Imagine a world where every student has access to quality education, regardless of their socioeconomic status or location. This is the vision shared by the United Nations in its Sustainable Development Goal number 4 (SDG 4). Achieving this goal has proven challenging, particularly in underserved populations and low- and middle-income countries around the world.
There’s near unanimity towards education as the foundation of a prosperous and thriving society. And yet, the lack of access to quality education and skills gap persists on a global scale. In the aftermath of the COVID-19 pandemic, progress towards SDG 4 has been limited, and in fact conditions in critical parts of the world have worsened. One of the few silver linings of this period was the uptake of technological solutions, namely online learning.
The recent explosion of AI technology, and the ensuing claims about its transformational potential, beg the question: Is AI a bona fide game-changer in global education? Can AI help us understand what’s getting in the way of progress, create and deliver opportunities for effective learning experiences in all contexts, and meaningfully and substantially contribute to SDG 4 (and all 17 Sustainable Development Goals)? And to what extent is the development roadmap for AI taking into consideration the whole of humanity’s biggest and most urgent problems, in ways that are accessible and reproducible to all?
Quality education, SDG 4 and Artificial Intelligence: An overview
AI has, at least in the imagination of professionals all over the world, the potential to be a game-changer in education and learning. From tutoring and homework assistance, to the development of next-gen personalized learning experiences from the ground up, speculation about AI’s potential existed long before the advent of ChatGPT and the so-called Large Language Models (LLMs).
Education has long been considered a critical tool for sustainable development. Among the indicators targeted by SDG 4 we have:
- Enrollments, participation and completion rates of children and young people in educational programs
- Completion rates and parity by sex
- Proficiency in subjects like reading, math and ICT skills
- Level of psychosocial development, including well-being, health and achievement of comparable learning benchmarks
- School services coverage and availability of teacher qualifications
In properly equipped schools and educational regions, we’ve already witnessed the introduction of AI-powered educational tools, and this trend will only grow in reach, volume, depth and sophistication. A growing number of young learners are becoming familiar with AI and the skills required to both embrace and further its development.
Unfortunately, much like other means of progress towards sustainable development brought by Information and Communication Technologies, these benefits have largely arrived solely for already advantaged learners. Those on the lower rungs of the national income ladder—ultimately, those learners who need the most support—struggle to access the benefits of this progress, exacerbating existing inequalities.
7 challenges and limitations of using AI in service of SDG 4
New AI developments will continue to yield only marginal benefits in low and middle-income regions if investments in AI tool creation aren’t also accompanied by investments dedicated to providing equitable access.
Unchecked development of AI could generate negative consequences for the poorest nations and learners in other ways. For example, if the rate of investment in computing power required to fuel newer generations of LLMs and other models is to be maintained (or accelerated), the negative environmental effects, including CO2 generation, e-waste creation, and water availability, will only worsen. Many emerging economies are also dependent on wealthy nations offshoring or outsourcing certain tasks and roles. Many of these duties, whether low-skilled or “white collar”, are already finding in AI a more than suitable replacement.
Even assuming ICT investments take place in the short and medium term, with an emphasis on increasing access to the connectivity required for everyone to make use of AI and other tools, a complex set of issues remain. These include:
№ 1. Providing equitable access to quality educational content, experiences, and opportunities
Among the massive and constantly growing list of AI tools already available for learning and education, there’s a limited number that are free or affordable in the long run, freemium versions notwithstanding. An even smaller number are open-source tools that are both readily available for disadvantaged educational systems, and not prohibitively expensive to use.
№ 2. Guaranteeing access to professionals properly skilled in both education and AI topics
A key factor in the quick adoption of LLMs in recent months has been usability. Through almost literally intuitive conversation-based interfaces, a sprawling cottage industry of “text-to-anything” solutions is cropping up, wherein the “anything” could be code, images, video, entire apps, as well as text itself. Specialists will still be required in this field, and guaranteeing access to them will be a key challenge going forward. Indeed, we’ve also witnessed the rise of the latest (and first?) LLM-based occupation: the prompt engineer, a person capable of getting the best possible output out of these systems.
