AI Literacy for High School Students: Why Systems Thinking and Critical Thinking Matter (Part I)

What is AI, why does it matter, and what promise does it offer when approached critically?

Image courtesy of Canva

Artificial intelligence is often narrowly understood, leaving its full potential and far reaching implications barely explored. AI is generally seen as a tool, something we can either embrace or reject, use or avoid, celebrate or fear. However, AI is better understood as a system, involving multiple components for its functionalities. It is made up of technologies, data, people, institutions, incentives, infrastructures, regulations, markets, cultures and environmental costs. In viewing AI from a system perspective, we go beyond asking questions such as, “What can this tool do?” to interrogating: What does it change? Who benefits? Who is excluded? What behaviours does it encourage? What new risks does it create? What becomes easier, and what becomes harder to trust? The system perspective matters as AI has far reaching implications. It is shaping classrooms, workplaces, public services, financial systems, creative industries, healthcare, research, development work and everyday communication. Its consequences are not only technical but are social, ethical, economic and ecological.

This is why, at the Rondil Scholars Pre-University Program, we teach students to approach emerging technologies, including AI, with a critical and interdisciplinary mindset, combining futures thinking, innovation, real-world problem-solving and learner agency. Our goal is not simply to encourage tool use, but to help students understand how the technologies work, where their limitations lie and how they can be used responsibly, ethically and thoughtfully within the wider systems they shape.

This first article in our three-part series focuses on the promise of AI. It explores how AI can support learning, research, creativity, discovery and problem-solving when used thoughtfully. The second article will turn to the risks and consequences of AI, including misinformation, bias, exclusion, environmental costs, cognitive dependency and the responsibilities that come with using these technologies. The last part will introduce how Rondil Scholars is developing a framework that supports shaping scholar career pathways with the ethical use of AI.

**This article is not only about generative AI, although it discusses generative AI explicitly where relevant. It considers AI more broadly, including machine learning, deep learning, natural language processing, computer vision, automation and other AI-enabled systems that are already shaping education, work, research, public services and everyday life.**

What is AI?

“There is no single universally agreed definition of artificial intelligence” as noted by The UK Parliament, and that this lack of one precise definition has allowed AI to be adapted across many contexts, but it also creates challenges for regulation and public understanding.

In simple terms, AI describes technologies that enable computers and machines to carry out tasks we usually associate with human intelligence. Across different definitions, this includes the ability to learn, understand information, solve problems, make decisions, recognize patterns, respond to new inputs, analyze large amounts of data, and in some cases generate creative outputs. IBM Google Cloud SAS

A brief history helps to put today’s excitement in perspective. The idea of artificial beings or self-moving machines is ancient, but modern AI grew out of twentieth-century computing. Alan Turing’s 1950 paper “Computing Machinery and Intelligence” proposed the famous imitation game, now known as the Turing Test, as a way of thinking about machine intelligence. The term “artificial intelligence” was coined in the 1950s, and early research explored symbolic reasoning, problem solving and neural networks. Tableau Since then, AI has moved through cycles of optimism, disappointment and renewed progress. Recent advances have been driven by increased data volumes, improved algorithms, cheaper computing power, cloud infrastructure, machine learning, deep learning and, most visibly, generative AI. SAS

AI includes several overlapping fields. Machine learning allows systems to learn from data and make predictions or identify patterns without being explicitly programmed for every situation. Deep learning uses neural networks with many layers to process complex data such as images, speech and text. Natural language processing enables machines to work with human language. Computer vision enables systems to interpret images and video. Robotics process automation uses software to automate structured, repetitive tasks. FINRA Generative AI, the type currently dominating public conversation, can produce text, images, audio, video, code or other media in response to prompts. IBM

It is also important to separate the AI we use today from the idea of artificial general intelligence. Most current AI is narrow AI, meaning it is built to perform tasks such as translating text, recognizing images, detecting fraud, or generating written content. Artificial general intelligence, often called AGI, would be much broader: a system able to learn and apply knowledge across many different tasks with something closer to human flexibility i.e., adaptable to unfamiliar situations beyond a narrow set of functions. For now, AGI remains a theoretical possibility rather than a technology that already exists. Google Cloud

Why systems thinking matters

At Rondil Scholars, we teach our scholars to approach AI through systems thinking, encouraging them to examine not only the immediate outputs the tool produces, but also the assumptions, limitations and wider consequences behind them. This approach is central to AI literacy for students, equipping them to critically evaluate AI tools rather than simply use them. For example, a chatbot may help a student draft an essay, a manager summarize reports, or a researcher explore literature faster. But the system around that tool includes most important the data it was trained on, the energy used to run it, the company that owns it, the rules governing its use, the assumptions embedded in its outputs, the user’s level of expertise, and the consequences for trust, assessment and decision-making.

