November 2023 Issue
ARTIFICIAL INTELLIGENCE IN EDUCATION by Sangeeta Dutta
Advancements in Computer Technology have revolutionized education especially conventional teaching methods. One such tool is Artificial Intelligence (AI). Imitation of human intelligence processes like speech and visual recognition, translation of the languages and virtual decision-making by machines and robots have become the bedrock of AI. The ability of machines to think and behave like human beings, has given AI a special place in today’s education.
Over the years, a computer has been generally used as an educational resource like a library or laboratory, as well as for maintaining databases of student information. What we have found is the inability of computers to conduct conversations with the student in his/her mother tongue. Computers find it difficult to understand the subject being taught, thus not being able to diagnose unanticipated responses. It also finds difficulty in deciding what should be taught next. Further it is unable to anticipate, diagnose, and understand the student’s mistakes and misconceptions. Thus, it becomes difficult to improve or modify current teaching strategies or learn new ones. What if we could overcome these obstacles? Thankfully, today AI has made major strides in natural language understanding, knowledge representation, planning, expert systems and learning.
Research within the field of artificial intelligence (Al) is having a positive impact on educational applications. For example, there now exist systems to teach or tutor many different subjects. Major research has happened in the development of learning environments that are designed to facilitate student-initiated learning. Another major application is the use of expert systems to assist with educational diagnosis and assessment. There are programmes where the student is actively engaged with the instructional system, and the student’s interests and misunderstandings drive the tutorial dialogue. Many such tools and techniques from AI have been employed in order to make these systems both flexible and sensitive.
Today’s systems are capable of analysing a wide range of student responses using domain expertise in the particular topic that it is teaching or tutoring. Expert systems have now been developed in a variety of areas, including chemistry, geology, computer-system configuration, medical diagnosis and mathematics. Often it contains a student model that represents both the student’s knowledge and misconceptions, as well as a component containing information regarding appropriate teaching strategies. What the student has mastered as well as difficulties, misconceptions, and unmastered skills are gleaned from existing answers, implicit problem-solving behaviour, and historical information regarding the student’s past experiences. There are sophisticated algorithms for determining which skills the student has not mastered; this can be a nontrivial task, especially when the student’s difficulty arises on a task requiring multiple skills. Another problem is bridging the gap between the various steps of a student’s solution; this may involve conjecturing missing steps or discovering the underlying motives or plans that led to the student’s observed sequence of steps.
Currently, AI systems can make intelligent decisions regarding what, when, and how to teach. Such tools employ expertise to determine whether a student’s answer is correct. Sometimes rather than just checking a student’s answer as an end-product, systems evaluate the process or procedure by which the student achieved the answer. In so doing, the system determines where the student has gone astray and hence finds which skills are lacking. The system then provides explanations, hints, and further examples and a variety of teaching strategies and selects the most appropriate for the student and task in question. A good teacher does not employ the same teaching methodology for all students or all lessons. Teaching strategies mostly coach the student within a particular activity, such as a game-playing situation; the intent is to manipulate the environment and the coaching such that a particular skill or general problem-solving ability is acquired. They question the student to encourage reasoning about current knowledge and to modify or formulate concepts; this may involve simulations or games in which the student can discover facts or laws; providing tasks for the student and evaluating the responses in order to detect the student’s misconceptions.
Today for example, AI has mastered the design to teach facts about world geography. Instead of storing geographic information in the form of prewritten frames, geographic facts about industries, exports, populations, and capitals are designed to manipulate the knowledge base to generate factual questions, to answer the student’s questions and to evaluate the student’s answers. There are examples of mixed-initiative systems where both the system and the student can initiate dialogue by asking questions. Sophisticated tools can model a variety of projects, including making movies, composing tunes, writing concrete poetry, as well as investigating both simple and complex domains, such as elementary geometry, elementary motion, and balance problems in physics, differential geometry, and Newtonian physics. We all know about Aesop’s well-known fable regarding the tortoise and the hare.
Today, this can easily be programmed by AI to simulate the tortoise and also represent the more complex, unpredictable hare.
There are AI tools which provide the student with a learning environment in which to acquire problem-solving skills by trying out ideas. It helps the student to develop, test, and debug appropriate hypotheses. The student is presented with a malfunctioning piece of electronic equipment. He must locate the faults by taking appropriate measurements. The student can ask questions about the measurements or which hypotheses should be considered. When the student forms a hypothesis, the AI tool evaluates it and, if necessary, helps the student debug it. Improvements have been incorporated to include troubleshooting games and sophisticated debuggers and explainers, including some tools necessary for student modeling and coaching. An important component of any teaching system is a method for representing, diagnosing, and correcting misconceptions or bugs.
