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 1–50 of 5,065 guides

5,065 guides
PART 0D · SUBPART 92 · GUIDE 1196

Detecting AI Errors Before They Affect Research

If the AI Is Wrong, Will Your Research Workflow Catch the Error Before It Matters?

AI errors are not always avoidable, but research workflows can be designed to detect consequential mistakes before they affect participants or scientific findings. Learn how to identify failure points, establish independent checks, test detection controls, and decide when an AI-assisted procedure is insufficiently protected.

PART 0D · SUBPART 92 · GUIDE 1195

Choosing Between AI and Non-AI Research Methods

Is There a Simpler or More Reliable Non-AI Method for the Research Task?

AI is not necessarily the best tool for every research task. Researchers should compare AI-assisted approaches with simpler validated alternatives, considering accuracy, methodological suitability, reproducibility, verification effort, and the actual benefits of automation.

PART 0D · SUBPART 92 · GUIDE 1194

Evaluating AI Research Quality Versus Speed

Is AI Actually Improving the Research or Merely Making the Task Faster?

Faster research is not necessarily better research, but efficiency can still be a legitimate benefit. Learn how to distinguish AI-generated time savings from improvements in scientific quality, account for verification and correction costs, and decide whether AI provides sufficient value for a particular research task.

PART 0D · SUBPART 92 · GUIDE 1193

Defending Research Under Full AI Transparency

Would the Research Still Be Defensible if Every AI Contribution Were Made Completely Transparent?

Full AI transparency raises a question beyond whether researchers complied with disclosure requirements: would the research remain scientifically defensible if its AI contributions were examined? Learn how to evaluate methodological validity, independent verification, accountability, and the limits of transparency as a research-integrity safeguard.

PART 0D · SUBPART 92 · GUIDE 1192

Taking Responsibility for AI-Assisted Research

Can You Take Full Responsibility for the Final AI-Assisted Research Output?

Researchers remain accountable for the scientific work they submit, even when AI contributes to writing, code, analysis, or interpretation. Learn what full responsibility means, what cannot be delegated to AI, and how researchers should respond when they cannot adequately verify or defend the final output.

PART 0D · SUBPART 92 · GUIDE 1191

Explaining AI Use to Research Stakeholders

Would You Be Comfortable Explaining Your AI Use to an Editor, Reviewer, Participant, Funder, or Ethics Committee?

Researchers may need to explain their AI use to stakeholders with different responsibilities and expectations. Learn what each stakeholder needs to know, how to substantiate AI-assisted procedures, and why being comfortable with disclosure is a useful reflection exercise but not a substitute for compliance or scientific validity.

PART 0D · SUBPART 92 · GUIDE 1190

Reconstructing AI-Assisted Research Processes

Can You Preserve Enough Information to Reconstruct the AI-Assisted Research Process?

A saved AI conversation may not be enough to reconstruct a research procedure. Learn how to preserve the inputs, model information, outputs, transformations, verification evidence, and human decisions needed to explain how AI-assisted findings were produced.

PART 0D · SUBPART 92 · GUIDE 1189

Documenting AI Tools and Models in Research

Can You Document Exactly Which AI Tool and Model Were Used?

Naming an AI application may not identify the model that produced a research output. Learn how to document providers, model identifiers, configurations, execution dates, and system dependencies, and how to report uncertainty when exact model information is unavailable.

PART 0D · SUBPART 92 · GUIDE 1188

Evaluating AI Privacy and Security Protections

Does the AI Tool Have Appropriate Privacy and Security Protections for the Research Task?

Security claims and enterprise subscriptions do not automatically make an AI tool suitable for confidential research. Learn how to evaluate provider terms, data flows, access controls, retention, and institutional authorization before selecting an AI system for a research task.

PART 0D · SUBPART 92 · GUIDE 1186

Determining Whether Research Information Can Be Shared With AI

Are You Allowed to Give the Necessary Research Information to the AI System?

Researchers may have legitimate access to information without permission to share it with an AI provider. Learn how to assess consent, confidentiality, institutional authorization, contractual restrictions, and provider processing arrangements before submitting research data to AI systems.

PART 0D · SUBPART 92 · GUIDE 1185

Independently Verifying AI Research Outputs

Can You Independently Verify the AI Output?

AI-generated answers can be persuasive without being scientifically correct. Learn what independent verification means, how to select suitable checks for different research tasks, and when insufficient evidence should prevent researchers from relying on an AI output.

