Qualitative and Mixed-Methods Research

Introduction and Context

Research in artificial intelligence and machine learning is often associated with numerical datasets, predictive models, benchmark scores, and statistical tests. Yet many important research questions cannot be answered adequately through quantitative evidence alone. Questions such as “How do clinicians understand an AI-assisted diagnosis system?”, “Why do data scientists distrust a model-monitoring tool?”, or “How does a community experience automated decision-making?” concern meanings, experiences, practices, institutional contexts, and social consequences.

Qualitative research provides systematic approaches for studying these issues. It investigates how people interpret events, how practices develop in context, and how meanings are negotiated. Mixed-methods research combines qualitative and quantitative approaches within a coherent research design. It is particularly valuable in AI and ML, where technical performance must often be considered alongside usability, trust, fairness, adoption, safety, and organizational impact.

This chapter explains interviews, focus groups, observation, document analysis, coding, thematic analysis, triangulation, credibility, reflexivity, and the integration of qualitative and quantitative evidence. The emphasis is not on treating qualitative research as an informal version of quantitative research, but on understanding its distinctive logic and standards of quality.

Learning Objectives

By the end of this chapter, you should be able to:

1. Explain the assumptions and purposes of qualitative and mixed-methods research.

2. Select appropriate qualitative data-collection methods for a research problem.

3. Design and conduct interviews, focus groups, observations, and document analysis ethically.

4. Organize, code, and interpret qualitative data using thematic analysis.

5. Evaluate credibility, dependability, confirmability, and transferability.

6. Explain reflexivity and identify the researcher’s influence on a study.

7. Distinguish major mixed-methods designs and integration strategies.

8. Apply triangulation without assuming that different data sources must produce identical findings.

9. Integrate qualitative and quantitative evidence in AI and ML research.

1. Foundations of Qualitative Research

1.1 The nature of qualitative inquiry

Qualitative research studies phenomena through participants’ meanings, language, actions, interactions, documents, and contexts. Data may include interview transcripts, field notes, photographs, policy documents, discussion-board posts, system logs interpreted alongside human accounts, or observations of work practices.

Qualitative inquiry is commonly associated with several assumptions:

• Contextuality: Human behavior is shaped by social, institutional, cultural, and technological settings.
• Meaning: Participants interpret events, and these interpretations are central evidence rather than noise.
• Process: Researchers may examine how practices change over time rather than measuring only final outcomes.
• Inductive and abductive reasoning: Concepts may emerge from data, while researchers also move iteratively between evidence and existing theory to develop the most plausible explanation.
• Researcher involvement: The researcher is an instrument of data collection and interpretation. Their identity, assumptions, relationships, and decisions influence the study.
• Depth over breadth: The aim is often detailed understanding rather than population estimation.

Qualitative research is not necessarily unstructured. A rigorous qualitative study has a clear question, justified sampling, a documented procedure, an analytic strategy, ethical safeguards, and an evidence-based account of findings.

1.2 Paradigms and methodological alignment

A research paradigm is a broad set of assumptions about reality, knowledge, and how knowledge should be produced. Interpretivist and constructivist approaches emphasize that social reality is understood through meanings and interactions. Critical approaches examine power, inequality, exclusion, and the consequences of apparently neutral systems. Pragmatism focuses on the research question and supports using methods that provide useful answers, including combinations of qualitative and quantitative evidence.

These paradigms are not rigid labels. However, researchers should maintain alignment among their research question, assumptions, methods, analysis, and claims. For example, a study claiming to understand how marginalized users experience an automated welfare system should not rely only on a decontextualized satisfaction score. It should examine participants’ accounts, institutional processes, and possibly documentary or observational evidence.

1.3 Sampling in qualitative research

Qualitative sampling is usually purposive rather than based on random selection. Participants are selected because they can provide relevant and information-rich perspectives. Common strategies include:

• Criterion sampling: Selecting all participants who meet specified conditions, such as having used an AI tutoring system for at least one semester.
• Maximum-variation sampling: Seeking diverse cases to examine different experiences across roles, locations, or levels of expertise.
• Typical-case sampling: Selecting cases that represent ordinary circumstances.
• Critical-case sampling: Selecting cases whose findings may have particularly important implications.
• Theoretical sampling: In an evolving analysis, selecting additional participants or cases to develop, refine, or challenge emerging concepts.
• Snowball sampling: Asking participants or organizations to recommend others, often useful for hard-to-reach populations but vulnerable to network bias.

