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 801–850 of 5,065 guides

5,065 guides
PART 0D · SUBPART 26 · GUIDE 396

AI Coding Changes Across Prompt Reruns

Can AI Coding Change When the Same Prompt Is Rerun?

Generative AI may assign different codes to the same research response even when the prompt is unchanged. Learn why coding decisions vary across runs, how to measure their stability, and what researchers should document before relying on repeated AI classifications.

PART 0D · SUBPART 26 · GUIDE 395

Systematic Bias in AI Research Coding

Can AI Coding Introduce Systematic Bias Into Research Data?

AI coding can systematically misclassify particular experiences, languages, or respondent groups, potentially distorting research findings even when overall accuracy appears high. Learn how these biases arise, how to detect them, and what researchers can do before using AI-coded data to support conclusions.

PART 0D · SUBPART 26 · GUIDE 394

Manual Verification of AI-Coded Data

How Much AI-Coded Research Data Should Be Manually Checked?

Should researchers manually check 10%, 20%, or every AI-coded response? There is no universally valid percentage. Learn how to design a verification sample, identify consequential coding errors, and determine whether the remaining AI-coded data can support defensible research conclusions.

PART 0D · SUBPART 26 · GUIDE 393

AI Coding Without Human Review

Can AI Coding Be Used in Research Without Human Review?

Researchers may not need to manually recode every AI-classified response, but eliminating human oversight entirely raises substantial methodological concerns. Learn when automated coding may be defensible, how to evaluate its performance, and why human-approved AI outputs are not automatically valid.

PART 0D · SUBPART 26 · GUIDE 392

AI Coding Confidence Scores

Should AI Provide Confidence Scores for Research Coding Decisions?

AI confidence scores may help researchers prioritize uncertain coding decisions, but a reported confidence of 90% does not necessarily mean a 90% probability of correctness. Learn how to interpret, calibrate, and evaluate confidence scores before using them to guide human review.

PART 0D · SUBPART 26 · GUIDE 391

AI Detection of Ambiguous Coding Cases

Can Generative AI Identify Ambiguous Cases During Data Coding?

Generative AI can help researchers identify responses that are difficult to classify, including ambiguous statements, overlapping categories, and insufficient contextual evidence. Learn how to distinguish genuine ambiguity from model uncertainty and develop a defensible process for reviewing difficult coding decisions.

PART 0D · SUBPART 26 · GUIDE 390

AI Coding Consistency vs. Validity

Does Consistent AI Coding Mean the Coding Is Valid?

Generative AI may assign the same codes repeatedly without interpreting the data correctly. Learn why coding consistency differs from validity, what agreement statistics can and cannot establish, and how to evaluate whether AI-generated classifications represent the concepts your research intends to measure.

PART 0D · SUBPART 26 · GUIDE 389

AI Consistency in Large-Scale Text Coding

Can Generative AI Code Thousands of Text Responses Consistently?

Generative AI can process thousands of textual responses, but processing more data does not guarantee consistent coding. Learn how batch size, coding instructions, model variation, and systematic errors affect large-scale AI classification, and how researchers can evaluate the reliability of their results.

PART 0D · SUBPART 26 · GUIDE 388

AI Application of Research Codebooks

Can Generative AI Apply a Human-Developed Research Codebook?

Generative AI can apply a human-developed codebook to interviews, survey responses, and other textual data, but supplying coding definitions does not guarantee accurate classification. Learn how to prepare a codebook for AI use, handle ambiguous cases, and evaluate whether the model follows your coding rules.

PART 0D · SUBPART 26 · GUIDE 387

AI-Generated Coding Categories

Can Generative AI Create Coding Categories Automatically?

Generative AI can suggest coding categories from qualitative research data, but the resulting labels may overlap, omit important meanings, or reflect assumptions imposed by the model. Learn how to evaluate, refine, and document AI-generated categories without confusing automated classification with qualitative interpretation.

PART 0D · SUBPART 26 · GUIDE 386

AI Categorization of Survey Responses

Can Generative AI Categorize Open-Ended Survey Responses?

Generative AI can organize open-ended survey responses into meaningful categories, but accurate classification requires more than identifying similar words. Learn how to design a coding workflow, handle overlapping responses, and verify AI-generated categories before analyzing your results.

PART 0D · SUBPART 26 · GUIDE 385

AI-Assisted Research Data Coding

Can Generative AI Help Code Research Data?

Generative AI can help researchers identify, organize, and apply codes to qualitative data, but its outputs require careful validation. Learn what AI-assisted coding can accomplish, where it may introduce errors, and how to use it responsibly in research.