Another important component is the interface with the student through a conversational system capable of a natural language dialogue. Computers simply do not have the same level of flexibility or adaptability as a good human teacher. Such systems are still restrictive in their use of natural language. The natural language interface is based upon semantic grammar. In semantic grammars the possible statements to the system are characterized in terms of the underlying concepts of the domain. As the domain is small, hence, there is a limited number of underlying concepts with which the system can deal.
Today’s AI environment is conducive to spontaneous learning, the kind of learning by which a young child learns to talk or to walk. This is developed in accordance with two fundamental heuristics of spontaneous learning. First, start from previous knowledge; a person is not able to make sense of a new experience and assimilate it unless it can be related to previous experience. Second, the learner should use the new ideas to ‘make them his own’: concepts are learned and remembered if they are important to the learner. In a programming environment, the child is the programmer; he/she manipulates the equipment, controls and builds the programming environment. It scores an attempt to provide an appropriate, friendly environment in which student-initiated learning can take place.
AI also looks at an important aspect of teaching which is the ability to anticipate and diagnose a student’s misconceptions. More than simply noting the child’s errors, the tool is able to determine the underlying cause of the errors. Some of the systems make a concerted effort in this regard and attempt to maintain an accurate and current model of the student’s knowledge, skills, errors, and misconceptions. There are children within the regular school system who are experiencing severe learning difficulties. In order to plan an appropriate instructional program for a child with learning problems, the exact nature of the child’s difficulties is determined. This educational diagnosis is initiated and assessment takes place within the regular school environment by a teacher who is familiar to the child, preferably the regular classroom teacher or the resource-room teacher. The teacher or diagnostician would perform the required task, such as obtaining the requested data, or administering the appropriate test, and would supply this information to the system. After this new data had been assimilated and analyzed, the system would propose the next step, and so on. Eventually, the system would provide a summary of its diagnostic findings along with a prescription, including appropriate remedial activities and instructional techniques.
Expert systems guide the user through the various stages and levels of diagnosis from the initial suspicion that a reading problem may exist, through to the point at which sufficient information has been gathered to plan an appropriate remedial program. Assessment begins with gathering of relevant data concerning the child’s physical, mental, emotional, social, and academic developmental history. In addition to the assessment of the child’s general skills in academic areas such as reading, spelling, and arithmetic, the expert system examines psycho-educational correlates that include those intellectual, visual auditory, and language skill deficiencies that might be related to learning disabilities. As the assessment of the child’s learning disabilities progresses, academic skills are subjected to finer and finer scrutiny until the nature of the child’s problems has been pinpointed exactly. For example, within the area of reading, one might assess the child’s knowledge of sight words, decoding skills, phonics skills, reading comprehension, etc. Within each of these categories there is an enormous amount of information which is collected. For example, when reading aloud, what type of errors does the child make: substitutions, omissions, word reversals, etc.? Does the child have more difficulty reading isolated words than short paragraphs, where context clues can be exploited? Does the child make more errors at the beginning of words, or are there particular word parts or endings that cause difficulties? Just within the area of phonics there are many skills that are assessed. For example, one can examine the child’s ability to decode words containing single consonants or consonant blends at the beginning, in the middle, and at the end of words. The tools examine whether there are particular consonants or combinations that cause difficulties? Are there certain substitutions that the child makes? What vowels or vowel combinations create problems? There are many skills which are assessed, and phonics is a small subset of the general area of reading. There are many related abilities which are assessed, such as interpreting pictures, developing sentences or short stories, expressing one’s own ideas or those presented in a given story. Because reading is not an isolated ability but a collection of many interrelated skills, assessing a child’s reading is a long, arduous task. Hence, part of the expert’s job is to determine what to assess in each individual case and to perform an efficient diagnosis with regard to both the expert’s and the child’s time.
Examples as mentioned above are some of the advances made within the AI research areas, such as knowledge representation, natural language understanding, reasoning inferencing and learning/discovery. These will undoubtedly continue to contribute towards the educational field.
Sangeeta Dutta is an Indian national who spent the better part of 27 years in Muscat – therefore, beautiful Oman is her second home. She loves the friendly people and outdoor life in this country. Her hobbies are gardening, listening to music, and going for early morning long walks, and visiting art galleries and museums. She likes to visit new countries. Apart from a Master’s in History and a Bachelor of Education, she has completed several certified curricula and professional development courses. She has been teaching in Muscat with an international experience of nurturing inclusiveness and diversity through teaching in multi-cultural, multi-skill level, and multi-age group classrooms. She builds a strong rapport with children, considering their moods and personalities, and has a passion for nurturing creativity and curiosity in them. She cherishes teaching because each year is a new journey and brings with it fresh challenges.
This article was part of an e-book, Perspectives on Education and Learning: Where Is Education Head? Login to our portal to read.