ARTIFICIAL INTELLIGENCE IN EDUCATION
A Revolution in Learning, a Commercial Juggernaut, or a Stealth Technology Insinuating Itself into Districts, Schools, and Classrooms?
Here’s the short answer. We think artificial intelligence (AI) in education may be all three: revolution, juggernaut, stealth technology, and here’s why . . .
In 2019, members of Citizen Advocates for Public Education-Ohio (CAPEOhio.org) focused on artificial intelligence (AI) after it made repeated appearances in the news about scientific breakthroughs. We were curious about whether AI was having equally ground-breaking impacts on learning and teaching.
We began to gather information on AI. As “citizen advocates” and not data scientists, we initially found the topic rather dizzying. We began to see quickly, however, even as non-experts, that there were clear, rapid, ongoing developments for AI in education. Few of these developments were positive.
AI Gets Great Press
AI often gets great press. It is now an entrenched and invaluable partner in scientific inquiry. There are countless examples of connections between AI and scientific progress. AI and the ubiquitous use of algorithms is also an outwardly benign fact of our daily lives. AI algorithms advise us about our potential choices in entertainment, music, or news articles, what we might be interested in purchasing, where to turn while navigating unfamiliar landscapes, people we should know about, and groups we might choose to join on social media. A debate has even surfaced as to whether some AIs have crossed the barrier into sentience. At the same time, AI ethicists are admonishing AI developers, who are both competitors and allies on an unregulated technological frontier, to develop a “code of conduct.”
It is accepted that AI will continue to expand and, increasingly, intrude in our lives (2022 AI Index Report). Global forces driving the development and adoption of AI are predicted to cause dramatic shifts in the nature of work. In the midst of its pervasive growth, abuses of AI have emerged, and there are ongoing legislative efforts at AI regulation. In the U.S., a set of recommendations regarding AI protections for citizens has been drafted to provide guidance to government agencies and, perhaps, private corporations. Moreover, the ability to “create” AI will eventually rest in the hands of anyone who cares to learn a nontechnical approach to AI development. This growing “no code” industry is developing programs that will allow non experts to use AI to perform tasks like visual recognition with simple icons rather than any knowledge of coding.
Many educators have already become de facto, boots-on-the-ground AI users out of necessity. Each day they manage AI for and with their students and contribute to the ever-increasing amounts of data public education produces. Recent changes – many driven by the Covid pandemic – have redefined “classroom” and “school day” for many students. Resulting rapid adoptions of new technologies and the expanding spectrum of students’ needs, leave busy educators with little time to ponder the details of how AI works or to contemplate the practical and ethical implications of AI in the future of education.
For individuals who have the time and interest, however, the internet offers a universe of information about all things AI: news items, videos, magazine articles, studies in research journals, websites, Twitter interest groups, AI conference proceedings, press releases, book reviews, university departments, online discussions, marketing promotions, TED talks, grant proposals.
Locating information about AI in education, however, can be challenging. Search terms range far beyond the obvious Google query of “AI in Education.” Searches for relevant AI information include terms such as machine learning, learning platforms, personalized learning, intelligent tutoring, and learning analytics. To understand how a personalized prescription for learning is constructed, searches expand to the field of psychometric measurement which includes computer adapted testing (CAT), item response theory (IRT), Rasch methodology, logits, and growth measurement. It remains a challenge to find any independent examinations of the instructional effectiveness of combining computer adapted testing (CAT) with AI-selected resources – current state of the art for most learning platforms. Because these personalized learning systems are proprietary, it is difficult to see the mechanisms that are operating. For most educators this process remains a “black box,” a typical feature of AI-powered systems.
The following materials represent our attempt at something of a curated set of current AI information relevant to education. This is a survey of publicly available resources. Keep in mind that many academic studies reside behind paywalls.
Artificial Intelligence in Education
Returning to our initial question, “Is artificial intelligence in education a revolution in learning, a commercial juggernaut, or a stealth technology insinuating itself into districts, schools, and classrooms?”
We found that AI in education is all three. AI may eventually cause a revolution in learning. AI in educational settings is quickly evolving into a ubiquitous commercial juggernaut. And, AI has shown up in classrooms largely unannounced and stealthily because of recent shifts to more virtual instruction and/or the promotion of personalized learning systems. These current trends of AI’s impact on education are likely to intensify as educators search for solutions to what the media has labeled “learning loss” related to COVID 19.
CAPE’s deep dive into AI resulted in several conclusions about AI’s current and future impact on students, educators, and schools. These conclusions are derived from not only our current research into AI, but also from our members’ many years of experience working in and with public schools in Ohio.
After our experience in researching this topic, we would also like to insert a note of caution about how easily thinking about AI becomes anthropomorphic. There is a tendency to view AI as live, animate, even willful, when AI programs exhibit human-like communication abilities. We encourage you to resist this line of thinking. AI is very advanced computer programming. Although a handful of experts believe differently, at present that’s it. AI is lots and lots of computer coding piled up until it actually becomes a “black box,” even to the developers.