PART 0D · SUBPART 25 · GUIDE 384

Verifying AI Data Cleaning Changes

How Do Researchers Verify Every Change Made or Suggested by AI During Data Cleaning?

AI-generated data corrections should never be accepted solely because a tool reports successful completion. Learn how to verify proposed and executed changes, detect unintended modifications, test cleaning scripts, and confirm that the final dataset remains faithful to the original research evidence.

PART 0D · SUBPART 25 · GUIDE 383

Researcher-Specified Data Cleaning Decisions

Which Data-Cleaning Decisions Must Be Explicitly Specified by the Researcher?

AI can execute data-cleaning procedures, but researchers must define the rules governing corrections, exclusions, missing values, and substantive transformations. Learn which decisions require explicit approval, how to specify them, and why leaving methodological choices to AI can compromise research findings.

PART 0D · SUBPART 25 · GUIDE 382

AI Data Cleaning Audit Trails

How Do Researchers Create an Audit Trail for AI-Assisted Data Cleaning?

An audit trail makes AI-assisted data cleaning traceable by recording what changed, why it changed, who approved it, and how the result was verified. Learn how to document transformations, preserve dataset versions, and reconstruct the cleaning process without relying on chatbot conversations alone.

PART 0D · SUBPART 25 · GUIDE 381

AI Access to the Master Research Dataset

Should AI Be Allowed to Modify the Master Research Dataset Directly?

Allowing AI to modify the master research dataset can introduce undocumented errors, unintended deletions, and irreversible changes. Learn when AI should have read-only access, how to implement approved transformations safely, and why protecting original observations is essential for reproducible research.

PART 0D · SUBPART 25 · GUIDE 380

AI for Distinguishing Data Errors

Can Generative AI Distinguish a Data Error From a Legitimate Unusual Observation?

Generative AI can help researchers investigate unusual observations, but statistical anomalies are not necessarily errors. Learn how to verify suspicious values, distinguish genuine variation from incorrect records, and make defensible decisions about correction, retention, or exclusion.

PART 0D · SUBPART 25 · GUIDE 379

AI for Suspicious Research Data Patterns

Can Generative AI Detect Suspicious Patterns in Research Data?

Generative AI can help researchers identify unusual response patterns, repeated sequences, implausible timings, and statistical anomalies. Learn how to investigate these warning signs without confusing suspicious data with evidence of fabrication or research misconduct.

PART 0D · SUBPART 25 · GUIDE 378

AI for Detecting Data Entry Errors

Can Generative AI Detect Data Entry Errors?

Generative AI can help identify potential data entry errors, including invalid codes, inconsistent records, and suspicious numerical values. Learn how to distinguish suspected mistakes from confirmed errors, verify corrections against original records, and maintain an auditable dataset.

PART 0D · SUBPART 25 · GUIDE 377

Automatic AI Missing-Data Imputation

Should AI Ever Fill Missing Research Data Automatically?

AI can automate approved missing-data procedures, but it should not invent participant responses or silently overwrite original observations. Learn when automated imputation may be appropriate, which decisions require researcher approval, and how to protect the integrity of statistical findings.

PART 0D · SUBPART 25 · GUIDE 376

AI for Inferring Missing Research Data

Can Generative AI Infer Missing Values in Research Data?

Generative AI can suggest plausible values for missing research observations, but predictions are not recovered participant responses. Learn how missing-value inference works, what assumptions it requires, and why uncertainty must be preserved before estimated values enter a research analysis.

PART 0D · SUBPART 25 · GUIDE 375

AI for Cleaning Free-Text Research Data

Can Generative AI Help Clean Free-Text Research Data?

Generative AI can help prepare free-text research data by identifying transcription errors, inconsistent formatting, and potentially identifying information. Learn which changes are appropriate, why participant wording matters, and how to verify AI-assisted cleaning without compromising qualitative evidence.

PART 0D · SUBPART 25 · GUIDE 374

AI for Standardizing Text Categories

Can Generative AI Standardize Inconsistent Text Categories in Research Data?

Generative AI can help standardize inconsistent text categories by identifying spelling variations, abbreviations, and potentially equivalent labels. Learn how to verify proposed mappings, preserve meaningful distinctions, and avoid introducing classification errors into research data.

PART 0D · SUBPART 25 · GUIDE 373

AI for Data Recoding Rules

Can Generative AI Suggest Data Recoding Rules?

Generative AI can help researchers propose recoding rules, reverse-score questionnaire items, and transform variables for analysis. Learn which decisions require methodological justification, how to prevent coding errors, and how to verify every transformation.

PART 0D · SUBPART 25 · GUIDE 372

AI for Research Data Outliers

Can Generative AI Identify Potential Outliers in Research Data?