Sample adequacy is not determined by a universal numerical rule. Researchers consider the study’s purpose, scope, participant diversity, data richness, and whether additional data are contributing substantially to the developing analysis. The term thematic sufficiency is often more appropriate than mechanically claiming “saturation,” especially when the study does not seek to develop a formal theory.

2. Qualitative Data-Collection Methods

2.1 Interviews

An interview is a purposeful conversation designed to elicit participants’ experiences, interpretations, knowledge, or reasoning. Interviews may be:

• Structured: All participants receive the same questions in the same order.
• Semi-structured: The researcher uses an interview guide while allowing probing and follow-up questions.
• Unstructured or conversational: The discussion is guided broadly by the research topic.

Semi-structured interviews are common in university research because they balance comparability and flexibility. A good interview guide moves from accessible questions to more specific or sensitive issues. Questions should be open-ended and neutral. “Can you describe how you used the recommendation system?” is generally better than “Did the recommendation system improve your work?” The latter presupposes a particular type of effect.

Useful probes include “Can you give an example?”, “What happened next?”, “How did that affect your decision?”, and “Were there circumstances in which your experience was different?” Researchers should distinguish between questions that elicit description and those that invite evaluation or explanation.

Interview quality depends on preparation, rapport, attentive listening, appropriate pacing, accurate recording, and careful transcription. Consent should cover recording, data storage, confidentiality, withdrawal conditions, and possible quotations. Researchers should not promise absolute anonymity when combinations of role, organization, and event could identify a participant.

2.2 Focus groups

A focus group is a moderated discussion with several participants who respond to questions and to one another. The interaction is itself data: participants may agree, disagree, explain shared norms, or reveal differences between professional groups.

Focus groups are useful for exploring language, collective understandings, product concepts, and contested issues. They are less suitable when the topic is highly sensitive, when participants have unequal power, or when confidentiality cannot reasonably be protected. A senior manager and junior employees, for example, may not speak freely in the same group.

The moderator should establish ground rules, manage dominant speakers, invite quieter participants, and avoid turning the session into a debate. Group composition should be considered carefully. Homogeneous groups may support openness, while contrasting groups may reveal differences in perspective. Analysis should attend not only to what was said but also to agreement, silence, hesitation, interruption, and interactional patterns.

2.3 Observation

Observation examines behavior and practices in their setting. It may be:

• Participant or non-participant: Depending on the researcher’s involvement.
• Overt or covert: Overt observation discloses the researcher’s role and is normally required in ethical institutional research; covert observation raises serious ethical concerns.
• Structured or open-ended: Structured observation uses a predefined schedule, whereas open-ended observation records broader activities and context.

In AI and ML research, observation can reveal workarounds that users do not mention in interviews. For example, a radiologist may verbally describe trusting an AI tool but routinely consult a colleague before accepting its output. Field notes should distinguish observed events from the researcher’s interpretations. A useful format records date, setting, participants or roles, activity, direct descriptions, preliminary interpretations, and questions for follow-up.

Researchers must consider privacy, access, reactivity, and safety. People may alter their behavior when observed. Sensitive workplaces may require organizational permission as well as individual consent, but institutional permission does not replace individual rights.

2.4 Document and digital-material analysis

Documents include policies, meeting minutes, technical manuals, incident reports, user complaints, model cards, system prompts, public statements, and regulatory guidance. Digital materials may include forum discussions, support tickets, or social-media posts. Documents should be treated as situated productions rather than transparent records of reality. Researchers should ask who created the document, for what purpose, for which audience, under what constraints, and what is omitted.

A document-analysis protocol may record document type, date, author or institution, intended audience, relevant passages, assumptions, representations of stakeholders, and links to other evidence. Public availability does not automatically make digital material ethically risk-free. Quoting a searchable post can expose its author even when a name is removed. Researchers should assess expectations of privacy, sensitivity, platform rules, and the possibility of re-identification.