Conclusions On Artificial Intelligence in Education
- AI is ubiquitous and, at present, largely unregulated, requiring educators to exercise due diligence when bringing AI into classrooms and school districts.
“It’s a Wild West out there for artificial intelligence. AI applications are increasingly used to make important decisions about humans’ lives with little to no oversight or accountability…”
“Today we encounter AI as our distant ancestors once encountered fire. If we manage this technology well, it will become a tremendous force for global good, lighting the way to transformative inventions. If we deploy it too quickly and without adequate foresight, AI will burn in ways we cannot control.”
2. AI should be part of the curriculum for both educators and students, including learning about who develops it, how it works, where it is used, and how to identify and evaluate AI’s positive and negative impact.
“Machine learning is changing, or will change, every industry, and leaders need to understand the basic principles, the potential, and the limitations…”
3. While AI continues to advance, it will be some time– if ever—for AI to think, reason or make decisions like a human. Unlike some other fields, AI in education is in its very early stages of development and effectiveness.
“I think I know what intelligence is; I think I know how brains do it. And AI is not doing what brains do… The vast majority of AI researchers don’t really embrace the idea that the brain is important. I mean, yes, people figured out neural networks a while ago, and they’re kind of inspired by the brain. But most people aren’t trying to replicate the brain. It’s just whatever works, works …”
4. In learning with and teaching about AI, it is important to identify trustworthy resources for policymakers, educators, students, parents, and members of the community.
“As the information technology sector has become more financially and politically powerful in the last decade or so, the voice of Silicon Valley has grown louder in the debates about the shape and direction of the education system. Many of its entrepreneurs have launched or invested in education businesses, often proudly ignorant of the history of education technology.”
“AI is a term used to describe a machine or computer program that uses features of human-like thinking, such as planning, problem-solving or logical action, to undertake a task. Many common computing applications, such as internet search engines, smart phone assistants, and social media facial recognition tagging technology, are powered by AI…. AI is usually invisibly infused through computing applications that can help us enhance our knowledge and judgement, and connect with others.”
“While we are probably far from creating machines that are self-aware, we should focus our efforts toward understanding memory, learning and the ability to base decisions on past experiences. This is an important step to understand human intelligence on its own.”
5. Machine learning is not the same thing as human learning. Much of current AI development arises from machine learning theory (pattern recognition within massive amounts of internet data scraping, reinforcement learning, word guessing, simple neural networks), which differs from current theories about the complexity of human learning (social, contextual, constructivist, identity-based, developed with a malleable brain impacted by social/emotional experience). This may result in disappointing or even dangerous unintended outcomes as AI in its current form fails to map out the complexities of how the human brain develops and learns over time.
“. . . Until a few years ago, I was a great believer in what might be called the ‘engineering’ model of personalized learning, which is still what most people mean by personalized learning. The model works as follows: You start with a map of all the things that kids need to learn. Then you measure the kids so that you can place each kid on the map in just the spot where they know everything behind them, and in front of them is what they should learn next. Then you assemble a vast library of learning objects and ask an algorithm to sort through it to find the optimal learning object for each kid at that particular moment. Then you make each kid use the learning object. Then you measure the kids again. If they have learned what you wanted them to learn, you move them to the next place on the map. If they didn’t learn it, you try something simpler. If the map, the assessments, and the library were used by millions of kids, then the algorithms would get smarter and smarter, and make better, more personalized choices about which things to put in front of which kids. I spent a decade believing in this model—the map, the measure, and the library, all powered by big data algorithms. Here’s the problem: The map doesn’t exist, the measurement is impossible, and we have, collectively, built only 5% of the library … We also don’t have the assessments to place kids with any precision on the map. The existing measures are not high enough resolution to detect the thing that a kid should learn tomorrow.”
AI Thinking Is Not the Same As Human Thinking
6. Given that “AI thinking” is not the same as human thinking, great caution should be exercised in adopting AI-powered systems that purport to do the thinking of teachers or supplant their role in instruction.
“We want to avoid teachers and students using AI-systems that ‘feel more and more like magic’ and where educators are unable to explain why a machine made a decision that it did in relation to student learning. The very basis of education is being able to make ‘fair calls’ and to transparently explain educational action and, importantly, to be accountable for these decisions.”
“Just because the algorithms want a kid to learn the next thing doesn’t mean that a real kid actually wants to learn that thing. So we need to move beyond this engineering model. Once we do, we find that many more compelling and more realistic frontiers of personalized learning opening up. Which brings me to the question that I hope might kick off your conversation: “What did your best teachers and coaches do for you—without the benefit of maps, algorithms, or data—to personalize your learning?” There are many ways to answer to this question. Each might be a doorway to the future of personalized learning.”
