1.1 How Do Artificial Intelligence and Machine Learning Appear in Interactive Learning Experiences?
Technology is transforming how we solve complex problems, as well as how we share information. In this chapter, we look at an innovative learning environment from the perspectives of an enrolled student, a teaching assistant, and the professor of a fluid dynamics course with 100+ enrolled students. The scenario and research provide insight into the value of incorporating artificial intelligence and machine learning into the learning experience.
-
Student
-
Halfway through the pursuit of their undergraduate degree in chemical engineering, Alex Rhimes, age 20, from Baltimore, Mary’s Lake, was planning on taking the foundational fluid dynamics class—the most notoriously difficult class in the major. The learning management system and recommendation engine used by the university suggested taking this course early based on Alex’s good grades and internship experience. Like most students, Alex logged into ratemyprofessors.com before selecting the class. With ratings in the high 4s, Alex tabbed over to the university course site and clicked the big blue “Register” button on the screen. According to the reviews, Prof. Gomez went above and beyond to create a highly personalized environment for each student. As soon as Alex registers, an email notification is received: Pre-Course Simulation Game.
-
Instructor
-
In Charrysville, Virgonne, Dr. Riley Gomez, second year associate professor, wakes up early on the first day of class to check new emails. Rolling over in bed groggily, Prof. Gomez reaches for the phone on the bedside table, scrolling past the unfiltered junk emails. Prof. Gomez is anticipating notifications from students submitting their last-minute survey responses to the self-assessment exercise shared a week ago. As student enrollments have increased and class sizes increased, Prof. Gomez started incorporating artificial intelligence and machine learning methods into the classroom environment. It was the only feasible way to reach the 100+ students enrolled in the fluid dynamics class.
-
“Lecture-style classes with the sage on the stage are a thing of the past,” explains Prof. Gomez. “Using algorithms is the most effective way to manage classes in the face of increased enrollment.”
-
Prof. Gomez swipes left, left, and down in order to load the most recent results. It looks like the class is spread all over the place with experience levels and interest in fluid dynamics. It is not uncommon for students to drop out of this class and fail to persist. However, Prof. Gomez is adamant about ensuring that every student feel supported and hopeful that the simulation assessment provided an opportunity for students to learn some basics before the first session. The left chart on the dashboard shows passive traits that were monitored during the simulation. Indicators such as eye movements and facial expressions are tracked in blue and orange. The middle chart on the dashboard illustrates anticipated knowledge gaps and opportunities for support based on data mined from last year’s class.
The role of artificial intelligence (AI
) in US education is continuing to expand (
Artificial Intelligence Market in the US Education Sector 2018-2022—Key Vendors are Cogni, IBM, Microsoft, Nuance Communications, Pixatel, and Quantum Adaptive Learning—ResearchAndMarkets.com 2018). As education moves toward providing customized learning paths, the use of artificial intelligence (AI) in learning systems increases, creating scaffolding that extends the ability and reach of an instructor (Tsinakos
2006) much as a physical exoskeleton combined with augmented reality enables a worker to see more than what is in front of them, and accomplish tasks they are not able to complete on their own (Srinivasan
2018). Chatbots (Bradeško and Mladenić
2012; Fonte et al.
2016; Albayrak et al.
2018; Eicher et al.
2018), autograders (Wang et al.
2018; Kyrilov
2014), and systems that passively monitor and then direct student progress (Paaßen et al.
2018) use AI, machine learning
(ML), and deep learning technologies to store and process data and then communicate it to students and instructors. This exploitation of AI in education requires substantial funding and time for research, implementation, and assessment for the education community to understand the efficacy of the technology and its role in student persistence and subsequent on-the-job performance or success in graduate studies (Marr
2018; Polachowska
2019).
Artificial intelligence is the study of how to make computers perform tasks that appear to require intelligence when performed by humans. Machine learning and deep learning fall under this broad definition of artificial intelligence. Machine learning
focuses on parsing and analyzing data in an automated fashion, without human intervention, to learn models for decision-making. Machine learning is considered a data mining technique. An algorithm that clusters data according to its similarities and differences is an example of machine learning. Deep learning is a subset of machine learning that relies on networks that mimic the way the human brain processes data and creates patterns to acquire decision-making ability.
A chatbot is a software program that converses with a human user. Chatbot ability ranges from those that conduct a shallow dialogue over a broad range of topics to those with deep knowledge and conversational ability over a well-scoped domain of discourse. The best are difficult to distinguish from human conversants. Autograders are software programs used to evaluate work produced by students with little or no human intervention. They can perform tasks ranging from scoring multiple choice tests to analyzing and grading essays. The findings produced by autograders range from binary (correct/incorrect) to conceptual feedback. Automatic review of essays is often combined with human review of the essays, with subsequent closer human examination of an essay if the automated and initial human results disagree. Passive monitoring and guidance can be integrated into an online learning system to compare a student’s activities to expected behavior. An instructor might learn through the system that several students engaged in online activities appear to be making similar errors, or when a particular individual appears to be lagging behind. They can also suggest interventions to the instructor, tailored to the difficulty encountered. Likewise, these systems can provide students hints about what they might try or modules to review, as well as feedback regarding how the student is progressing relative to the rest of the class.
Applications of AI-based education technology support learning in four ways: through automated tutoring, personalizing learning, assessing student knowledge, and automating tasks normally performed by the instructor (Lu and Harris
2018). Intelligent tutoring systems (ITS)
produce statistically significant improvements in student learning outcomes, such as mastery and retention, when compared to traditional classroom teaching, independent textbook use, and non-AI computer-based instruction (Ma et al.
2014). However, experts point out that ITS curricula are rather inflexible due to technical challenges in accommodating user feedback, modified core standards, or content changes.
In addition to supporting improved student learning outcomes, the use of AI and ML in education has the potential to lead to improved teacher satisfaction (VanLehn et al.
2019a,
b; Dietrich
2015). AI coupled with ML can provide 24/7 student support. It supports tracking student performance and aggregating student concerns. It can facilitate personalizing and adapting learning materials to individual students. These automated tools enable timely and passive assessment and more finely grained tracking of student knowledge and skill gains (Aleven et al.
2010; Arroyo et al.
2014). This assistance empowers the instructor, who can feel more confident in student opportunity to succeed, knowing that the students are receiving needed support that the instructor might otherwise struggle to provide. The instructor is able to devote time to creative activities and feedback beyond what the automated systems can provide, such as affective feedback and support (Wu et al.
2016; Duo and Song
2012).
The implementation of an online learning system
requires a sophisticated digital ecosystem that incorporates the complex interactions among students, instructors, and content. It must include a sophisticated human-computer interface that supports access, monitoring, feedback, and assessment (Reyna
2011; Rezaei and Montazer
2016). The system can be built upon an existing e-business solution or learning management system
, or arise from an array of independent modules. These systems are often cloud-based, providing services over the Internet, to provide maximum accessibility.