Redistributing Epistemic Labour: The Impact of Generative AI on Teachers' Instructional Decision-Making in Vietnamese High Schools

Redistributing Epistemic Labour: The Impact of Generative AI on Teachers’ Instructional Decision-Making in Vietnamese High Schools

Dataset Overview

This repository contains the dataset associated with a longitudinal study investigating how Vietnamese high school teachers interact with Generative Artificial Intelligence (AI) during instructional planning and decision-making activities. The study focuses on understanding how AI-related capability, perceptions, trust, and cognitive engagement develop over time and contribute to instructional decision quality.

The dataset was collected as part of a broader research project examining human–AI collaboration in educational settings and the redistribution of cognitive and epistemic work between teachers and AI systems.

Research Objectives

The dataset was developed to examine:

• The influence of teacher AI self-efficacy on perceptions of artificial agency and collaborative trust.
• The role of perceived artificial agency in shaping trust toward AI-supported instructional activities.
• The contribution of AI collaborative trust to epistemic labour redistribution.
• The relationship between epistemic labour redistribution and instructional decision quality.
• The developmental dynamics of teacher–AI collaboration across multiple stages of interaction.

Study Design

A three-wave longitudinal survey design was employed.

Data collection was conducted across one academic year to capture the temporal development of teacher–AI interaction.

Measurement waves included:

• Time 1 (T1):
• Teacher AI Self-Efficacy (TAI)
• Perceived Artificial Agency (PAA)
• Time 2 (T2):
• AI Collaborative Trust (ACT)
• Epistemic Labour Redistribution (ELR)
• Time 3 (T3):
• Instructional Decision Quality (IDQ)

The temporal separation between measurement waves was intended to reduce common method bias and strengthen interpretation of causal ordering among constructs.

## Data Collection Period

Data were collected between September 2024 and March 2025 using a three-wave longitudinal survey design.

The data collection process was organized across three measurement occasions:

- Time 1 (September–October 2024): Teacher AI Self-Efficacy (TAI) and Perceived Artificial Agency (PAA).
- Time 2 (November–December 2024): AI Collaborative Trust (ACT) and Epistemic Labour Redistribution (ELR).
- Time 3 (February–March 2025): Instructional Decision Quality (IDQ).

Temporal separation between measurement waves was implemented to reduce common method bias and to capture the developmental dynamics of teacher–AI interaction over time.

Participants

The final matched dataset contains responses from 576 Vietnamese high school teachers.

Participant characteristics include:

• Teachers from public high schools located in urban, semi-urban, and rural areas.
• Participants representing diverse teaching disciplines, including mathematics, literature, English, natural sciences, social sciences, information technology, arts, and physical education.
• Teachers with varying levels of professional experience, ranging from early-career educators to highly experienced teachers with more than twenty years of service.
• All participants reported prior experience using at least one Generative AI platform for instructional, professional, or administrative purposes.

Generative AI Usage Context

Participants reported active use of Generative AI across multiple professional activities.

Commonly used AI platforms include:

• ChatGPT
• Gemini
• Claude
• Microsoft Copilot
• Perplexity AI
• DeepSeek
• Grok
• Poe
• Qwen
• Kimi

Reported purposes of AI use include:

• Lesson planning and instructional design.
• Learning material and content preparation.
• Assessment and rubric development.
• Classroom activity design and differentiation.
• Student feedback and learning support.
• Information search and pedagogical idea generation.
• Administrative reporting and documentation.
• Professional communication and document drafting.

Dataset Contents

The dataset contains the following categories of variables:

Demographic Variables

• Gender
• Teaching experience
• Subject area
• School location

Generative AI Usage Variables

• Primary AI platform used
• Frequency of AI use across teaching-related activities
• Patterns of AI-supported instructional practice

Research Variables

The dataset includes responses to five latent constructs:

• Teacher AI Self-Efficacy (TAI)
• Confidence in evaluating, adapting, and using AI-generated outputs.
• Perceived Artificial Agency (PAA)
• Perceptions of AI autonomy, responsiveness, adaptability, and initiative.
• AI Collaborative Trust (ACT)
• Willingness to rely on AI-generated recommendations during instructional work.
• Epistemic Labour Redistribution (ELR)
• Redistribution of selected cognitive and reasoning tasks between teachers and AI systems.
• Instructional Decision Quality (IDQ)
• Perceived effectiveness, coherence, and pedagogical value of AI-supported instructional decisions.

All constructs were measured using:

• Four survey items per construct.
• Five-point Likert response scales.
• Response options ranging from 1 (Strongly Disagree) to 5 (Strongly Agree).

Data Collection Procedures

Data were collected between September 2024 and March 2025.

Key procedures included:

• Online questionnaire administration.
• Distribution through participating schools and professional teacher networks.
• Anonymous participation.
• Longitudinal response matching using self-generated identification codes.
• Removal of duplicate records, unmatched responses, and cases with substantial missing data.

The final dataset includes only participants who completed all three measurement waves successfully.

Potential Uses of the Dataset

This dataset may be useful for:

• Artificial Intelligence in Education (AIED) research.
• Human–AI collaboration studies.
• Teacher professional learning research.
• Educational technology adoption research.
• Trust in AI research.
• Distributed cognition and cognitive offloading studies.
• Longitudinal structural equation modelling.
• Cross-cultural and comparative educational research.

Researchers may use the dataset for replication studies, secondary analyses, model comparison, methodological investigations, and theory development related to AI-supported educational decision-making.

Ethical Approval

The study received ethical approval from Vietnam National University, Hanoi under Decision No. 3073/QĐ-ĐHQGHN dated 28 June 2024.

Participation was voluntary and based on informed consent.

No personally identifiable information is included in the publicly shared dataset.

Funding

This research was supported by Vietnam National University, Hanoi under the Science and Technology Project QG.24.87.