn how to identify the disputed claim, select an appropriate verification source, evaluate conflicting recommendations, and document an evidence-based resolution.

PART 0D · SUBPART 84 · GUIDE 1105

When AI Models Agree but Are Wrong

Can Multiple AI Models Agree and Still Be Wrong?

Several AI models can independently appear to confirm the same answer while sharing an underlying error. Learn how correlated mistakes, overlapping information, and common assumptions create false confidence, and how researchers can distinguish model agreement from genuine scientific corroboration.

PART 0D · SUBPART 84 · GUIDE 1104

Majority Voting Across AI Models

Should Researchers Use Majority Voting Across Multiple AI Models?

Majority voting across AI models can sometimes improve classification or decision performance, but agreement does not guarantee correctness. Learn how model dependence, correlated errors, task design, and independent validation determine whether voting is scientifically defensible.

PART 0D · SUBPART 84 · GUIDE 1103

Do More AI Citations Mean Greater Reliability?

Should Researchers Trust the AI Model That Provides More Citations?

An AI answer containing numerous citations may appear more credible than one with only a few sources. Learn why citation quantity is not a measure of correctness, how AI can misrepresent genuine publications, and how researchers should assess the evidence behind AI-generated claims.

PART 0D · SUBPART 84 · GUIDE 1102

Does More Detailed AI Advice Mean Greater Reliability?

Should Researchers Trust the AI Model That Gives the More Detailed Answer?

A detailed AI response can appear more credible than a concise alternative, but length and confidence do not establish accuracy. Learn how researchers can evaluate explanatory depth, detect unsupported claims, and verify AI-generated research advice without being persuaded by presentation alone.

PART 0D · SUBPART 84 · GUIDE 1101

When Two AI Models Disagree

What Should Researchers Do When Two AI Models Give Different Answers?

Two AI models may provide different answers to the same research question. Learn how to identify substantive disagreements, investigate their causes, verify claims against independent evidence, and avoid confusing model agreement with scientific correctness.

PART 0D · SUBPART 84 · GUIDE 1100

Why Identical Prompts Produce Different Research Advice

Why Can the Same Prompt Produce Different Research Advice?

The same AI prompt can produce different methodological recommendations because language models generate context-sensitive responses rather than applying a fixed scientific decision rule. Learn how to distinguish ordinary response variation from conflicting advice, identify missing research assumptions, and evaluate recommendations using methodological evidence.

PART 0D · SUBPART 84 · GUIDE 1099

When AI Gives Different Answers

What Should Researchers Do When AI Gives a Different Answer Every Time?

AI may produce different answers when researchers repeat the same question. Learn how to distinguish harmless wording variation from consequential analytical disagreement, investigate its causes, and establish whether an AI-assisted workflow is reliable enough for research.

PART 0D · SUBPART 83 · GUIDE 1098

Environmental Cost in Responsible AI Research

Should Environmental Cost Be Part of Responsible AI Research Planning?

Responsible AI research planning may need to account for electricity consumption, emissions, infrastructure, and resource demands. Learn when environmental assessment is warranted, how to incorporate it without compromising scientific rigor, and what researchers can reasonably document.

PART 0D · SUBPART 83 · GUIDE 1097

AI Model Performance vs. Computational Cost

How Should Researchers Balance AI Model Performance Against Computational Cost?

Choosing an AI model requires more than comparing accuracy or price. Learn how researchers can evaluate performance gains against computational costs, account for consequential errors, and select models that meet scientific requirements without unnecessary expenditure.