The world of education is going to be deeply affected by the introduction of new AI-based technologies, and that's a fact. However, it is hard to tell if those changes are really going to push toward a positive evolution of our society. Education, in general, has a tremendous impact on our entire society and is one of the cornerstones of human evolution. The science of learning and instruction has changed significantly over the course of the last century, and it may be argued that many of the current behavioral changes of the latest generations can be attributed to the evolution in education we have witnessed. Increased use of artificial intelligence in education certainly holds immense potential for improving learning and teaching, but are these improvements going to build a better society and a better world?
The Current Scenario
Whether the results are going to be good or bad, AI in education is going to boom. According to recent reports, the sector's growth has been predicted at 47.5 percent through 2021 in the U.S. market alone. Machine learning has already been added by some of the largest tech giants in the tools used to aid students in performing their tasks. For example, IBM’s Watson Analytics is able to answer natural language questions about info included in its database, while Google's G Suite for Education app uses natural language processing to write complex formulas at the request of students and teachers. (For more on machine learning in education, see How Machine Learning Can Improve Teaching Excellence.)
As a side note, here we can already see one of the potential unexpected generalized effects of implementing AI in schools. Voice chats are becoming the latest technology trend and a must-have in many businesses. AI can now perfect its ability to recognize and understand human voices by feeding on a data set as huge as the entire education system. How long will it take before all offices will start using talking AI to stimulate meaningful and efficient communication and collaboration between team members? Am I the only one who's thinking about Mass Effect's AI EDI here?
Things are not so different overseas as well. In China, semi-sentient robots are already being used to automate the grading process, reducing the workload of teachers. Their smart artificial minds can understand the general logic and meaning of an essay and generate an almost human-like judgment about its quality. And at least 60,000 schools have already implemented them with apparently great outcomes.
The Amazing Potential
One of the most evident AI benefits is the ability to automate menial operations, expediting many administrative and organizational tasks. Checking homework, grading papers, looking through illness records and absence sheets, and preparing report cards are just some examples of the tasks where educators spend most of their time – tasks that an AI can perform with almost no errors in just a few minutes.
AI can also help digitize textbooks and create customizable "smart" content for students of all age ranges, helping them with memorizing and learning. Virtual characters and augmented reality can be powered by AI to create believable social interactions such as those experimented with by the University of Southern California (USC) Institute for Creative Technologies. These virtual environments can be used to assist students in their endeavors and learning process, or as substitutes for tutors, lecturers and teaching assistants. No one can ever work all day and night and provide students with 24/7 responses... unless he or she is a robot, of course!
Drawbacks and Controversies
So far, everything about AI and education has seemed amazing, hasn't it? Things are never so simple in the real world, however. To achieve the results it is designed for, AI requires one thing above all else: data. Data must be fed to the algorithm so it can "learn" about the environment, and which ones are the "good" and the "bad" outcomes. But what if the entire data set about student learning we're using is, at best, unreliable if not completely worthless?
For example, the vast majority of the studies that try to measure student learning use uninterpretable or unrealistic metrics such as self-reported "learning gains" or (even worse) student grades. But what does a student grade measure other than acting as an extremely vague performance indicator? Recently, during an experiment that received significant media attention, an AI was able to pass the UK's GP (general practitioner) exam, obtaining a superb 81 percent score. This "grade" is, therefore, nothing but a final score – which does not reflect in any way the validity of the learning process or the teaching method, either for the AI or for any other student. But that's the only data we can easily collect, even if it lacks any educational meaningfulness. How much time will humans need to learn how to "cheat" AI-driven tests and grab positive grades with little or no effort?
By solely focusing on performance, the risk is to focus on marginal or irrelevant learning theories. Current data sets draw their data from a wide range of educational databases, yet many of them are old, and the teaching methods used are obsolescent. Teachers who spent decades teaching a class are not necessarily better at their jobs than those who are younger, simply because there's a huge difference between what our society is now and what it was 30 years ago. Yet, all this data is merged together in an unintelligible swamp of info that AI cannot really discriminate more than its designers could. (To learn more about advances in education, check out Virtual Training and E-Learning: How Digital Technology Is Paving the Future of Advanced Education.)
AI can stimulate technology addiction and further make our future generations highly dependent on all kinds of devices if their exposure starts with childhood. Especially if the alleged "quality" content that AI will use to teach is drawn from an immensely vast pool of junk content that was selected by a handful of companies.
AI can help skyrocket our ability to educate and teach new generations, freeing a lot of time for human professors who could (in theory) focus solely on the things that matter.
However, this fantastic world of efficiency comes at a steep price. If we're not careful, we risk providing our students with low-quality content, taught in the wrongest way possible, that they can still avoid studying by cheating their AI teachers. If we do not want to live in a society full of cognitively passive, socially-unadapted adults addicted to technology, we need to adjust the sights now rather than later.