The term artificial intelligence (AI) is well known but often misunderstood. To the average person, AI may conjure science-fiction images of an allknowing entity that has superior cognitive abilities in comparison to humans. This imagery plays on the idea of an artificial general intelligence, technology that is still years away (Haenlein & Kaplan, 2019).
Human-Technology Collaboration PHD Student, George Washington University, Graduate School of Education & Human Development, 2134 G ST, NW, Washington, DC 20052.
Human-Technology Collaboration PHD Student, George Washington University, Graduate School of Education & Human Development, 2134 G ST, NW, Washington, DC 20052.
Today, AI falls into what experts call narrow AI; AI designed for a specific purpose or task (Haenlein & Kaplan, 2019). While AI experts agree on the general versus narrow AI macrolevel categorization, discussion still exists on the best way to categorize and describe both what AI currently is and is not at the microlevel (West, 2018). This complicates the dissemination of AI’s definition to other fields of nonexperts. Within education, researchers have frequently failed to define AI in their work. In a recent literature review of higher education research on artificial intelligence, Zawacki-Richter et al. (2019) found that only 5 out of 146 (3.4%) articles clearly defined the concept of AI in their studies. The lack of emphasis on defining and communicating AI’s conceptual boundaries contributes to misunderstanding and intimidation around the term for many, including those in education. In this article we introduce various definitions of AI, discuss emerging conceptual frameworks for AI in education, and explain why definitions of AI matter for online educators.
Professor of Education Technology, George Washington University, 2134 G ST, NW, Washington, DC 20052. Telephone: (202) 994-1884.
Professor of Education Technology, George Washington University, 2134 G ST, NW, Washington, DC 20052. Telephone: (202) 994-1884.
What is Artificial Intelligence?
Although Alan Turing is often credited with the general concept of AI, John McCarthy was the first to coin and define the term in the mid-1950s (McCarthy, 2007; West, 2018). McCarthyʼs (2007) definition of AI is “the science and engineering of making intelligent machines, especially intelligent computer programs” (p. 2). Within the education literature, Baker and Smith (2019) defined AI as a computer that performs cognitive tasks. But what does “intelligent” and “cognitive” mean in practice and why define a concept so broadly? Baker and Smith (2019) noted that they purposefully defined AI broadly because it “does not describe a single technology,” but instead is made up of a variety of techniques, most recognizably machine learning (p. 10).
AI differs from traditional computer programming in the fact that the computer system does not simply stick to rules written by a developer. The computer is programmed in a way that allows it to learn from its past experiences, using statistical error, with a large dataset and to make judgments and/or predictions. Some examples of the techniques used by AI to learn are natural language processing (NLP), computer vision, and machine learning, among others. These techniques each have overlap, with many specialists still debating the boundaries of the conceptual relationships. However, the debate around the conceptual boundaries of techniques is beyond the scope of this article.
Natural language processing is the “conversion of human language into structured data’’ (Sorin et al., 2020, p. 640). In online education, natural language processing will be at the foundation of emerging technologies that assist instructors with essay grading and that make virtual tutors/ assistants like chatbots possible (Page & Gehlbach, 2017; Saleheen et al., 2018; Smith et al., 2020). Computer vision is a larger area that focuses on computers “extracting information” from videos and pictures (Meske & Bunde, 2020, p. 54). In online education, computer vision may contribute to emerging technologies that seek to identify boredom, frustration, or burnout in online students during learning (Ashwin & Guddeti, 2020; Behera et al., 2020). Machine learning is the “science of programming computers so they can learn from data” (Géron, 2019, p. 4). Machine learning is often categorized as shallow or deep learning. Shallow learning uses more traditional statistical techniques, and deep learning builds a complex network of neural layers in which information passes back and forth. Machine learning techniques, especially deep learning neural networks, are used in both natural language processing and computer vision. In online education, machine learning algorithms are likely to be at the center of adaptive learning systems that use baseline data for each student to adjust the learning content/ review sequence based on the machine’s prediction for what each student needs (du Boulay, 2019; Smith, 2018).
How is Education Conceptualizing Al?
Researchers in the field of education have slowly been conceptualizing the role and vernacular of artificial intelligence as applicable to its context. Two examples of conceptualizing AI in education come from the groups Artificial Intelligence (AI) for K– 12 initiative (AI4K12) (https://ai4k12.org/) and Nesta (https://www.nesta.org.uk/project/artificial-intelligence/). AI4K12, a group sponsored by the Association for Advancement of Artificial Intelligence and the Computer Science Teachers Association, aims to create national guidelines for AI in K–12 education. The group’s “5 Big Ideas in AI” are also helpful for fleshing out what AI means. They state that AI (1) perceives the world using sensors, (2) creates representations of the world that it uses in reasoning, (3) can learn from data, (4) interacts naturally with humans, and (5) has both positive and negative societal impacts (AI4K12, 2020). See Ai4K12 (2020) for a graphical illustration of these 5 Big Ideas in AI, as well as in-depth descriptions of these ideas.
Baker and Smith (2019), part of the U.K. innovation foundation Nesta, organized educational AI into three categories: learner-facing AI, teacher-facing AI, and system-facing AI. Learner-facing AI focuses on the needs of students. Examples are learning platforms that are “adaptive,” “differentiated,” or “individualized” based on individual student needs (Baker & Smith, 2019, p. 11). For instance, teacherfacing AI involves easing teachers’ administrative tasks, like use of automated essay grading and/or technologies that assist teachers in effective grouping of students and tracking of their individual academic progress. On the other hand, system-facing AI “make[s] or inform[s] decisions made by those managing and administrating schools or our education system as a whole” (Baker & Smith, 2019, p. 14). Currently, system-facing AI is the least developed of the three although it has potential for data analytics not only within individual educational contexts but also the possibility for connecting data between these contexts as well.
Why do Al Definitions Matter for Online Educators?
As demonstrated above, the concept of AI is complex and interdisciplinary, and the technologies associated with it are frequently evolving. Although AI is currently available to many online educators in some capacity, “Experts see AI as accelerating rapidly now, and more intense and widespread impacts will soon become prevalent” (Roschelle et al., 2020, p. 3). Consequently, although educators are not expected to be experts on the topic of AI, fostering a basic understanding is a key first step in determining how it might be used meaningfully and with purpose. In addition, it is critical for online educators to not only be aware of what AI can mean and how it might be used for teaching and learning, but it is also equally—if not more important—to use this awareness to comprehend potential ethical, data privacy, bias, and other problematic issues associated with AI. Moreover, it is essential to be clear about one’s conceptualizations and definitions of AI, particularly when conducting and sharing research and communicating with students and colleagues about AI.
Clearly defining AI may help improve communication around the topic in education and, potentially, also demystify it by cultivating a shared understanding. Additionally, the more that educational stakeholders understand AI’s boundaries and basic techniques, the more they will be able to pass that knowledge on to students. The students of today will be the AI users of tomorrow, and it is paramount that we have a society that thinks critically about both the benefits and drawbacks of AI. The more that individuals are able to grapple with what AI can realistically do, the more that those individuals can question not only the outcomes of AI but the fairness of the data and training methods used to train the system.
