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 301–350 of 5,065 guides

5,065 guides
PART 0D · SUBPART 66 · GUIDE 896

Reference Standards for AI Evaluation

What Should Count as the Reference Standard When Evaluating an AI System?

An AI performance score is only as defensible as the standard used to judge its outputs. Learn how researchers can establish reliable reference answers, handle expert disagreement, assess ambiguous cases, and avoid treating imperfect benchmarks as unquestionable ground truth.

PART 0D · SUBPART 66 · GUIDE 895

Evaluating AI Performance in Research

How Should Researchers Evaluate AI Performance in a Research Study?

Evaluating AI performance requires more than calculating an accuracy score or selecting impressive examples. Learn how to define the evaluation objective, establish reference standards, select appropriate metrics, account for variability, and interpret results within the conditions actually tested.

PART 0D · SUBPART 65 · GUIDE 894

Reproducibility of Changing AI Systems

How Do Researchers Make Studies of Rapidly Changing AI Systems Scientifically Reproducible?

Commercial AI systems can change during or after a research study, making exact replication difficult. Learn how to preserve experimental conditions, document model behavior, distinguish reproducibility from response stability, and produce scientifically defensible findings even when the original system becomes unavailable.

PART 0D · SUBPART 65 · GUIDE 893

When AI Model Research Becomes Outdated

Can Research About an AI Model Become Outdated Before It Is Published?

An AI model may be updated, replaced, or withdrawn before a study evaluating it reaches publication. Learn why this does not automatically invalidate the research and how to preserve the scientific value of findings about rapidly changing AI systems.

PART 0D · SUBPART 65 · GUIDE 892

Measuring Prompt Sensitivity in AI Research

What Is Prompt Sensitivity, and How Should Researchers Measure It?

Prompt sensitivity describes how an AI model's responses change when the wording or structure of its instructions changes. Learn how researchers can measure these effects, distinguish meaningful variation from superficial differences, and report prompt robustness without overstating model reliability.

PART 0D · SUBPART 65 · GUIDE 890

Identical Prompts in AI Model Comparisons

Should Different AI Models Be Tested With Identical Prompts?

Identical prompts can strengthen comparisons between AI models, but they do not automatically guarantee fairness. Learn when exact prompt matching is appropriate, when model-specific adaptations are defensible, and how to preserve methodological comparability.

PART 0D · SUBPART 65 · GUIDE 889

Comparing AI Models Fairly

How Do Researchers Compare AI Models Fairly?

A fair AI model comparison requires more than submitting the same questions to different systems. Learn how to establish comparable testing conditions, select appropriate performance measures, account for response variability, and avoid misleading conclusions about which model performs better.

PART 0D · SUBPART 65 · GUIDE 888

Accounting for Nondeterministic AI Responses

How Do Researchers Account for Nondeterministic AI Responses in a Study?

Generative AI may produce different responses to identical prompts, complicating reproducibility and performance evaluation. Learn how researchers can measure this variability, control experimental conditions, analyze repeated outputs, and report findings without overstating their stability.

PART 0D · SUBPART 65 · GUIDE 885

Preserving AI Research Prompts

Should Researchers Preserve Every Prompt Used to Study an AI Model?

Prompts are often central methodological materials when researchers study generative AI systems. Preserving only successful prompts may conceal selection bias, while incomplete conversation records can make evaluations difficult to reproduce. Learn which prompts should be archived, how to document exploratory and formal testing, and when confidentiality or practical constraints justify restricted preservation.

PART 0D · SUBPART 65 · GUIDE 884

Recording AI Model Testing Dates and Times

Should Researchers Record the Date and Time When an AI Model Was Tested?

Recording when an AI model was tested helps researchers establish the temporal context of their findings, especially when commercial systems may change without notice. Learn when testing dates are sufficient, when individual response timestamps are necessary, and how time zones, model updates, and longitudinal designs affect the reproducibility and interpretation of AI research.

PART 0D · SUBPART 65 · GUIDE 883

Recording AI Model Versions in Research

Should Researchers Record the Exact AI Model Version Being Studied?

