Post by Tom Daccord
Artificial Intelligence in education is a fast-growing field that is attracting much educator interest and commercial investment. Of late, much attention has been paid to ChatGPT, an AI assistant that can write essays, poetry, and code, roleplay, explain difficult concepts, point-out incorrect premises and more. The following Q&A aims to help EdTechTeacher readers understand Artificial Intelligence and ChatGPT in context and outline how they are impacting education.
What is Artificial Intelligence?
Artificial intelligence (AI) is the ability of a computer, machine, robot, or tech system to demonstrate characteristics of human intelligence. AI is commonly associated with platforms or systems that present human intellectual capabilities, such as the ability to reason, solve problems, use language, or learn from past experience. AI technologies are being applied today in many areas, such as robot control (self-driving cars), understanding human speech (Siri and Alexa), chatbots (ChatGPT), written language (Sudowrite), game systems (Go), customer profiling (e.g., Google and Amazon), planning and scheduling, and medical diagnosis and treatment. Advances in Machine Learning have made AI more powerful in recent years and AI systems will undoubtedly play a greater role in our personal and professional lives in the years to come.
What is Machine Learning?
Machine Learning is a subset of Artificial Intelligence and denotes how machines “learn” by embarking on a series of trial-and-error processes to determine how they can improve. Advanced AI systems use large amounts of data to identify patterns and can predict what a user might do in a given situation (an “if-then” scenario). Machine Learning also enables AI systems to perform advanced tasks without following explicit instructions. As such, Machine Learning focuses on developing computer systems that can access data and learn to use it for themselves. Machine Learning applications include robot assistants, military training simulation, and medical surgeries, among others.
What is ChatGPT?
ChatGPT is a recently released language-processing system developed by OpenAI, a San-Francisco based research-and-development company. ChatGPT is an artificial intelligence assistant trained to help answer questions and provide information on a wide variety of topics. ChatGPT functions as an advanced chatbot that can (among other things) write essays, craft poetry, produce song lyrics, write and debug code, explain difficult concepts, identify incorrect premises, roleplay in imaginary scenarios, and more. ChatGPT employs a user-friendly conversational style and can understand natural language input. And it can provide extensive responses with only a small amount of inputted text. But its content is limited to 2021, so it cannot be relied upon to address current events.
ChatGPT is a variation on GPT-3, the largest language model ever created. GPT-3 (Generative Pre-trained Transformer) absorbs and analyzes an enormous amount of text on the Internet and in various publications and requires only a small amount of text to write stories, create characters, craft poetic language, and do so “impressively” and “human-sounding.” ChatGPT is essentially leveraging GPT-3 to become the world’s most advanced chatbot.
GPT-3 produced content raises many important implications for educators, which I outline in this article. The recent release of ChatGPT has only heightened fears that students will use AI technology to cheat on their homework and other assignments.
How is Artificial Intelligence being used in education now and how might ChatGPT impact its use in the future?
AI is being used today in education in the form of personal tutors, dialogue simulators, AI-assisted subject and topic apps, course management platforms, automated writing systems, and more.
An Intelligent Tutoring System (ITS) is the most commercially prevalent example of AI in schools and universities. ITSs act as a personal tutor for students by leveraging AI to adapt the content or style of instruction to the perceived needs of the user. An ITS typically furnishes curriculum content sequentially in the form of activities, including quizzes that are automatically graded. ITSs provide just-in-time feedback and assessment via data collected from student responses. These responses are analyzed to determine the information and activities to be presented next to the user. In this way, an Intelligent Tutoring System adapts its activities to match (repeatedly) the perceived learning level of the student. The fundamental goal is to generate a personalized pathway to student mastery of curriculum content. Examples include Algebra Tutor PAT, Mathematics Tutor, and Why2-Atlas.
A Dialogue-based Tutoring System (DBTS) can perform both language processing and language generation and typically complements an ITS. In short, a DBTS simulates dialogue between a human tutor and a student. (AI-powered chatbots often encountered in commercial websites are now appearing in education platforms.) Dialogue-based Tutoring Systems enhance Intelligent Tutoring Systems because when a user writes or speaks a response to a question. DBTS can evaluate the response in terms of a level of understanding and also provides feedback to help the user understand the answer.