3. From Data to Findings

3.1 Preparing and managing data

Data management begins before analysis. Researchers should create consistent file names, maintain a data inventory, separate identifying information from research data, record versions of transcripts and codebooks, and secure files according to institutional requirements. Automated transcription can save time but must be checked against audio, especially for accents, technical terminology, names, and overlapping speech. AI-assisted transcription or summarization should not be treated as a substitute for researcher verification.

A transcript may be verbatim, lightly edited for readability, or selectively transcribed. The choice should match the research question. Conversation analysis requires detailed representation of pauses and turn-taking; a thematic study may use a less detailed transcript. The method should be reported transparently.

3.2 Coding

Coding assigns a concise label to a segment of data that represents a feature relevant to the research question. A code may be descriptive, such as “difficulty locating explanation,” or interpretive, such as “accountability shifted to the user.” Coding is not merely highlighting interesting quotations; it is a systematic bridge between raw data and analytical claims.

A practical coding process includes:

1. Familiarization: Read or review the data repeatedly and record initial observations.

2. Initial coding: Code relevant segments, allowing multiple codes where appropriate.

3. Code development: Define codes, clarify boundaries, merge duplicates, and identify relationships.

4. Focused coding: Give greater attention to codes that are frequent, significant, theoretically relevant, or explanatory.

5. Category development: Group related codes into broader categories.

6. Theme development: Construct patterns of shared meaning that answer the research question.

7. Review and refinement: Test themes against the full dataset and revise them.

8. Writing and evidencing: Explain the interpretation with carefully selected quotations and contextual detail.

A codebook commonly contains the code name, definition, inclusion criteria, exclusion criteria, example quotation, and related codes. Software such as NVivo, ATLAS.ti, MAXQDA, or qualitative functions in general data tools can support retrieval and organization. Software does not perform interpretation automatically and does not make a study rigorous by itself.

3.3 Thematic analysis

Thematic analysis identifies and interprets patterned meanings across a dataset. A theme is more than a topic. “Training” may be a topic; “Training made responsibility visible but did not make the system understandable” is a possible interpretive theme. Themes should have internal coherence, meaningful distinction from one another, and clear relevance to the research question.

Researchers should decide whether the analysis is primarily semantic, attending to explicit statements, or latent, examining underlying assumptions and structures. They should also state whether themes are developed inductively, deductively, or through a combination. A deductive analysis may use a framework such as trust, autonomy, or fairness; an inductive analysis may allow unexpected concerns to emerge.

A strong thematic report connects claims to evidence while preserving variation. Researchers should not present only quotations that support their preferred interpretation. Negative cases, contradictory accounts, and minority perspectives can refine or qualify a theme.

3.4 Illustrative coding example

Suppose three participants discuss an AI-based student-advising tool:

• “It gives an answer quickly, but I still check the policy before acting.”
• “Students think the system is official, even when it is only a suggestion.”
• “When the recommendation is wrong, nobody knows whether to blame the data or the adviser.”

Possible initial codes include speed valued, verification required, authority misinterpreted, recommendation versus decision, and unclear accountability. These may contribute to a broader theme such as Efficiency without settled responsibility. The theme is not simply a summary of the word “trust”; it interprets how speed, authority, verification, and accountability interact.

4. Quality, Credibility, and Reflexivity

4.1 Quality criteria

Qualitative quality is often discussed through four related criteria:

• Credibility: The findings are plausible and well supported by the data and research process.
• Dependability: The process is logical, traceable, and documented sufficiently for readers to understand how conclusions were reached.
• Confirmability: Findings are grounded in the data rather than being arbitrary expressions of the researcher’s preferences.
• Transferability: Readers can judge whether insights may be relevant in another context because the study provides sufficiently rich contextual description.

These criteria do not require pretending that qualitative research is free from interpretation. They require transparent, disciplined interpretation.