Recording the exact AI model version helps researchers establish which system produced the observations in a study. However, commercial AI services do not always expose immutable model identifiers, and a model name may refer to a changing deployment. Learn what version information should be preserved, why model identity matters for reproducibility, and how to report uncertainty when the exact model state is unavailable.

PART 0D · SUBPART 65 · GUIDE 882

Studying Changing Commercial AI Models

How Do Researchers Study a Commercial AI Model That Can Change Without Notice?

Commercial AI models may change through updates to their underlying models, system instructions, safety mechanisms, or supporting infrastructure, sometimes without detailed public notice. These changes complicate reproducibility and longitudinal research. Learn how researchers can document testing conditions, detect behavioral changes, and produce scientifically defensible findings even when the AI system cannot be frozen or fully inspected.

PART 0D · SUBPART 65 · GUIDE 881

Studying AI as a Research Subject

What Changes Methodologically When AI Itself Is the Object of a Research Study?

When artificial intelligence itself becomes the object of a research study, researchers must treat model behavior as the phenomenon being investigated. This changes how observations are collected, experimental conditions are controlled, responses are evaluated, and findings are generalized. Learn why model versions, prompts, conversation history, repeated testing, and system updates become important methodological considerations.

PART 0D · SUBPART 64 · GUIDE 880

AI Simulation vs. Empirical Research

When Is AI Simulation a Legitimate Methodological Tool Rather Than a Substitute for Empirical Research?

AI simulation can be a legitimate methodological tool when researchers use it to examine computational behavior, test research procedures, or investigate explicitly defined hypothetical conditions. However, simulation cannot automatically replace empirical evidence about real people or events. Learn how to distinguish scientifically defensible simulations from inappropriate substitutes for data collection.

PART 0D · SUBPART 64 · GUIDE 879

Preventing Confusion Between AI Simulations and Empirical Evidence

How Do Researchers Prevent AI-Simulated Participants From Being Mistaken for Empirical Evidence?

AI-simulated participants can produce convincing survey responses and interview narratives, creating a risk that artificial observations will be mistaken for genuine human evidence. Researchers can reduce this risk through clear labeling, reliable provenance records, separation of datasets, transparent reporting, and appropriate analytical safeguards. Learn how to preserve the distinction throughout the research lifecycle.

PART 0D · SUBPART 64 · GUIDE 878

Limits of Scientific Claims From AI-Simulated Participants

What Scientific Claims Cannot Be Made From AI-Simulated Participants?

AI-simulated participants can provide useful evidence about model behavior and methodological procedures, but they cannot independently establish what real people experience, believe, or do. Learn which scientific claims exceed the evidentiary limits of simulated responses, why statistical significance does not resolve these limitations, and how researchers can avoid unsupported conclusions.

PART 0D · SUBPART 64 · GUIDE 877

Scientific Claims From AI-Simulated Participants

What Scientific Claims Can Researchers Make From AI-Simulated Participants?

AI-simulated participants can support scientific claims about model behavior, simulation performance, and certain methodological procedures. However, their responses do not independently establish what real people believe, experience, or do. Learn how researchers can frame defensible findings, distinguish computational observations from human evidence, and avoid unsupported population-level conclusions.

PART 0D · SUBPART 64 · GUIDE 876

AI Simulation of Rare Research Scenarios

Can Generative AI Simulate Rare Scenarios for Research Methods Development?

Generative AI can help researchers construct rare or difficult-to-observe scenarios for testing research instruments, analytical procedures, and methodological decisions. However, simulated scenarios reflect model assumptions rather than independent observations of rare events. Learn when AI-generated edge cases are useful, how to evaluate them, and why methodological stress testing should not be confused with empirical validation.

PART 0D · SUBPART 64 · GUIDE 875

Testing Coding Frameworks With AI Responses

Can AI-Generated Responses Help Researchers Test a Coding Framework?

AI-generated responses can help researchers test qualitative coding frameworks by creating controlled examples, ambiguous passages, and cases that challenge code definitions. However, successful classification of synthetic text does not establish that a framework captures the meaning of genuine participant accounts. Learn how to use artificial excerpts for methodological testing while preserving empirical and interpretive validity.