PART 0D · SUBPART 92 · GUIDE 1184

Recognizing AI Errors Through Researcher Expertise

Do You Understand the Research Task Well Enough to Recognize When AI Is Wrong?

Researchers may receive convincing AI-generated analyses without possessing the expertise needed to recognize errors. Learn how to evaluate your understanding of an AI-assisted task, distinguish implementation mistakes from methodological problems, and determine when independent evidence or qualified expertise is necessary.

PART 0D · SUBPART 92 · GUIDE 1183

A Responsibility Test for AI Research Tasks

Should You Use AI for This Research Task? A Practical Responsibility Test

The availability of AI does not establish that it is appropriate for every research task. A practical responsibility test can help researchers evaluate methodological suitability, verification, authorization, potential harm, documentation, and whether AI offers a defensible advantage over alternative procedures.

PART 0D · SUBPART 91 · GUIDE 1182

Archiving AI Contributions for Research Transparency

What Should Researchers Archive So Future Researchers Can Understand Exactly How AI Contributed?

Future researchers may need more than an AI disclosure statement to understand how a study was conducted. Learn which model details, prompts, outputs, data references, validation evidence, and human decisions should be preserved, how to organize them, and when confidentiality limits what can be shared.

PART 0D · SUBPART 91 · GUIDE 1181

Auditing AI-Assisted Research Before Publication

How Do Researchers Audit an AI-Assisted Research Workflow Before Publication?

A prepublication AI audit examines whether AI-assisted research procedures are sufficiently documented, verified, and scientifically defensible. Learn how to trace AI contributions, assess methodological errors, review confidentiality and disclosure obligations, and decide whether unresolved issues must be corrected before submission.

PART 0D · SUBPART 91 · GUIDE 1180

Managing AI Model Changes During Research

How Should Researchers Handle AI Model Changes During a Long Research Project?

AI model updates can change research outputs even when prompts and datasets remain the same. Learn how to document model versions, assess changes in performance, protect longitudinal comparability, and decide when revalidation or reanalysis is necessary.

PART 0D · SUBPART 91 · GUIDE 1179

Information Researchers Should Not Enter Into AI Systems

How Do Researchers Decide Which Information Must Never Be Entered Into an AI System?

Researchers should not enter information into AI systems merely because the technology can process it. Learn how to identify prohibited or restricted inputs, assess confidentiality and reidentification risks, and distinguish public AI services from appropriately authorized research environments.

PART 0D · SUBPART 91 · GUIDE 1178

Documenting AI Interactions in Research

Which AI Interactions Should Researchers Log?

Researchers do not necessarily need to preserve every conversation with AI. Consequential interactions should be documented when they affect research methods, evidence, analysis, or conclusions. Learn what to record, how much detail is necessary, and how to protect confidential information in AI research logs.

PART 0D · SUBPART 91 · GUIDE 1177

AI Disclosure Requirements in Research

Which Uses of AI in Research Require Disclosure?

Not every interaction with AI requires the same disclosure. Researchers should distinguish AI-assisted writing from methodological contributions, understand journal and institutional requirements, and report consequential AI involvement with enough detail to support scientific scrutiny.

PART 0D · SUBPART 91 · GUIDE 1176

AI Research Privacy and Security Review

Which Uses of AI in Research Require Privacy or Information-Security Review?

AI use may require privacy or information-security review when research information is transferred, stored, accessed, or processed under new arrangements. Learn how to assess sensitive data, external providers, retention practices, access controls, and authorization before introducing AI into a research workflow.

PART 0D · SUBPART 91 · GUIDE 1175

AI Research Ethics Review Requirements

Which Uses of AI in Research Require Ethics Review?

Not every use of AI in research requires ethics committee approval. The need for review depends on whether AI changes participant-facing procedures, introduces risks, affects consent or eligibility, or alters the processing of protected research information. Learn how to identify activities that warrant formal review or an institutional determination.

PART 0D · SUBPART 91 · GUIDE 1174

AI Research Tasks Requiring Human Double-Checking

Which AI-Assisted Research Tasks Require Human Double-Checking?

Some AI-assisted research tasks require qualified humans to examine individual outputs, while others can be evaluated through independent calculations or systematic validation. Learn when human double-checking is necessary, how it differs from routine approval, and how to design review procedures that detect consequential errors.