Since language processing and language generation are so fundamental to achieving an enhanced Intelligent Tutoring System (ITS), it is to be expected that GPT-3 technology will substantially improve ITS capabilities. GPT-3 will likely lead to enhanced personal assistants that can, among other things, explain difficult concepts at age-appropriate levels, provide detailed feedback on student written work, roleplay to further student understanding, take and summarize notes from various materials, and also serve as a search engine. (From a market perspective, it is difficult to gauge which companies will gain access to GPT-3. Moreover, GPT-4 is expected sometime this year.)
Already, there is an increasing number of commercially available AI-assisted educational apps. For example, PhotoMath is an AI-assisted math for learning Mathematics and AIDA is a AI-driven calculus learning app. Blue Canoe is an application for students to learn spoken English skills, while Amira is designed to accelerate reading comprehension. In addition, the popular Quizlet app uses machine learning to help students study learning sets. With access to GPT-3 all of these could provide an enhanced personalized learning experience.
And let’s not forget Learning Management Systems (LMS). They also integrate AI, which is often used to automate course management tasks. So, GPT-3 could be leveraged by teachers to automate scheduling and create course correspondence. But it could also be leveraged to help teachers analyze and grade student work, provide a virtual tutor for students, create explanations of difficult concepts, and help identify and address student knowledge or skill gaps.
How are schools reacting to ChatGP3 and how can they adapt?
ChatGPT is already causing significant angst in schools. The immediate concern amongst educators is that students will use ChatGPT to cheat on assignments. Educators are also concerned that students will be unmotivated to learn if AI can write their essays, solve their math problems, and more. Moreover, ChatGPT makes mistakes so information that students receive may be wrong.
The New York City school system, the nation’s largest, decided to ban students and teachers from accessing ChatGP3 on their system’s network. Other school districts may follow. Already, individual educators and academic departments are prohibiting students from using ChatGPT.
But, students can access ChatGPT on their home computers or cell phones, so any school ban will be of limited effectiveness. Furthermore, ChatGPT produces original content so it likely will not be discovered by an anti-plagiarism-detector. So, finding proof that students used ChatGPT to cheat on an assignment may prove difficult. In all, schools may find that any attempt to ban ChatGPT is ultimately fruitless.
But most importantly, GPT-3 technology will not someday go away and schools must plan accordingly. In all likelihood ChatGPT (and rival technologies) will improve significantly and become more prevalent in our society. So schools will be forced to reckon with GPT-3 and it may prompt many educators to rethink their instructional strategies.
In many ways schools still put much focus on routine cognitive tasks that can be automated or outsourced. For instance, teachers often assign generic homework questions whose answers can be looked up on Google. Filling in worksheets is still a common activity in many schools. Moreover, much of the rote content delivered in teacher lectures and course textbooks can easily be found online. So, ChatGPT may force teachers to reexamine their assignments and think of ways to personalize them to highlight student voice and identity.
One response would be for schools to provide opportunities for students to work actively with ChatGPT. Students could evaluate what ChatGPT does well, what it does not so well, and how it might be leveraged to help generate ideas and enhance creativity. For students struggling to get started on an assignment, ChatGPT can show students how they might organize their thoughts and help them create an outline to frame their ideas. For students “stuck” during the process of addressing a task, and not sure what do next, ChatGPT can engage them in conversations about potential paths they might follow — even roleplaying with students to develop creative responses. Since ChatGPT can create imaginary scenarios (such as how a historical figure might address a contemporary problem) it can provide creative, non-traditional information for students to consider. It can even remix a collection of student work — notes, an essay, a dialogue — and prompt new ways to articulate student ideas. All the while, teacher and student together could analyze ChatGPT responses and evaluate their effectiveness, embarking on a conversation about how to leverage it skillfully and appropriately.
As AI continues to automate many tasks, it presents us with a unique opportunity to focus on what makes us different than technology. Human thinking and actions are not limited by data inputs; when faced with the complicated dynamics of human problems, we develop creative, human solutions. Humans can also communicate emotions, context, insights, and experiences to each other; our social-emotional abilities extend beyond algorithms or statistics and ChatGPT will never match us in this regard. As we think about learning in an AI world, we should remember that humans have the advantage of mental flexibility and creativity over rule-based technology, and our schools can help enhance these abilities. As such, an AI-aware experience for students would be working with ChatGPT to access and evaluate new forms of knowledge, to think creatively with new information, and to demonstrate understanding in diverse ways.