PART 0D · SUBPART 64 · GUIDE 874

AI-Simulated Qualitative Interviews

Can Generative AI Simulate Qualitative Interview Responses?

Generative AI can produce realistic interview-style responses from hypothetical participants, making it potentially useful for interviewer training, protocol development, and qualitative coding tests. However, AI-generated narratives are not genuine accounts of lived experience. Learn how simulated interviews can support qualitative methodology without being misrepresented as empirical testimony.

PART 0D · SUBPART 64 · GUIDE 873

AI-Generated Participants for Pilot Testing

Can AI-Generated Participants Be Used for Research Pilot Testing?

AI-generated participants can support preliminary research pilot testing by helping researchers examine questionnaire wording, survey logic, interview procedures, and analytical workflows. However, simulated respondents cannot independently establish how actual people understand questions or behave during research. Learn how to use AI simulations appropriately while preserving the role of genuine human pilot testing.

PART 0D · SUBPART 64 · GUIDE 872

LLM Responses vs. Human Survey Data

Can Large Language Model Responses Be Treated as Human Survey Data?

Large language models can generate plausible questionnaire responses, but these outputs are not equivalent to observations collected from human participants. Researchers must distinguish model-generated answers from empirical survey data, particularly when reporting sample sizes, estimating population characteristics, or interpreting statistical significance. Learn when LLM responses can be used legitimately and which scientific claims require genuine human evidence.

PART 0D · SUBPART 64 · GUIDE 871

AI Prediction of Human Survey Responses

Can Generative AI Predict How a Human Population Would Respond to a Survey?

Generative AI can produce survey responses that resemble selected patterns observed in human populations, but predicting actual population attitudes requires empirical validation. Learn how AI-based survey simulations work, why demographic prompting does not guarantee representativeness, and how researchers should evaluate prediction accuracy, bias, and generalizability.

PART 0D · SUBPART 64 · GUIDE 870

AI-Generated vs. Real Behavioral Research Participants

Can AI-Generated “Participants” Replace Real Participants in Behavioral Research?

AI-generated participants can simulate selected human responses and support methodological experiments, but they cannot automatically replace real participants in behavioral research. Their outputs may reproduce some empirical patterns while failing to capture human experiences, contextual influences, and population diversity. Learn when simulated participants are useful and what evidence is required before drawing conclusions about actual people.

PART 0D · SUBPART 64 · GUIDE 869

AI-Simulated Research Participants

Can Generative AI Simulate Human Research Participants?

Generative AI can simulate responses from hypothetical research participants, including survey answers, interview narratives, and behavioral decisions. However, these outputs reflect model behavior rather than genuine human participation. Learn when AI-simulated participants can support methodological research, how their responses should be evaluated, and why realistic language does not establish empirical validity.

PART 0D · SUBPART 63 · GUIDE 868

Reporting Synthetic Data in Research Papers

How Should the Use of Synthetic Data Be Reported in a Research Paper?

Synthetic data must be reported transparently so readers can distinguish artificially generated observations from genuine empirical evidence. Researchers should explain why synthetic data were used, how they were generated, which validation procedures were performed, and what conclusions the data can support. Learn how to document synthetic-data methods without overstating their scientific significance.

PART 0D · SUBPART 63 · GUIDE 867

Labeling Synthetic Research Data

Should Synthetic Data Be Clearly Labeled in a Research Dataset?

Synthetic data should be clearly labeled so researchers, collaborators, and future users can distinguish artificial records from genuine observations. Learn when dataset-level labels are sufficient, when record-level indicators are needed, and how documentation and provenance help prevent misleading analyses and research-integrity problems.

PART 0D · SUBPART 63 · GUIDE 866

Validating Synthetic Research Data

How Do Researchers Validate Synthetic Data?

Synthetic data validation determines whether artificially generated records are suitable for a particular research purpose. Statistical similarity alone is insufficient. Researchers may need to examine analytical utility, subgroup representation, privacy risks, and the reliability of the generation process. Learn how to design a defensible validation strategy and avoid overstating what synthetic data can establish.