A systems thinking approach to AI. Image courtesy of Canva.

In other words, AI is not simply a productivity tool. It is an intervention in a system, and interventions create feedback loops.

If AI makes writing faster, more people may (or will) produce more content. That increases the amount of mediocre work within digital spaces, making trustworthy sources harder to identify. If AI makes cheating easier, institutions may respond with stringent detection tools, which reduces trust between students and educators. If AI tools become essential for work, the inequalities further widen as people without reliable internet, high-quality devices or are fluent in English may fall further behind. If AI saves time in one sector, it may increase energy demand elsewhere. Systems thinking helps us hold these tensions together.

The promise: acceleration, creativity and discovery

AI offers tangible benefits. It can take on repetitive tasks, spot patterns across large datasets, support better decision-making, reduce certain kinds of human error and keep services running continuously. It can also help people find insights more quickly, automate high-volume digital work and support discovery through data. However, these systems still depend on humans to design them, guide their use, interpret their outputs and ask the right questions. SAS IBM

AI is used across a wide variety of sectors including agriculture, education, engineering, finance, transport, healthcare, justice and manufacturing. In research and innovation, AI can help scan large bodies of literature, detect patterns, generate hypotheses, discover new materials, support modelling, assist with translation, and help researchers communicate complex findings to wider audiences. In healthcare, AI can support medical imaging, risk prediction, personalized treatment and drug discovery. Cutting across health and innovation, Novartis is using AI for target identification, molecule design and safety prediction. In one kidney disease program, it narrowed thousands of genes to five promising targets in under a year, while another project computationally screened 15 million potential compounds and reduced laboratory testing to about 60 molecules. These approaches can replace parts of a process that traditionally takes several years, helping researchers identify viable drug candidates faster while reducing failures, costs and reliance on animal studies.

A US FDA-approved AI tool now autonomously screens retinal photographs for diabetic retinopathy by detecting microscopic eye lesions with up to 93% accuracy. This allows primary care clinics to instantly identify at risk patients immediately and prevent blindness without needing an on-site ophthalmologist. Diabetes in Control The UK’s NHS also uses AI-powered prediction software drawing on routinely collected hospital data to identify patients who require immediate preventative support to reduce the likelihood of future, unplanned hospital visits.  

In agriculture, AI can help monitor crops, forecast prices and improve resource efficiency. For instance, the agricultural equipment company, John Deere, uses its See & Spray technology which combines boom mounted cameras and AI processors to scan fields and distinguish weeds from crops in real-time, reducing chemical herbicide usage by up to two thirds. The technology has reportedly saved farmers an estimated 8 million gallons of herbicide (equating to savings of up to 59%) with translated economic savings of $15.7 per acre. In education, AI can assist with lesson planning, marking, scheduling, with tools such as Gradescope, providing consistent and objective grading of assignment and optimizing class timetables and resource allocation. AI can also further support responses to learner queries and help identify individual learning needs. Platforms such as Carnergie Learning, provide personalized feedback and support, adapting to individual learning styles and needs to help students understand complex concepts and improve academic performance.

Used prudently, AI can become part of a learning journey rather than a shortcut around it. It can help participants brainstorm, compare perspectives, simplify complex concepts, practise writing, translate ideas across audiences, analyze datasets, prepare presentations, test assumptions and reflect on their own reasoning. For students and professionals, AI can should be a thinking partner, not a replacement for thinking. The phrase “used prudently” is doing important work here. AI is most valuable when learners understand both its capabilities and its limits. A program that teaches AI use should not only show people which tools to use but it should help them ask better questions, evaluate outputs, check sources, recognize bias (all part of the art of “prompting”), protect data, disclose use appropriately and understand when human judgement is essential. These considerations vary by sector, as public health practitioners, lawyers, teachers, social entrepreneurs, finance professionals, journalists, engineers and community organizers face different use cases, risks and responsibilities.

AI presents significant promise to extend human capacity: helping people explore ideas and solve complex problems more efficiently. But these benefits do not exist in isolation. Every AI tool sits within a wider system of data, design choices, access, incentives and consequences. In the next article, we will examine the other side of that system: the risks, unintended consequences and responsibilities that emerge when AI becomes part of how we learn, work and make decisions.

Written by Dr Sandile Mtetwa-Omuthe, Dr Manuel Kernen and Dr Ronald Omuthe