Practical techniques include prolonged engagement where feasible, triangulation, peer debriefing, participant feedback, negative-case analysis, reflexive journaling, an audit trail, and thick description. Member checking may involve asking participants to comment on a transcript, summary, or interpretation. It is not a vote that determines whether the researcher’s analysis is correct; participants may disagree for many reasons, and the researcher must explain how feedback was handled.

4.2 Reflexivity

Reflexivity is the continuing examination of how the researcher’s position, assumptions, relationships, and decisions shape the study. A researcher who designs an AI system may interpret user criticism differently from an independent evaluator. Professional status, language, gender, institutional affiliation, technical expertise, and funding relationships may affect access and interaction.

A reflexive journal can document expectations before data collection, emotional responses, changes in interpretation, difficult interactions, analytic decisions, and evidence that challenges initial assumptions. Reflexivity is not a confessional exercise. Its purpose is to make the production of knowledge more transparent and critically examined.

4.3 Triangulation

Triangulation compares perspectives, methods, investigators, theories, or data sources. For example, a study of an AI hiring tool might compare interviews with recruiters, observations of workflow, system documentation, and hiring records. Convergence can strengthen confidence, but divergence is also informative. Recruiters may describe a system as advisory while documents or observations show that its output effectively determines shortlists.

Triangulation should not be used as a simplistic rule that three sources make a finding true. Different sources answer different questions and may represent different realities. The researcher should explain why sources were selected, what each contributes, and how agreements and contradictions were interpreted.

5. Mixed-Methods Research

5.1 Purpose and logic

Mixed-methods research intentionally integrates qualitative and quantitative approaches in one study or program of inquiry. It is appropriate when a single form of evidence cannot adequately address the research problem. Qualitative data may explain why a quantitative pattern exists; quantitative data may indicate the distribution or prevalence of experiences identified qualitatively.

Mixed methods should not mean merely placing interviews and a survey in the same report. Integration must occur in the research question, sampling, data collection, analysis, interpretation, or reporting.

5.2 Major designs

Convergent design: Qualitative and quantitative data are collected during a similar phase, analyzed separately, and then compared or integrated. This is useful when the researcher wants complementary perspectives on the same issue.

Explanatory sequential design: Quantitative data are collected and analyzed first, followed by qualitative research to explain surprising, important, or unclear results. For example, a survey may show low trust among experienced users, after which interviews investigate why.

Exploratory sequential design: Qualitative research comes first and informs a later quantitative instrument, intervention, or test. Interviews with students may generate items for a questionnaire measuring perceptions of an AI tutor.

Embedded design: One method is nested within a larger study. For example, a randomized evaluation of an AI feedback system may embed interviews to examine how participants experienced the intervention.

Multiphase design: Several connected qualitative and quantitative studies address a broad research program over time.

The design should be selected because it fits the problem, not because mixed methods appears more sophisticated.

5.3 Integration strategies

Integration may occur through:

• Connecting: Results from one phase determine who or what is studied in the next phase.
• Building: Findings from one method inform the instrument, intervention, sampling plan, or protocol of another.
• Merging: Qualitative and quantitative results are brought together for comparison.
• Embedding: One dataset supports or contextualizes another within a larger design.

A joint display is a table or visual structure that places qualitative themes and quantitative results together. For example:

| Quantitative finding | Qualitative explanation | Integrated interpretation |

|---|---|---|

| Users with high system exposure reported lower trust | Experienced users described recurring false alerts and difficult correction procedures | Greater exposure may reveal operational weaknesses rather than automatically increasing trust |

Integration requires attention to convergence (similar findings), complementarity (different but mutually informative findings), expansion (each method addresses a different part of the problem), and discordance (findings conflict). Discordance should be investigated rather than hidden.

6. Applied Case: Evaluating an AI-Based Clinical Decision-Support Tool

This is an illustrative case.

Background

A hospital introduces an AI tool that flags patients who may be at risk of deterioration. In an initial quantitative evaluation, the tool shows acceptable predictive performance and a moderate reduction in response time. However, adoption differs substantially between wards.

Problem

The hospital wants to know not only whether the tool performs well statistically, but why staff use it inconsistently and whether it changes clinical responsibility or communication.