7. AI can be biased because an AI is only as good as the data set on which it has been trained. Data sets typically are selected by humans potentially resulting in outcomes that are neither objective nor neutral. This can be dangerous. AI’s literal, unquestioning acceptance of data, reliable or unreliable, can be seriously limiting given the decisions it may be asked to make. Even the most sophisticated AIs, which are experimental large language models [LLMs], are just now beginning to interpret, evaluate, or synthesize data to produce something that resembles “thinking.”
“Artificial intelligence is not an objective, universal, or neutral computational technique that makes determinations without human direction. Its systems are embedded in social, political, cultural, and economic worlds, shaped by humans, institutions, and imperatives that determine what they do and how they do it. They are designed to discriminate, to amplify hierarchies, and to encode narrow classifications. When applied in social contexts such as policing, the court system, health care, and education, they can reproduce, optimize, and amplify …”
8. The largest entities in the tech industry itself, those which have the most to gain financially, are primary funders of research and development in AI.
“There are some obvious contenders when it comes to commercial AI labs. U.S. Big Tech — Google, Facebook, Amazon, Apple and Microsoft — have all set up dedicated AI labs over the last decade. There’s also DeepMind, which is owned by Google parent company Alphabet, and OpenAI, which counts Elon Musk as a founding investor.”
9. Corporations compete by promising to improve education by managing data and instruction with AI technology, although it may not be advertised as AI per se. A vendor or corporation ends up possessing student data because AI is embedded in collecting, maintaining and/or sharing student data from daily classroom activities. This is a major privacy issue. AI, which is rarely bound by laws or government regulations, is often imported into the classroom stealthily without regard to the privacy issue.
“Google positions the company ‘as a free public service, divorced from marketplace contexts and concerns’… but this framing allows the company to obscure the intent to surveil students to create algorithmic identities. Google profits from securing future customers, profiling students for marketing, and extracting data.”
“The complaint says Google has infiltrated over half of the nation’s primary and secondary schools by offering free software and Chromebook laptops, and it alleges the company is illegally using the products to mine personal information of students under 13 — including their physical locations, browsing histories, contact lists, voice recordings and passwords.”
10. The impact that any AI-based instructional system might have on student learning–and the potential ethical concerns raised in using AI–need to be evaluated beyond claims from developers, corporations, vendors, or sales staff that promote AI in education.
“Adopting commercial digital platforms and learning programs can pose real risks to the integrity of schools’ curriculum and teaching. School and district leaders can minimize the risks by judiciously choosing and using products they adopt.”
“Controversies over data collection and sharing are likely to intensify with the expansion of [Google] Classroom. … research [has]highlighted how hundreds of external education technology providers are integrated into Classroom, potentially enabling Google to extend its data extraction practices far beyond the platform. The road map for Classroom confirms its plans to extend these integrations, through a “marketplace” of “Classroom add-ons” that teachers can then assign without requiring extra students’ log-ins. This makes Classroom itself the main gateway for students to access other non-Google resources.
These developments give Google extraordinary gatekeeping power in the education technology industry, as it sets the rules for other third-party providers to integrate with Classroom and for the exchange of data between them. In its new role as an LMS [learning management system], Classroom can even integrate with existing school information systems, acting as the key interface between a school and its student records.”
11. Large-scale studies documenting any lasting effects of AI on student learning don’t yet exist and may never exist because there may not be a disinterested funder large enough to conduct or sponsor such studies. Open-source educational resources (OERs) coming from diverse, often untested programs are now in the mix with AI-based learning platforms. It will be very difficult to determine whether such “personalized” models have any actual beneficial effect.
“. . . Technology in education has long been expected to have revolutionary impacts on learning. Computer-assisted instruction was expected to be effective because it places students at their precise level of proficiency, so that they need not repeat content they already know, and it advances students at their own pace, so that they can never fall behind. The computer, it is said, is patient, giving students as much time as they need to master the content, but moving forward rapidly if they are succeeding. The computer immediately provides answers, so students need not practice errors, but can correct themselves and move on. Every one of these points was made in a 1954 film that still exists on YouTube (“Teaching Machine and Programmed Learning”). Yet 67 years later, it is clear that technology programs based on these arguments, no matter how sensible they sound, have not transformed the outcomes of learning, not even in elementary mathematics”
https://www.vox.com/23815311/science-of-reading-movement-literacy-learning-loss Conclusion: The sciences of reading continue to evolve. While cognitive scientists have been mapping brain functions related to reading, the programs that use this research to actually improve students’ reading performance have yet to be developed. It’s “buyer beware” when adopting programs marketed as “science of reading” products ….
The Research Is In! Teaching Is Life-Changing What really moves the needle for educators and, by extension, their students? Being an effective teacher is about more than just improving test scores—it’s also about making a difference in students’ lives. When we asked our readers to describe the traits of a life-changing teacher, they said that great…
Learn More About the Issues and Challenges. Listen to “Looking Forward Our Way” Our guests for this episode are Dr. Margaret Kasten and Dr. James Bishop. They both give a better insight into what CAPE-Ohio is about and how we can do better with our public schools. Looking Forward Our Way Get new content delivered…