PART 0D · SUBPART 63 · GUIDE 865

Synthetic Data and Participant Privacy Leakage

Can Synthetic Data Still Leak Information About Real Research Participants?

Synthetic data can reduce the exposure of genuine research records, but they may still reveal information about individuals represented in the original dataset. Generative models can memorize records, reproduce distinctive attributes, or permit membership and attribute inference. Learn how privacy leakage occurs, how researchers can evaluate the risks, and why synthetic generation does not automatically guarantee anonymity.

PART 0D · SUBPART 63 · GUIDE 864

Synthetic Data and Research Participant Privacy

Can Synthetic Data Protect Research Participant Privacy?

Synthetic data may help protect research participant privacy by reducing the need to share original individual-level records. However, artificial generation does not automatically guarantee anonymity or prevent information leakage. Learn how synthetic data can support responsible data sharing, which privacy risks remain, and what researchers should evaluate before releasing generated datasets.

PART 0D · SUBPART 63 · GUIDE 863

Synthetic Data for Prediction Models

Can Synthetic Data Be Used to Develop Prediction Models?

Synthetic data can support prediction-model development by providing artificial training examples, testing analytical pipelines, and representing selected scenarios. However, models trained on generated data may learn artificial patterns that do not generalize to real populations. Learn when synthetic data are useful, how to evaluate their contribution, and why independent empirical validation remains essential.

PART 0D · SUBPART 63 · GUIDE 862

Testing Research Code With Synthetic Data

Can Synthetic Data Be Used to Test Research Analysis Code?

Synthetic data can help researchers test analytical code before working with genuine research observations. By generating artificial datasets with known characteristics and expected results, researchers can identify programming errors, verify data transformations, and improve reproducibility. Learn how to design meaningful tests and recognize what synthetic data cannot establish.

PART 0D · SUBPART 63 · GUIDE 861

Can Synthetic Data Replace Real Participants?

Can Synthetic Data Replace Real Participant Data in Research?

Synthetic data can replace real participant records for certain methodological tasks, but they generally cannot substitute for genuine human participation when a study aims to understand actual experiences, attitudes, or behaviors. Learn where synthetic data are useful, what claims they can support, and why empirical validation remains important.

PART 0D · SUBPART 63 · GUIDE 860

Legitimate Synthetic Data in Research

When Is the Use of Synthetic Data Legitimate in Research?

Synthetic data can be legitimate in research when artificial observations serve a justified scientific purpose, are validated for their intended use, and are accurately reported. Understand when synthetic datasets can support methodological development, simulation, and privacy-related applications, and when they cannot substitute for genuine empirical evidence.

PART 0D · SUBPART 63 · GUIDE 859

Synthetic Data vs. Fabricated Data

Synthetic Data vs. Fabricated Data: What’s the Difference?

Synthetic data and fabricated data may both contain artificially generated observations, but they differ fundamentally in how they are represented and used. Synthetic data can support legitimate research methods, while fabrication involves making up data or results and reporting them as genuine evidence. Learn how to distinguish the two and avoid misleading scientific claims.

PART 0D · SUBPART 63 · GUIDE 858

Synthetic Data vs. Simulated Data

Synthetic Data vs. Simulated Data: What’s the Difference?

Synthetic data and simulated data are both artificially generated, but the terms often emphasize different methodological purposes. Synthetic data commonly reproduce selected properties of existing datasets, while simulated data often represent outcomes generated under specified models or scenarios. Learn where the concepts overlap and why the distinction matters for research design and interpretation.

PART 0D · SUBPART 63 · GUIDE 857

Synthetic Data and AI Generation

What Are Synthetic Data, and How Can AI Generate Them?

Synthetic data are artificially generated records designed to represent selected characteristics or scenarios. AI can produce them using generative models, learned statistical relationships, or language-based generation. Understand the main approaches, research applications, validation requirements, and limitations of AI-generated synthetic datasets.

