Annotated Bibliography - Artificial intelligence in education: The three paradigms. (Ouyang and Jiao, 2021)

Ouyang, Fan, and Pengcheng Jiao. "Artificial intelligence in education: The three paradigms." Computers and Education: Artificial Intelligence 2 (2021): 100020.

In the paper "Artificial intelligence in education: The three paradigms.", by Fan and Jiao, the authors analyze the field of artificial intelligence use in education, which is commonly referred to as “AIEd” and manifest in applications such as intelligent tutoring systems, teaching botos, learning analytics dashboards, adaptive learning systems, human-computer interactions, among others. 

However, the author claims that the use of AI alone does not incur good educational outcomes. The use of technology must be aligned with the correct pedagogical perspectives. And that is the gap the authors aim to fill with the paper: how the different roles of AIEd are connected to existing educational and learning theories. They do that by proposing three AIEd paradigms: 1) AI-Directed, learner-as-recipient. 2) AI-supported, learner-as-collaborator, 3) AI-empowered, learner as leader. 

The methodology of the paper was a standard literature review, and the research question was: “what are the different roles of AI in education, how AI are connected to the existing educational and learning theories, and to what extent the use of AI technologies influence learning and instruction.”

Paradigm one: AI-Directed, learner-as-recipient

The authors identified that, in the first paradigm, the AI represents the domain knowledge and directs the learning process. On the other hand, the learner is the recipient of such knoweldge and follows a specific learning pathway.

This is based on the theory of behaviorism, that sees learning as a reinforcement of knowledge acquisition through programmed instructions.  It offers new concepts in a logical and incremental way and offers the users immediate feedback about responses and tries to maximize positive reinforcements. Then, the learner is the recipient of pre-defined sequences of knowledge. The A.I. system does not model itself to the learner's incoming knowledge and skills, nor adjust its feedback to the learner. It is the paradigm that is least learner centered. An example of implementation of AIEd in paradigm one is ITSs (Intelligent Tutoring Systems). 

One issue with AIEd in paradigm one is that the students´ characteristics, goals and needs are not taken into consideration. The system might collect information about the learner's knowledge and skills, but it is only used to put the student in the “right” step of a predefined knowledge path. 

Paradigm two: AI-Supported, learner-as-collaborator

The second paradigm classified by the authors sees the A.I. as a supporting tool, while the learner is the collaborator and the system has a focus on the individual's unique learning process. The theoretical theory which supports this paradigm is constructivism, that defends that learning occurs as individuals interact with people and that information and technology are socially situated constructs. Therefore, for this paradigm, the A.I. system and the learner should interact and work as a team to optimize the learner-centered, personalized learning experience. 

Some A.I. implementations that use paradigm two are the dialogue-based tutoring systems (DTSs) or the exploratory learning environments (ELEs). 

One issue with implementations in such a paradigm is the lack of continuous communication between the learners and the A.I. system. The interaction is very complex because neither the individual nor the system´s state is static. 

Paradigm Three: AI-Empowered, learner-as-leaders

The third paradigm, as described by the authors, view education as a complex adaptive system that requires a synergy between many entities (student, instructor, information and technology). The authors claim that this paradigm recognizes that the instructor has to be equipped with support to foster learner-centered learning. Therefore, it is up to the learner to take the agency of her own learning and manage the risks of the AI decision. Then, the ultimate goal of the application is to augment human intelligence and potential, but the human remains the leader of this process. Then, paradigm three defends the synergetic collaboration between the A.I. and human intelligence to produce adaptive and personalized learning. 

The paper continues to analyze that the most recent advances in human-computer interaction allowed for real time monitoring of the cognitive, emotional and social aspects of the learning process. The more recent algorithms can also interpret the real time collection and analysis of the students´ inputs to provide feedback. The authors conclude the paper claiming that the future of AIEd must be learner-centered, data-driven, personalized and up to date with the current knowledge. 

This was a very interesting paper that shows how the use of A.I. must be aligned with an educational framework and planned pedagogical outcome. It is an important reflection about how computer science and educational science must walk together to produce the best outcomes.