Learning Analytics and Intelligent Tutoring Systems

State of the Art and Practice in AI in Education

• Recent advancements in Artificial Intelligence (AI) have fueled high expectations for its future role in education and learning (AIED).
• These expectations are often based on misunderstandings of current AI capabilities, limited understanding of existing AIED systems, and overly simplistic views of education and its role in society.
• A review of AI systems in education reveals diverse pedagogical and educational philosophies underlying their design.

Introduction

• AI is frequently compared to "new oil," highlighting its significant economic potential and societal impact.
• There are substantial global investments in AI technology.
• Concerns exist regarding the potential use of AI in education, including the automation of teacher tasks and its impact on learning outcomes.
• AIED research and development has primarily been driven by computer scientists, but commercial interests have become increasingly prominent.
• The AIED market is projected to rapidly expand to over US$20 billion.

What is AI?

• Defining AI is complex and often contested; various perspectives exist in the literature and policy domains.
• The European Union (EU) focuses on AI products marketed for the EU common market, whereas discussions about AI often rely on science fiction or hypothetical scenarios of future AI capabilities.
• A working definition of Al is that it's a machine-based system capable of making predictions, recommendations, or decisions about real or virtual environments.

Data-Driven AI

• Data-driven AI involves using large datasets to train computer models to improve their accuracy in predictions.
• Key to this approach is the iterative adjustment of model parameters, a technique known as deep learning.
• Modern AI systems contain numerous interconnected "neural networks," with millions or billions of adjustable parameters.
• Training these complex models often requires significant computational resources, and large datasets.

Knowledge-Based AI

• Knowledge-based AI systems rely on representing human knowledge and expertise as rules in computer programs or expert systems.
• These systems often leverage reasoning and rule-based inference to make decisions.
• Knowledge-based AI systems have been widely used in simpler contexts but face challenges in complex or dynamic environments.

Taxonomy of AIED Systems

• AI in education (AIED) can be categorized as:
• Student-focused AIED: Includes intelligent tutoring systems (ITS), AI-assisted learning apps and simulations, support for students with disabilities, automatic essay writing, chatbots, automatic formative assessment, dialogic-based tutoring systems, and exploratory learning environments.
• Teacher-focused AIED: Focuses on aiding teachers, including plagiarism detection, smart curation of learning materials, classroom monitoring and Al-assisted feedback and summative assessment tools.
• Institution-focused AIED: addresses administrative functions such as admissions, course planning, scheduling, and dropout identification tools.

Roadblocks on the AIED Highway

• Ethical Concerns: Raises questions regarding student autonomy, data privacy, and potential biases embedded in AI systems.
• Personalization: A complex multifaceted concept that needs careful consideration to avoid homogenization of learning practices and ignore the social interaction aspect and collaboration in learning.
• Efficacy and Impact: Evidence of positive impact (especially in large scale) is often lacking.
• Techno-solutionism: The assumption that AI can solve all educational problems without considering underlying societal and cultural causes.
• AIED Colonialism: The use of AIED systems in inequitable power dynamics between Global North and Global South nations.
• Commercialization of Education: concerns about the potential for AIED systems to be monetized and prioritize profit, without adequate consideration for learner benefits and educational values.

Concluding Vision

• The future of AIED likely requires a combination of both knowledge-based and data-driven approaches.
• Future AIED systems will likely use data for pattern recognition for providing basic learning information processing.
• Developing AIED systems should consider the human development aspect.
• Further research in the areas of AIED, ethics, personalization, efficacy, techno-solutionism, and commercialization is necessary.