Analysis and approach

The research team adopts an explanatory sequential mixed-methods design. First, it analyzes usage logs and a staff survey measuring frequency of use, perceived usefulness, confidence, and workload. The results show that intensive-care staff use the tool frequently, while general-ward nurses often ignore alerts. Survey responses alone do not explain the difference.

The team then purposively samples nurses, physicians, and ward managers from both types of ward. It conducts semi-structured interviews, observes handover meetings and alert-review practices, and analyzes training materials, escalation policies, and system documentation. Interviews are recorded with consent, transcribed, and coded. The team keeps a reflexive journal because two researchers participated in the tool’s implementation project.

Thematic analysis identifies four themes:

1. Alerts compete with existing priorities: General-ward nurses receive alerts during periods of high workload and cannot easily distinguish urgent alerts from routine notifications.

2. Clinical meaning is not operational meaning: Staff understand the prediction as a probability, but the system interface does not explain what action is expected.

3. Trust is relational: Users trust alerts more when a senior clinician explains the tool and responds constructively to false positives.

4. Responsibility remains ambiguous: Staff fear that following an incorrect alert may be criticized, while ignoring it may also be criticized.

Observation reveals that general-ward staff frequently copy alert information into handover notes but do not discuss it during rounds. Document analysis shows that the official policy describes the tool as advisory, whereas training slides use language suggesting that staff should respond to every alert.

Integration and outcome

The integrated interpretation is that low adoption is not primarily a rejection of predictive analytics. It results from a mismatch between alert volume, workflow, explanation, training, and accountability. The quantitative pattern identifies where the problem is concentrated; qualitative evidence explains how the problem is produced in practice.

The hospital revises alert thresholds in collaboration with staff, adds action-oriented explanations, clarifies escalation responsibility, and introduces ward-specific training. A later evaluation combines usage metrics with interviews and observations. Alert response improves, but the team continues monitoring unintended effects, including alert fatigue and unequal access to specialist support.

Learning points

• Technical performance does not establish practical usefulness.
• Mixed methods can connect distributional evidence with contextual explanation.
• Contradictions between policy documents and training materials are analytically important.
• Trust depends on institutional arrangements, not only interface design.
• Ethical evaluation must consider workload, accountability, and patient consequences.

7. Ethical, Governance, and Professional Issues

Qualitative and mixed-methods research involves responsibilities concerning autonomy, privacy, fairness, and harm. Researchers should obtain informed consent in language participants can understand, avoid coercive recruitment, and explain foreseeable risks. Participants should know whether quotations will be used, how recordings will be stored, and who may access the data.

Confidentiality is particularly difficult in small organizations or specialized professions. Removing names may not prevent identification through job title, incident, or quotation. Researchers should minimize unnecessary personal data, use controlled access, establish retention and deletion procedures, and comply with applicable institutional and data-protection requirements.

AI tools introduce additional concerns. Automated transcription may transfer sensitive speech to external services. Large language models may retain or expose uploaded data depending on their configuration. Researchers should not upload identifiable transcripts to an unapproved service. AI-generated coding or summaries must be checked for omission, hallucination, stereotyping, and loss of context. The researcher remains responsible for the analysis and should disclose substantial use of automated tools.

Research involving employees, patients, students, or vulnerable communities requires careful attention to power. Participation should not affect employment, grades, treatment, or access to services. Findings should avoid stigmatizing groups, and community or participant perspectives should not be reduced to convenient quotations.

8. Challenges and Limitations

Qualitative studies may be criticized for small samples, researcher subjectivity, limited generalizability, or time-intensive analysis. These concerns should be addressed through appropriate claims rather than by imitating quantitative designs. A qualitative study generally supports contextual and conceptual transfer, not automatic statistical generalization to a population.

Access can shape the evidence: organizations may permit observation only of successful deployments, while dissatisfied users may be less willing to participate. Interview accounts are also situated and may contain memory errors, strategic presentations, or socially desirable responses. Focus groups may suppress minority views. Documents may reflect institutional interests rather than actual practice.

Mixed-methods studies introduce additional challenges, including unequal status between methods, incompatible assumptions, excessive workload, weak integration, and contradictory findings. A larger quantity of data does not guarantee a better study. Researchers should define the contribution of each method, plan integration early, and report limitations honestly.