Generative AI can help identify potential outliers using statistical and logical checks, but an unusual observation is not necessarily an error. Learn how to investigate flagged values, select appropriate detection methods, and avoid deleting legitimate research data.

PART 0D · SUBPART 25 · GUIDE 371

AI for Duplicate Research Records

Can Generative AI Detect Duplicate Records in a Research Dataset?

Generative AI can help researchers identify identical records, repeated participant identifiers, and potentially duplicated submissions. Learn why matching records are not necessarily errors, how to investigate suspected duplicates, and when removing them could compromise research findings.

PART 0D · SUBPART 25 · GUIDE 370

AI for Missing and Impossible Values

Can Generative AI Identify Missing or Impossible Values in Research Data?

Generative AI can help researchers identify blank responses, invalid codes, out-of-range values, and logically inconsistent records. Learn how to establish reliable validation rules, investigate flagged observations, and avoid replacing genuine research data with unsupported assumptions.

PART 0D · SUBPART 25 · GUIDE 369

AI for Research Data Cleaning

Can Generative AI Help Clean Research Data?

Generative AI can help researchers identify inconsistencies, inspect missing values, suggest corrections, and develop reproducible data-cleaning procedures. However, researchers must distinguish AI-generated suggestions from verified errors and retain control over changes to their datasets.

PART 0D · SUBPART 24 · GUIDE 368

AI Transcript Verification Workflow

What Is a Defensible Verification Workflow for AI-Generated Research Transcripts?

A defensible AI transcript verification workflow requires more than proofreading automatically generated text. Learn how to preserve source recordings, verify participant statements, document corrections, manage unresolved passages, and establish when transcripts are ready for research analysis.

PART 0D · SUBPART 24 · GUIDE 367

De-Identifying Research Transcripts for AI

Should Research Transcripts Be De-Identified Before Further AI Processing?

Removing participant names is not enough to make research transcripts anonymous. Learn when de-identification is necessary before AI processing, how indirect identifiers create privacy risks, and how to protect participants without removing information essential to qualitative analysis.

PART 0D · SUBPART 24 · GUIDE 366

Confidential Interviews and Cloud AI Transcription

Can Confidential Research Interviews Be Uploaded to Cloud-Based AI Transcription Services?

Uploading confidential research interviews to cloud AI transcription services may be permissible, but it requires more than selecting a secure-looking platform. Learn how consent, institutional approval, vendor data practices, and legal obligations determine whether external processing is appropriate.

PART 0D · SUBPART 24 · GUIDE 365

AI Speaker Identification in Transcripts

Can AI Transcription Correctly Identify Different Speakers?

AI transcription can distinguish voices and assign speaker labels, but recognizing that two statements came from the same voice is not the same as identifying the participant. Learn how speaker identification works, why attribution errors occur, and how researchers can verify who said what.

PART 0D · SUBPART 24 · GUIDE 364

Filler Words and Pauses in Transcripts

Should Filler Words and Pauses Be Preserved in Research Transcripts?

Filler words, repetitions, and pauses are not always irrelevant speech. Whether they should remain in a research transcript depends on the analytical method and the role of conversational detail in interpreting participant meaning. Learn how to choose and apply an appropriate transcription convention.

PART 0D · SUBPART 24 · GUIDE 363

AI Transcription and Participant Meaning

Can AI Transcription Change the Meaning of a Participant’s Statement?

AI transcription can change the apparent meaning of participant statements through surprisingly small errors, including omitted negations, altered qualifications, and incorrect punctuation. Learn how these distortions affect qualitative analysis and how researchers can distinguish faithful transcription from unintended rewriting.

PART 0D · SUBPART 24 · GUIDE 362

Simultaneous AI Transcription and Translation

Can AI Reliably Translate and Transcribe an Interview at the Same Time?

AI can transcribe and translate research interviews in a single workflow, but the resulting text may contain both speech recognition and translation errors. Learn when simultaneous processing is appropriate, why original-language records matter, and how to verify translated interviews without losing participant meaning.

PART 0D · SUBPART 24 · GUIDE 361

AI Transcription of Multilingual Interviews

Can AI Transcription Reliably Handle Multilingual Interviews?

AI can transcribe interviews conducted in multiple languages, but code-switching, language detection errors, and unequal recognition accuracy can affect the resulting transcripts. Learn how to evaluate multilingual transcription systems while preserving participants' original language and meaning.

PART 0D · SUBPART 24 · GUIDE 360

AI Transcription With Multiple Speakers

Can AI Transcription Reliably Handle Multiple Speakers?