PART 0D · SUBPART 91 · GUIDE 1173

AI Research Tasks Requiring Independent Verification

Which AI-Assisted Research Tasks Require Independent Verification?

AI outputs that materially affect research evidence, analysis, or conclusions require verification appropriate to their scientific role. Learn which tasks need independent checks, what counts as independent evidence, and when verification should go beyond asking another AI system.

PART 0D · SUBPART 91 · GUIDE 1172

AI Research Task Risk Classification

How Do Researchers Classify AI-Assisted Research Tasks by Risk?

AI research tasks differ in their scientific and ethical risks. Learn how to classify activities using the consequences of error, uncertainty, information sensitivity, detectability, and available safeguards, without mistaking a practical risk matrix for a validated universal standard.

PART 0D · SUBPART 91 · GUIDE 1171

AI Approval Before Participant Data Collection

Which Uses of AI Should Be Approved Before Participant Data Are Collected?

Researchers should establish necessary approvals before using AI in activities that affect participants, change approved procedures, or introduce new data-processing risks. Learn how to distinguish ethics approval, institutional authorization, privacy review, and methodological validation before participant data are collected.

PART 0D · SUBPART 91 · GUIDE 1170

Planning AI Use Before Research

Should AI Use Be Planned Before a Research Project Begins?

Planning AI use before research begins can prevent avoidable methodological, ethical, and documentation problems. Learn which AI activities should be specified in advance, when later changes are acceptable, and how to preserve flexibility without compromising scientific accountability.

PART 0D · SUBPART 91 · GUIDE 1169

Building a Responsible AI Research Workflow

How Do Researchers Build a Responsible AI-Assisted Research Workflow?

Responsible AI-assisted research requires more than selecting a tool and checking its final output. Learn how to design a workflow that establishes appropriate AI tasks, protects research information, verifies consequential results, preserves methodological decisions, and supports scientific accountability.

PART 0D · SUBPART 90 · GUIDE 1168

Using AI as a Research Assistant, Not an Authority

How Do Researchers Use AI as an Assistant Rather Than an Authority?

AI can help researchers generate possibilities, examine assumptions, and implement research procedures without becoming the authority behind scientific decisions. Learn how to distinguish useful assistance from unsupported AI influence and maintain evidence-based control over the research process.

PART 0D · SUBPART 90 · GUIDE 1167

AI Efficiency vs. Researcher Judgment

Should AI Save Researchers Time or Replace Researcher Judgment?

AI can reduce the effort required for research tasks, but efficiency should not come from abandoning the judgment needed to evaluate evidence and defend scientific decisions. Learn where automation is appropriate and how to distinguish legitimate time savings from uncritical delegation.

PART 0D · SUBPART 90 · GUIDE 1165

When Human Oversight Becomes Rubber-Stamping

When Is “Human in the Loop” Merely a Label for Rubber-Stamping AI Output?

A human approval step does not necessarily mean an AI-assisted research process has meaningful oversight. Rubber-stamping occurs when researchers approve outputs without sufficient evidence, competence, independence, or opportunity to challenge them.

PART 0D · SUBPART 90 · GUIDE 1164

Meaningful Human-in-the-Loop Oversight

When Is Human-in-the-Loop Oversight Actually Meaningful in AI-Assisted Research?

Human-in-the-loop oversight is meaningful only when reviewers can independently assess consequential AI outputs and intervene effectively. Learn how competence, evidence, authority, workload, and verification determine whether human supervision provides genuine protection.

PART 0D · SUBPART 90 · GUIDE 1163

Reproducing AI-Assisted Research Results

What Must Researchers Be Able to Reproduce Without Blindly Trusting AI?

Researchers should be able to reproduce or independently substantiate the consequential results they report, rather than relying on AI-generated answers. The required evidence depends on whether AI contributed calculations, classifications, interpretations, or other research outputs.

PART 0D · SUBPART 90 · GUIDE 1162

Defending AI-Assisted Research During Peer Review

What Must Researchers Personally Be Able to Defend to a Reviewer When AI Was Used?

Reviewers may reasonably question how AI influenced a study's methods, evidence, and conclusions. Researchers should be prepared to justify AI's role, present independent validation, explain limitations, and respond with verifiable documentation rather than relying on the tool's reputation.

PART 0D · SUBPART 90 · GUIDE 1161

Explaining AI-Assisted Research Work

What Must Researchers Personally Be Able to Explain About AI-Assisted Work?

