The Future of Learning in the Age of AI: Transforming education
Fears about plagiarism from ChatGPT (or "AIgerism") are understandable, but they also show how far our thinking needs to progress when it comes to education. Certainly, students need to be able to think and formulate written or verbal texts independently of AI, but neither should we imagine that AI should merely adapt to our existing methods of teaching.
The Current State of Education
If nuclear fusion is always 30 years away, a school which is not based on passivity, rote learning, and sluggish tendency to conform towards the median student is always just on the verge of becoming widespread. Despite centuries of proposals for educational reform, the great majority of students spend much of their time sitting in chairs listening to a teacher, collecting facts to be later regurgitated.
Flaws in existing education
Unequal Access
In 2019, the adult literacy rate for those aged 15 years and older was 86.5%. (Source: World Bank)
The percentage of the world population with a college education is 7%. In 2014, the gross enrollment ratio in tertiary education was approximately 9% in Sub-Saharan Africa and 21% in South Asia.
According to the OECD, the average percentage of 25- to 64-year-olds with any postsecondary degree across OECD countries increased from 30% in 2010 to 39% in 2020. In the United States, during the same period, the percentage increased from 42% to 50%. (Source: OECD)
Students Left Behind
- While many innovative ideas have been developed to adapt to students' different learning styles, schools often lack the resources to put these into practice, whether in terms of instructor preparation or teaching time, or other resources.
Possibilities Opened Up by AI
Interactivity
AI can help create interactive environments, be it educational games, chat-based experiences, or interactive coding for people learning programming. It can provide instant answers to certain questions that previously required reading through documentation.
Adapting to Learners
AI can adapt information to different levels of expertise, and will be increasingly capable of transforming knowledge between text, audio, and visual formats.
Reducing bureaucratic constraints
Currently, instructors are like Gulliver tied down by the Lilliputians. Many instructors first priority is to help their students learn and grow, but often find themselves instead focusing on other criteria. Where rules, regulations and standards are helpful in designing effective schools, AI may help by easing the task of fullfilling them or reporting on them while freeing the instructor to focus on the learning priority. For example, the ability to evaluate more flexibly through AI might enable less reliance on standardized testing.
Increasing access
Hundreds of millions of students lack access to educational resources, but may nonetheless have access to a smartphone. While the ideal teacher is a human, AI may be able to increase access to an interactive mode of learning that could reach students with few alternatives, and connect them to a global community of other learners.
Principles of AI-Enabled Learning
The future of learning with AI does not mean all-AI all the time, as if our goal is to immerse ourself in the Matrix. Instead we seek to define a proper role for AI which delimits its use while also making the most productive use of it.
Minds should be cultivated
AI is like a tractor for learning - a powerful tool, but not necessarily something you should use to plow over your entire intellectual life.
Interacting with AI like GPT can be a stimulating experience. But it's only one type of experience. Cultivating one's mind involves also creating space for "offline" thinking - for calm reflection, daydreaming, conversation and debate with peers, friends, family, and fellow citizens.
Comparative advantage
Humans are users of tools. AI is a fancy tool. Calculators and computers did not ruin our ability to do math. Whatever AI learns to do, it learns to do based on the programming of humans and the texts and other sources generated by humans (possibly also cats).
Recognize the limits of AI
All of us need to have a better understanding of the limits of large language models (LLMs) and other AI systems. This is a fast-moving, dynamic process. Even where clear flaws have been pointed out, such as the tendancy of LLMs to "hallucinate" incorrect answers to questions, or to exhibit bias against ethnic groups or other identities, researchers are quickly moving to address these problems, while also adding new capabilities that might reduce or increase those problems, such as the ability to incorporate internet search results.
Ethics
New rules norms, and practices need to be developed to clearly delineate human- from AI- or hybrid-generated texts or information, as well as to trace the source of facts, claims, photos, and other media to their origins.
Recommendations for Educators and Policymakers
1. Learning first, AI second (or 3rd, 4th...)
AI is not a panacea. Educational plans should not be judged naively by wether or not they incorporate the latest buzzwords from AI.
2. Don't be an ostrich, teach.
The AI horse is out of the barn. You may not like AI, and you may have reasonable fears about its consequences. If so you should advocate for rules or guidelines or regulations that prevent harm. But it is not enough to meekly hope it will just go away, any more than the Cold War meekly hoped that nuclear weapons could be un-invented.
3. Be willing to innovate, and allow innovation
Adapting to AI will take time, many iterations, and mistakes. Bureaucracies need to give instructors the space to explore new tools, even if in some cases the results are not encouraging. Ideally, this testing should be done a systematic way, such as through randomized control trials.
Sources and Further Reading
- On average in OECD countries, 4th and 8th grade students say that about 73-77% of class time is spent on "lecture-style presentations", while teachers say it is closer to 20-26%. See more data and breakdown for OECD countries at stats.oecd.org. Go to Education and Training / Measuring Innovation in Education / Part 2 Innovation as change in classrooms and schools
World Bank EdStats Series SE.ADT.LITR.ZS
- https://ourworldindata.org/tertiary-education, citing World Bank EdStats "Total enrollment in tertiary education (ISCED 5 to 8), regardless of age, expressed as a percentage of the total population of the five-year age group following on from secondary school leaving."