Whether it’s old skills like teaching or new skills like AI development and implementation, the demand for top talent is only increasing, and wealthy economic centers are likely to capture most of the talent.
№ 3. Accounting for AI’s inherent bias, lack of consideration and knowledge about low income contexts
The discussion on AI bias is broad, deep and complex and beyond the scope of our discussion. Nonetheless, AI bias presents significant challenges in lower-income contexts. For example, given how the corpus used to train LLMs is an internet repository of text, they capture perspectives, tones, styles and problems that are significantly more representative of the well-off populations. Furthermore, it’s unclear whether ChatGPT and similar tools can realistically provide help to learners and teachers on topics that are not currently found in their training datasets. (Paradoxically, since it wasn’t a thing by the end of 2021, ChatGPT cannot help you become a better prompt engineer to improve the work you give to itself.)
№ 4. Overcoming limited understanding of the skills gap and forward-looking competencies needed at regional and national levels
It has been difficult enough to identify future competency demands and skills gaps even in contexts where the “Industrial Revolution 4.0”—and “Education 4.0”—have been openly embraced. A further (perhaps more macroeconomically oriented) bias could potentially arise from AI support tools that try to answer local upskilling challenges with solutions stemming from completely different economic, educational and institutional models, not to mention social idiosyncrasies.
№ 5. Preventing misuse of AI technologies, and misunderstanding of their benefits and risks
If the way AI tools are introduced to society continues without “guardrails” or coordination between institutional, government, and other stakeholders, nascent problems will become widespread, joining other recent structural issues we’re dealing with. Societal information and communication troubles involving AI are about to become part of our lives, heaped upon modern issues we’re already dealing with such as fake news, unrealistic beauty and lifestyle standards, disinformation, and societal polarization.
№ 6. Solving our inability to assess effectiveness, and the problem of certifying AI-based education
It’s unlikely the world will arrive at a clear set of guidelines about the do’s and don’ts of AI in learning and education before the appearance of an avalanche of fully AI courses, programs or even entire degrees unaided by humans. Considering the challenges that institutions and learners already face in proving the robustness and value of their qualifications, one can only imagine the challenges of making “AI degrees” valid in the eyes of employers, certification boards, and other key institutionality.
№ 7. Increasing trust in AI, transparency and ownership
As interesting and as valuable as AI and LLM-based tools have become for millions of people already, the controversy around “who owns what” can only grow in intensity. Both from the private sector and government bodies a spirited conversation is drawing L&D leaders and policymakers, as haphazard as it might be for now, simultaneously receiving criticism that they’re too indifferent to the negative consequences of the tools they’ve created, and too ready to bow to calls for censorship. It becomes apparent that a well organized forum of plural views, perspectives and education on AI represents a step forward to concertation and cohesive action. Another major point of criticism involves whether intellectual property rights were violated when the AI models were originally trained using huge libraries of existing content. Add these to long-standing conflicts between “big tech” and society, such as data privacy and surveillance concerns, as well as the transparency, accountability and explainability of AI, and you’ve a complex web of issues to unravel.
One lingering question, especially in relation to the UN’s SDG 4, involves the ownership and distribution of economic benefits of AI usage, as well as the burden of negative impacts. The widespread introduction of AI across developmentally contrasting parts of the world, and an excessive or exclusive focus on AI towards profit motives, could increase income inequality and impede the vast majority of workers from reaping any of the economic or financial rewards of any incoming age of productivity.
Prompts for optimism?
Despite these challenges—and the many we haven’t had space to touch on, such as AI plagiarism—there’s widespread optimism about the role AI will ultimately play in education and learning experiences. After all, we’re reaping benefits from AI left and right, and have been for several years.
For starters, there’s a generalized agreement about the importance of establishing rules, from standards of practice all the way to fully fledged regulatory frameworks. These rules would provide guidelines for both educators and learners offering best-case scenarios of AI use in learning and education, while accounting for the ethical issues arising from the implementation. Several organizations and interest groups are already working on these guidelines.