Researchers should be able to explain what AI contributed, why it was appropriate, how consequential outputs were verified, and how human judgment shaped the final work. Technical detail should be sufficient for scientific scrutiny without implying that every internal model mechanism must be known.

PART 0D · SUBPART 90 · GUIDE 1160

Research Decisions Requiring Human Judgment

Which Research Decisions Must Researchers Personally Make Even When Using AI?

AI can suggest research questions, methods, interpretations, and publication strategies, but consequential scientific decisions must remain accountable to human researchers. Learn which decisions require human judgment and how AI can support them without becoming the authority.

PART 0D · SUBPART 90 · GUIDE 1159

What Researchers Must Verify in AI-Assisted Work

What Must Researchers Personally Verify When Using AI?

Researchers remain responsible for verifying consequential AI contributions before incorporating them into scientific work. Learn which claims, calculations, methodological decisions, and interpretations require checking, and how verification should reflect the risks of error.

PART 0D · SUBPART 90 · GUIDE 1158

What Researchers Must Understand About AI-Assisted Work

What Must Researchers Personally Understand Before Accepting AI-Assisted Work?

Researchers need not understand every technical detail of an AI model, but they must understand enough about its contribution to evaluate the methods, verify consequential results, recognize limitations, and defend the scientific work they accept.

PART 0D · SUBPART 90 · GUIDE 1157

Responsible AI Use in Research

What Does Responsible AI Use in Research Actually Look Like?

Responsible AI use is not simply disclosing that a tool was used or reviewing its final output. It requires a defensible workflow in which researchers select appropriate tasks, protect information, verify consequential contributions, document decisions, and remain accountable for the research.

PART 0D · SUBPART 89 · GUIDE 1156

The Strongest Reason Not to Use AI

What Is the Strongest Reason Not to Use AI Even When It Is Available?

The strongest reason not to use AI is that its use cannot be responsibly justified for the research task. Learn how scientific defensibility brings together reliability, authorization, verification, and accountability without reducing the decision to a simple preference for or against AI.

PART 0D · SUBPART 89 · GUIDE 1155

AI Complexity Without Research Quality Improvement

When Does AI Add Complexity Without Improving Research Quality?

AI can make a research workflow more complicated without making its findings more credible. Learn how to identify unnecessary AI components, assess their methodological costs, and decide when a simpler procedure offers a stronger scientific justification.

PART 0D · SUBPART 89 · GUIDE 1154

AI Time Savings vs. Scientific Value

Should Researchers Avoid AI When It Saves Time but Adds No Scientific Value?

Saving time can be a legitimate reason to use AI, but faster output does not necessarily improve research quality. Learn how to evaluate efficiency gains, verification burdens, scientific consequences, and whether AI provides enough value to justify its use.

PART 0D · SUBPART 89 · GUIDE 1151

Using AI Models That Cannot Be Documented

Should Researchers Avoid AI When the Model Cannot Be Documented Adequately?

Researchers should reconsider AI use when they cannot document enough about the model and its contribution to evaluate or reconstruct a consequential research procedure. Learn which records matter, when incomplete documentation is manageable, and when it undermines scientific defensibility.

PART 0D · SUBPART 89 · GUIDE 1149

Using AI With Unfamiliar Research Methods

Should Researchers Avoid AI When They Do Not Understand the Underlying Method?

AI can help researchers learn and implement unfamiliar methods, but it cannot replace the understanding needed to evaluate their suitability and defend the results. Learn when methodological unfamiliarity is manageable and when AI-assisted analysis should be postponed or avoided.

PART 0D · SUBPART 89 · GUIDE 1148

AI Tools With Unclear Data-Use Terms

Should Researchers Avoid AI When the Tool’s Data-Use Terms Are Unclear?

Unclear AI data-use terms can prevent researchers from determining whether a tool is appropriate for confidential or restricted information. Learn which provider commitments matter, how to assess uncertainty, and when to withhold data or select another system.

PART 0D · SUBPART 89 · GUIDE 1147

AI and Confidential Research Data

Should Researchers Avoid AI When Confidential Research Data Cannot Be Protected?

Researchers should not enter confidential research data into an AI system unless the transfer is authorized and appropriate protections are established. Learn how to evaluate confidentiality risks, consent limitations, provider arrangements, and safer research workflows.