AI can transcribe conversations involving multiple speakers, but overlapping speech and speaker attribution create additional challenges. Learn how multi-speaker transcription works, where errors occur, and how researchers can preserve the accuracy of group discussions and interviews.

PART 0D · SUBPART 24 · GUIDE 359

AI Transcription and Accents

Can AI Transcription Accurately Handle Different Accents?

AI transcription accuracy can vary across accents, dialects, and speaker populations. Learn what published research reveals about these differences, why accent-related errors matter for qualitative research, and how to evaluate transcription systems without misrepresenting participants.

PART 0D · SUBPART 24 · GUIDE 358

AI Transcription of Technical Terms

Can AI Transcription Accurately Handle Technical Terminology?

AI transcription systems can recognize many technical terms, but unfamiliar vocabulary, acronyms, and discipline-specific expressions remain sources of error. Learn why these mistakes occur, how they can affect research findings, and how to verify specialized terminology without changing participant meaning.

PART 0D · SUBPART 24 · GUIDE 357

Manual Verification of AI Transcripts

Should AI-Generated Research Transcripts Always Be Checked Manually?

AI-generated transcripts can contain errors that alter participant meaning, even when the text appears accurate. Learn why manual verification matters, when full checking is preferable, and whether selective review can ever be justified.

PART 0D · SUBPART 24 · GUIDE 356

AI Transcription Accuracy

How Accurate Is AI Transcription for Research Interviews?

AI transcription accuracy varies considerably across recordings, speakers, and speech recognition systems. Learn how accuracy is measured, why a high percentage can be misleading, and what researchers should evaluate before trusting an automated transcript.

PART 0D · SUBPART 24 · GUIDE 355

AI Transcription for Research Interviews

Can AI Reliably Transcribe Research Interviews?

AI can substantially reduce the time required to transcribe research interviews, but automated transcripts may contain errors that affect interpretation. Learn when AI transcription is appropriate, why accuracy varies, and how to use it without compromising research integrity.

PART 0D · SUBPART 23 · GUIDE 354

When Not to Use AI for Data Collection

When Is AI-Mediated Data Collection Inappropriate for Research?

AI-mediated data collection is not suitable for every research study. Learn when methodological requirements, participant vulnerability, sensitive disclosures, privacy constraints, or inadequate validation make automated interviewing inappropriate, and when human-led methods are preferable.

PART 0D · SUBPART 23 · GUIDE 353

Logging AI-Participant Research Interactions

What Should Researchers Log From an AI-Mediated Participant Interaction?

AI-mediated research requires more than saving participant answers. Researchers may need records of the questions actually asked, model configurations, automated decisions, protocol deviations, and safety events. Learn what to log, what to avoid collecting, and how to balance research reproducibility with participant privacy.

PART 0D · SUBPART 23 · GUIDE 352

Validating AI-Mediated Data Collection

How Do Researchers Validate an AI-Mediated Data Collection Process?

Validating AI-mediated data collection requires more than testing whether a chatbot works. Researchers must evaluate protocol adherence, question quality, response accuracy, measurement effects, participant safeguards, and reproducibility before relying on AI-collected evidence.

PART 0D · SUBPART 23 · GUIDE 351

Combining AI and Human-Collected Data

Can AI-Collected Research Data Be Combined With Human-Collected Data?

Research data collected through AI and human interviewers can sometimes be combined, but using the same questions does not guarantee comparable evidence. Learn how to assess interviewer-mode effects, measurement equivalence, and the methodological consequences of pooling data collected through different procedures.

PART 0D · SUBPART 23 · GUIDE 349

Imminent Harm Disclosures to AI Interviewers

What Happens if a Participant Discloses Imminent Harm to an AI Interviewer?

An AI interviewer may receive a disclosure suggesting imminent harm to the participant or another person. Researchers need predefined procedures for stopping automated questioning, assessing the limits of confidentiality, and arranging timely human intervention without relying on the chatbot to manage an emergency independently.

PART 0D · SUBPART 23 · GUIDE 348

AI Interviewers and Participant Distress

Can an AI Interviewer Respond Inappropriately to Participant Distress?

AI interviewers may fail to recognize participant distress, respond with inappropriate reassurance, or continue probing when an interview should stop. Understand these risks and the safeguards researchers need before using AI to interview human participants.

PART 0D · SUBPART 23 · GUIDE 347

Measurement Bias in AI Interviews

Can AI-Mediated Interviewing Introduce Systematic Measurement Bias?

AI-mediated interviews can introduce systematic measurement bias when chatbot behavior influences what participants disclose or how their answers are interpreted. Learn how unequal probing, leading questions, and interviewer effects can distort research findings, and how researchers can detect and reduce these risks.