Going forward, while it’s difficult to achieve meaningful and sustainable progress towards SDG 4 with the help of AI without addressing some of the structural issues and gaps mentioned previously, the technology could help optimize the delivery of solutions and resources that would yield measurable progress. Examples include:
- AI teaching and tutoring. There’s no technological substitution for human connection. However, AI can optimize the time of professionals like teachers, TAs, tutors, and other support and orientation roles. They all could leverage AI to increase their availability for quality time with learners.
- Personalized, autonomous learning. New frontiers in personalized learning may arise via use of AI tools that optimize the learning experience towards the learners’ preferences and personality, without a loss of quality and challenge. Beware, however, blind optimism about products heading in the right direction: AI is not a magic cloak protecting commercial products from utter business failure.
- Assessment practice. AIs can provide mock examination content and other experiences aimed at improving students’ performance prospects in standardized third-party administered testing.
- Global access to talent pools. Human capital magnets like Silicon Valley are few and far between, concentrating talent and raising the risk of “talent gentrification” and “brain drain” among developing nations. AI —likely coupled with open source eLearning solutions— could become the basis for solutions that make world-class talent, worldly available.
- Increased accessibility. Artificial General Intelligence (AGI), the purported stage in which AI achieves complete autonomy, won’t be with us any time soon. But today’s AI and LLM tools can provide some measure of auto-correction by becoming compliant with accessibility standards or adaptive technologies.
- General process optimization. Strengthening and optimizing processes, understanding and generating insights, and making data analysis more approachable are things AI is already helping with, including on global issues like climate change.
The stream of AI tools in education and learning is non-stop. Get exclusive access to our AI in Learning selection of tools —and popularity tracker—, free with your ticket for our AI in Learning virtual summit (also free).
Experts highlight the importance of individuals taking steps to become better acquainted with both the promise and peril of AI in Learning. Joti Balani, CEO of consultancy Freshriver.ai, suggests a three-pronged framework to advance positive, ethical use of AI both in your classroom and globally:
- Educate yourself. Understanding AI, or better yet, being part of the communities trying to make sense of it, is half the battle. Study the tools, put them into practice, and share your experiences and outcomes as freely as possible.
- Volunteer. Be part of experiments aimed at evaluating the impact of AI in global education. Luckily enough, there appears to be a consistent positive correlation between top AI talent and interest for philanthropic causes, which means that your involvement in SDG-friendly initiatives could benefit you professionally down the line. (Should that matter to you, of course.)
- Engage. Be it in the political and legislative sphere, within the global aid and development ecosystem, or through independent networks, help advance discussions of AI in Learning where it’s most needed. Help make sure the centers of development of AI technologies move in a globally beneficial direction that’s decentralized and inclusive.
Check out Joti among 20+ experts discussing tools, challenges and the future of AI in learning in our upcoming virtual summit of the eLearn Success Series: AI in Learning
To sum up—Here’s how you can be part of the AI-based solution
Despite these challenges and limitations, there’s reason for optimism when it comes to the role of AI in achieving SDG 4. It’s no secret that collective progress towards Sustainable Development Goals has been both subpar and regressing due to both global (COVID-19) and regional (conflict and war, or the localized impact of climate change) crises. However, overall there is generalized optimism among education and eLearning professionals, regarding the role and positive outcomes of AI, particularly LLMs, being introduced in the classroom.
As technology continues to improve, we can expect to see more sophisticated and effective AI-powered solutions benefitting more and more people. Awareness, open debate on several fronts (technical, ethical, and economic, just to name a few) and democratization of ownership are needed if we’re to collectively move forward in terms of the SDG targets.
Overall, progress on SDG 4 at the regional and national level will benefit from AI in proportion to available infrastructure, wider access to talent and healthy institutions. Embracing AI solutions integrally can lead to powerful, personalized learning experiences, and there’s the potential for substantial economic and development progress thanks to better coordination around filling out skill gaps. The debate is ongoing. Join our virtual event on AI in Learning and learn how AI can help close global skills gaps, while providing high-quality learning experiences to all students.