PART 0D · SUBPART 62 · GUIDE 856

Human Accountability for AI Research Misconduct

Why Does Human Accountability Remain Central When AI Performs a Prohibited Research Action?

AI may generate fabricated data, alter scientific evidence, or perform prohibited research actions, but responsibility does not automatically transfer to the technology. Learn how human accountability operates, why intention and control matter, and how researchers and institutions should govern AI-assisted workflows.

PART 0D · SUBPART 62 · GUIDE 855

Detecting AI-Assisted Research Falsification

Can AI-Assisted Research Falsification Be Detected Reliably?

AI-assisted research falsification cannot be detected reliably through any single tool or statistical test. Although forensic analysis, reproducibility checks, and provenance records can reveal inconsistencies, establishing falsification requires evidence of how research materials, data, or results were manipulated and represented. Learn what detection methods can establish and where their limitations remain.

PART 0D · SUBPART 62 · GUIDE 854

Synthetic Data vs. Real Research Observations

How Must Synthetic Data Be Distinguished From Real Research Observations?

Synthetic data and real research observations may look similar, but they have different evidentiary origins. Researchers must distinguish them through accurate labeling, provenance records, methodological documentation, and appropriate reporting so that artificial records are never mistaken for observations that actually occurred.

PART 0D · SUBPART 62 · GUIDE 853

Legitimate AI-Generated Synthetic Research Data

Can Generative AI Be Used Legitimately to Generate Synthetic Research Data?

Generative AI can legitimately create synthetic research data for software testing, methodological development, simulation, and certain privacy-related applications. However, artificial observations must be clearly distinguished from empirical evidence, validated for their intended purpose, and reported with appropriate methodological limitations.

PART 0D · SUBPART 62 · GUIDE 852

AI vs. Manual Research Data Fabrication

Is Using AI to Fabricate Research Data Different From Fabricating Data Manually?

Using AI to fabricate research data does not make the conduct ethically different from inventing data manually. Both misrepresent observations that were never collected. However, AI can change the scale, appearance, and traceability of fabricated material, creating additional challenges for research verification and accountability.

PART 0D · SUBPART 62 · GUIDE 851

AI and Research Fabrication Detection

Can Generative AI Make Fabricated Research Harder to Detect?

Generative AI may make fabricated research more difficult to recognize by producing plausible datasets, scientific images, and textual evidence. However, no universal detector can establish authenticity or fabrication. Learn why provenance, independent verification, and appropriate investigation matter more than appearance alone.

PART 0D · SUBPART 62 · GUIDE 850

AI-Generated Fabricated Research Datasets

Can Generative AI Fabricate a Convincing Research Dataset?

Generative AI can produce research datasets that appear statistically plausible, complete with realistic variation and relationships among variables. However, convincing numerical patterns do not establish that observations were genuinely collected. Learn how fabricated datasets differ from legitimate synthetic data and why provenance matters more than appearance.

PART 0D · SUBPART 62 · GUIDE 849

AI-Generated Laboratory and Statistical Data

Can Generative AI Create Fake Laboratory or Statistical Data?

Generative AI can produce convincing laboratory measurements, statistical tables, and experimental results without conducting experiments or analyzing genuine data. Learn when generated values constitute fabrication, how legitimate simulation differs, and what researchers should verify before reporting scientific findings.

PART 0D · SUBPART 62 · GUIDE 848

AI-Generated Fake Survey Responses

Can Generative AI Create Fake Survey Responses?

Generative AI can produce convincing survey responses, including Likert-scale ratings, demographic profiles, and open-ended answers. However, presenting artificial responses as genuine participant data constitutes fabrication. Learn how to distinguish legitimate simulation from false empirical evidence and protect the integrity of survey research.

PART 0D · SUBPART 62 · GUIDE 847

AI-Generated Participants and Fake Interviews

Can Generative AI Create Fake Research Participants or Interview Transcripts?

Generative AI can produce convincing fictional participants, interview transcripts, and quotations, but these outputs are not evidence of actual human experiences. Learn when simulated interviews are legitimate, when their use constitutes fabrication, and how researchers can preserve qualitative data integrity.