9. Best Practices for Student Researchers

1. Begin with a question that requires the type of evidence you plan to collect.

2. Justify the paradigm, design, sampling strategy, and role of each method.

3. Pilot interview and observation protocols.

4. Use open, neutral questions and probe for concrete examples.

5. Record contextual information and distinguish description from interpretation.

6. Create a versioned codebook and retain an audit trail of analytic decisions.

7. Search actively for negative cases and rival explanations.

8. Use reflexive notes throughout the project, not only in the final chapter.

9. Protect identities in quotations, transcripts, recordings, and digital data.

10. Integrate methods explicitly through connecting, building, merging, or embedding.

11. Report both convergence and discordance.

12. Make claims proportionate to the sample, context, and evidence.

Contemporary Developments

Contemporary qualitative research increasingly examines human interaction with algorithmic systems, platform-mediated work, online communities, and sociotechnical infrastructures. Researchers may combine interviews with interaction logs, participatory design workshops, usability studies, and algorithmic impact assessments. Participatory and community-based approaches involve affected groups in defining questions, interpreting evidence, and deciding how findings are used.

There is also growing interest in responsible and explainable AI. Qualitative methods can evaluate whether explanations are understandable, actionable, and trusted in context rather than merely technically available. In mixed-methods evaluation, fairness metrics may identify group differences while interviews and observations investigate how institutional processes create or respond to those differences. Such work reinforces a central principle: AI systems are used within human organizations, and their effects cannot be understood from model metrics alone.

Chapter Summary

Qualitative research provides systematic ways to study meaning, experience, interaction, process, and context. Interviews, focus groups, observation, and document analysis generate different forms of evidence and should be selected according to the research question. Coding and thematic analysis transform organized data into interpretive findings through an iterative, transparent process.

Credibility, dependability, confirmability, and transferability provide useful quality criteria. Triangulation compares perspectives and methods, but disagreement can be as informative as convergence. Reflexivity requires researchers to examine how their position and decisions influence knowledge production.

Mixed-methods research intentionally integrates qualitative and quantitative evidence. Convergent, explanatory sequential, exploratory sequential, embedded, and multiphase designs serve different purposes. Integration may connect, build, merge, or embed methods, and should be visible in the analysis rather than occurring only as separate reports. In AI and ML research, these approaches connect technical performance with usability, trust, accountability, fairness, workflow, and social consequences. Ethical practice requires informed consent, privacy protection, responsible technology use, attention to power, and claims that remain proportionate to the evidence.

Key Terms

Qualitative research: Systematic study of meanings, experiences, practices, and contexts using non-numerical or primarily interpretive data.

Mixed-methods research: Intentional integration of qualitative and quantitative approaches in one study or research program.

Interpretivism: An approach emphasizing how people construct and understand social reality.

Purposive sampling: Selecting information-rich participants or cases relevant to the research purpose.

Theoretical sampling: Selecting further data sources to develop or refine emerging concepts.

Semi-structured interview: An interview guided by prepared questions while allowing probing and flexibility.

Focus group: A moderated group discussion used to study views and interaction among participants.

Observation: Systematic study of behavior and practices in their natural or designed setting.

Document analysis: Examination of written, visual, or digital materials as situated sources of evidence.

Code: A label assigned to a relevant segment of qualitative data.

Codebook: A documented set of codes, definitions, boundaries, and examples.

Theme: A patterned and interpretive meaning relevant to a research question.

Thematic analysis: A method for identifying, developing, and interpreting themes across data.

Reflexivity: Critical examination of how the researcher influences the research process and interpretation.

Credibility: The plausibility and evidential support of qualitative findings.

Transferability: The extent to which readers can judge relevance to other contexts.

Triangulation: Comparison of multiple methods, sources, investigators, or theories.

Joint display: A table or visual device integrating qualitative and quantitative findings.

Member checking: Inviting participants to comment on data, summaries, or interpretations.

Audit trail: A documented record of research procedures, decisions, and analytic development.

