Role of AI in Blended Learning: A Systematic Literature Review

📝 Paper Summary

AI in Education (AIEd) Blended Learning

A systematic review of 30 studies reveals that AI in blended learning is predominantly used for online asynchronous tasks as a direct mediator, with limited application in connecting online and offline activities.

Core Problem

Blended learning faces challenges in balancing flexibility with structure, stimulating interaction, and supporting self-regulation, which traditional methods often struggle to address effectively.

Why it matters:

Blended learning is becoming a primary instructional approach post-COVID-19, requiring scalable solutions for personalization and efficiency.
Without technological support, high flexibility can lead to low engagement, isolation, and poor self-regulation among students.
Prior reviews often focus generally on AIEd or specific technologies (like ITS) rather than the specific intersection of AI's role within the unique constraints of blended learning.

Concrete Example: In a typical flipped classroom, students might watch videos online but struggle to self-regulate or connect concepts to the in-person class. Teachers lack visibility into this gap. The review finds AI often addresses the online part (e.g., adaptive quizzes) but rarely acts as a bridge connecting the online data to the face-to-face classroom activities.

Key Novelty

Taxonomy of AI Roles in Blended Learning Challenges

Categorizes AI applications into three roles: 'New Subject' (e.g., chatbots), 'Direct Mediator' (e.g., adaptive platforms), and 'Supplementary Assistant' (e.g., learning analytics).
Maps these roles directly to four persistent challenges of blended learning: flexibility/autonomy, interaction, learning processes, and affective climate.
Identifies a critical gap: AI is mostly used for 'online' components rather than integrating or enhancing the 'face-to-face' component.

Evaluation Highlights

76.7% of reviewed studies (23/30) applied AI exclusively to the online asynchronous component of blended learning.
40.0% of studies identified AI's primary role as a 'Direct Mediator' (e.g., intelligent tutoring systems managing instruction directly).
Only 23.3% of studies utilized AI to connect online activities with classroom-based offline activities.

Breakthrough Assessment

4/10

A solid systematic review that provides a useful categorization framework (Roles vs. Challenges). However, it is a synthesis of existing work rather than a new technical method, and the sample size (30 papers) is relatively small.

⚙️ Technical Details

Problem Definition

Setting: Systematic Literature Review of empirical studies published between Jan 2007 and Oct 2023

Inputs: Search queries combining 'blended learning', 'hybrid learning', 'flipped learning' with 'artificial intelligence' or 'AIEd' on Google Scholar, ERIC, and Web of Science

Outputs: Synthesis of 30 eligible empirical studies analyzed via two frameworks: Boelens et al.'s challenges and Xu & Ouyang's AI roles

Pipeline Flow

Database Search (Google Scholar, ERIC, WoS)
Manual Search (Specific EdTech Journals)
Screening (Titles/Abstracts)
Eligibility Check (Full-text review against criteria)
Data Extraction & Coding (Types of BL, AI Roles, Challenges)
Synthesis & Analysis

System Modules

Search Strategy

Identify potential studies

Model or implementation: Boolean queries on databases + manual journal review

Screening & Selection

Filter studies based on inclusion criteria

Model or implementation: PRISMA framework

Coding Framework

Categorize studies

Model or implementation: Dual Framework: Boelens et al. (2017) Challenges + Xu & Ouyang (2021) AI Roles

Novel Architectural Elements

Integration of two specific conceptual frameworks (Blended Learning Challenges + AI Roles) to map the landscape of the field

Comparison to Prior Work

vs. Chen et al.: Focuses specifically on the 'blended learning' context rather than general education
vs. Bergdahl et al.: Expands scope beyond just Learning Analytics to include agents, tutors, and other AI roles
vs. General AIEd reviews: Specifically maps AI functions to the four pedagogical challenges of blended learning (flexibility, interaction, process, affect)

Limitations

Small sample size of only 30 studies
Limited coverage of Generative AI (ChatGPT) due to the cutoff date of the search (October 2023) relative to the technology's emergence
Did not include conference proceedings, which often contain the latest AIEd prototypes

Reproducibility

The paper documents search terms, databases, inclusion/exclusion criteria, and the PRISMA flow (Figure 2). The list of 30 analyzed studies is not explicitly provided as a separate appendix in the text but referenced throughout the results tables.

📊 Experiments & Results

Evaluation Setup

Systematic review analysis of 30 empirical papers

Benchmarks:

N/A - Qualitative Synthesis (Literature Review)

Metrics:

Frequency of AI roles (Direct Mediator, Supplementary Assistant, New Subject)
Frequency of AI application context (Online vs. Offline)
Distribution of research contexts (Higher Ed vs. K-12)
Statistical methodology: Descriptive statistics (counts and percentages) of coded categories

Key Results

Benchmark	Metric	Baseline	This Paper	Δ
Reviewed Papers (N=30)	Percentage applied to Online Component only	Not reported in the paper	76.7	Not reported in the paper
Reviewed Papers (N=30)	Percentage applied to Both Online and Offline	Not reported in the paper	23.3	Not reported in the paper
Reviewed Papers (N=30)	Percentage utilizing AI as Direct Mediator	Not reported in the paper	40.0	Not reported in the paper
Reviewed Papers (N=30)	Percentage utilizing AI as Supplementary Assistant	Not reported in the paper	20.0	Not reported in the paper
Reviewed Papers (N=30)	Percentage utilizing AI as New Subject (e.g., Agent)	Not reported in the paper	16.7	Not reported in the paper
Reviewed Papers (N=30)	Percentage in Higher Education	Not reported in the paper	66.7	Not reported in the paper

Main Takeaways

AI is primarily used to automate the online, asynchronous part of blended learning (e.g., adaptive quizzes, tutors), reinforcing the 'distance' aspect rather than integrating with the 'face-to-face' aspect.
There is a lack of research on using AI to bridge the gap between online self-study and offline collaboration (e.g., using online data to group students for in-class activities).
AI acts as a 'Direct Mediator' most often, helping to control student flexibility and provide immediate feedback, which addresses the challenge of self-regulation in blended environments.
Future research opportunities lie in 'AI as a New Subject' (chatbots/robots) and systems that actively support teacher orchestration across both learning modalities.

📚 Prerequisite Knowledge

Prerequisites

Understanding of blended learning models (flipped, rotation, etc.)
Basic knowledge of AI applications in education (ITS, chatbots, learning analytics)

Key Terms

Blended Learning: An instructional approach combining face-to-face classroom methods with computer-mediated activities

Flipped Learning: A type of blended learning where direct instruction moves to the individual learning space (usually online video), and the group space is transformed into a dynamic, interactive learning environment

ITS: Intelligent Tutoring System—computer systems that provide immediate and customized instruction or feedback to learners without intervention from a human teacher

Pedagogical Agent: Virtual characters or bots designed to facilitate learning, often by simulating social interaction

NLP: Natural Language Processing—AI technology enabling computers to understand, interpret, and generate human language

Learning Analytics: Measurement, collection, analysis, and reporting of data about learners and their contexts to understand and optimize learning

Direct Mediator: An AI role where the technology acts as the primary bridge or platform for instruction (e.g., an adaptive learning system)

Supplementary Assistant: An AI role where the technology supports the teacher or student indirectly, often through data insights (e.g., dashboards)

New Subject: An AI role where the technology acts as a participant, such as a peer, tutee, or social robot