Review Questions

Short Answer Questions

1. What distinguishes qualitative research from quantitative research?

2. Why are semi-structured interviews commonly used in qualitative studies?

3. Define a qualitative research code and explain its purpose.

4. What is reflexivity, and why is it important?

5. Name two criteria used to assess qualitative research quality.

Long Answer Questions

1. Explain how interviews, focus groups, observation, and document analysis produce different types of evidence. Give an appropriate use for each method.

2. Describe the major stages of thematic analysis from familiarization to reporting.

3. Compare convergent, explanatory sequential, and exploratory sequential mixed-methods designs.

4. Discuss triangulation and explain why contradictory findings should not automatically be treated as a research failure.

5. Explain the ethical challenges of conducting qualitative research on AI systems using sensitive organizational or personal data.

Analytical / Application Questions

1. A survey shows that users with more experience of an AI decision-support tool trust it less. Design an explanatory sequential study to investigate this result, including sampling, qualitative methods, integration, and ethical safeguards.

2. During interviews, employees say an algorithm is only advisory, but observations show that managers routinely follow its recommendations. How would you analyze this contradiction and report it credibly?

3. Develop an initial coding framework for studying students’ experiences of an AI tutoring system. Include at least six possible codes, two broader categories, and one possible interpretive theme.

Multiple Choice Questions with Answers

Question 1

Which feature is most characteristic of qualitative research?

A. Estimating a population mean from a random sample

B. Understanding meanings and practices in context

C. Maximizing predictive accuracy on a test set

D. Testing a null hypothesis with a significance threshold

Correct Answer: B. Understanding meanings and practices in context

Question 2

What is the main advantage of a semi-structured interview?

A. It eliminates researcher interpretation

B. It guarantees identical answers from all participants

C. It combines comparable questions with flexible probing

D. It requires no preparation or interview guide

Correct Answer: C. It combines comparable questions with flexible probing

Question 3

Which sampling strategy selects participants because they meet specified relevant conditions?

A. Criterion sampling

B. Convenience sampling

C. Cluster sampling

D. Systematic random sampling

Correct Answer: A. Criterion sampling

Question 4

Which statement best describes a qualitative theme?

A. Any frequently used word in a transcript

B. A patterned and interpretive meaning relevant to the research question

C. A participant’s demographic characteristic

D. A statistical correlation between two variables

Correct Answer: B. A patterned and interpretive meaning relevant to the research question

Question 5

What is the purpose of a reflexive journal?

A. To replace participant consent forms

B. To calculate interquartile ranges

C. To document how the researcher’s position and decisions shape the study

D. To guarantee that participants agree with the findings

Correct Answer: C. To document how the researcher’s position and decisions shape the study

Question 6

In an explanatory sequential mixed-methods design, which sequence is normally used?

A. Qualitative data followed by quantitative data

B. Quantitative data followed by qualitative investigation

C. Only qualitative data collected in several phases

D. Qualitative and quantitative data permanently kept unrelated

Correct Answer: B. Quantitative data followed by qualitative investigation

Question 7

Which is an example of triangulation?

A. Using only one interview quotation repeatedly

B. Comparing interviews, observations, and organizational documents

C. Replacing field notes with an automated summary

D. Increasing the number of survey response categories

Correct Answer: B. Comparing interviews, observations, and organizational documents

Question 8

Which quality criterion concerns whether readers can judge relevance to another context?

A. Transferability

B. Randomization

C. Statistical power

D. Internal consistency

Correct Answer: A. Transferability

Question 9

What should a researcher do when qualitative and quantitative findings conflict?

A. Delete the less convenient dataset

B. Assume the quantitative result is always correct

C. Investigate the possible reasons for the discordance

D. Report only the finding supported by more participants

Correct Answer: C. Investigate the possible reasons for the discordance

Question 10

Which practice is most appropriate when using an AI tool to transcribe sensitive interviews?

A. Upload identifiable recordings to any free service

B. Use the output without checking it against the recording

C. Verify the transcript and follow approved data-protection procedures

D. Allow the tool to determine the final themes automatically

Correct Answer: C. Verify the transcript and follow approved data-protection procedures