AI Foundations and Literacy

 AI Foundations and Literacy

Mary Lourdes Bonnici MBA | Learning • Knowledge • Growth | 30 June 2026

Introduction

Artificial intelligence (AI) increasingly shapes how people search, study, create, communicate and make decisions. AI literacy is the ability to understand what these tools do, evaluate their claims and use them responsibly. This article develops the fundamental concepts, practical skills and ethical judgement needed for informed participation in an AI-enabled world.

1. Understanding Artificial Intelligence

AI refers to computational systems capable of performing tasks associated with human intelligence, including classification, prediction, language processing and content generation. It does not necessarily mean that a machine thinks or understands as humans do.

Machine learning (ML) is a subset of AI that learns statistical patterns from examples. Deep learning is a subset of ML based on multilayer neural networks. Conventional automation follows programmed rules; learning systems adapt patterns from data. Most deployed AI remains specialised rather than universally capable.

2. How AI Developed

Alan Turing explored machine intelligence in 1950. The 1956 Dartmouth proposal helped establish AI as a research discipline. Expert systems, statistical learning and improvements in computing later advanced the field. The transformer architecture described by Vaswani and colleagues in 2017 enabled major advances in language processing. Generative AI became widely accessible during the 2020s. Progress is substantial but varies across tasks.

3. How Machine Learning Works

A typical workflow defines the problem, gathers suitable data, cleans and labels examples where necessary, trains a model, validates its settings and evaluates performance on held-out data. Once deployed, the model performs inference: generating predictions or outputs from new inputs. Monitoring is needed because conditions and data can change.

Supervised learning uses labelled examples; unsupervised learning detects patterns without provided labels; reinforcement learning learns from rewards and feedback; self-supervised learning creates training signals from data itself.

Overfitting occurs when a model learns training examples too narrowly and performs poorly on new cases. Data leakage, unrepresentative samples and changing real-world conditions can also undermine results.

4. Generative AI and Large Language Models

Generative AI creates text, images, audio, video or code based on learned patterns. Large language models commonly process text as tokens and generate likely continuations conditioned on the prompt and context. Their fluent language can be helpful without guaranteeing factual correctness.

Hallucinations are outputs that are inaccurate or unsupported, including invented sources and misleading numbers. Models may also reflect bias, misunderstand instructions or omit important context. Connecting a model to search or documents can help but does not eliminate these risks.

5. The Essential Skills of AI Literacy

Recognition: identify when an AI system may be shaping content or recommendations.

Understanding: distinguish prediction, generation, automation and evidence-based reasoning.

Prompting: communicate the objective, audience, constraints and expected format.

Verification: examine claims, sources, calculations, dates and uncertainty.

Privacy: minimise personal or confidential information shared with tools.

Agency: preserve meaningful human judgement and responsibility.

6. Effective Prompting and Fact-Checking

A useful prompt includes purpose, audience, relevant context, constraints, desired format and quality criteria. For example: “Explain the greenhouse effect to a beginner in 250 words, distinguish weather from climate, include two common misconceptions, and identify reliable scientific sources for checking.”

Verification routine: identify consequential claims; locate primary evidence; confirm the date and context; compare with independent trustworthy sources; and correct or remove unsupported claims. Always recalculate important figures independently. A polished answer is not proof of accuracy.

7. Ethics, Bias and Privacy

AI systems can amplify unfair patterns present in training data or design choices. Fairness must be tested in the actual context of use. Transparency matters when people are affected by automated decisions. Privacy demands careful treatment of personal information, informed choices about data sharing and appropriate security.

AI can also facilitate impersonation, phishing and synthetic misinformation. Copyright and ownership issues depend on jurisdiction, source material and the use of generated content. Human accountability remains essential. UNESCO, OECD and NIST provide widely used principles and frameworks for responsible AI.

8. Everyday Applications

Education: generate practice questions, simplify difficult explanations and support language learning, while independently checking facts.

Accessibility: captions, transcription and text simplification can assist users, although errors remain possible.

Creativity: brainstorm concepts and draft outlines while preserving originality and checking rights.

Personal productivity: organise notes and plan tasks without disclosing unnecessary private information.

High-stakes decisions: health, legal, safety and financial matters require appropriate expertise and authoritative evidence rather than reliance on AI alone.

9. Governance and the Future

AI governance combines risk assessment, documentation, technical evaluation, monitoring, accountability and routes for correcting harmful outcomes. The EU AI Act establishes a risk-based legal framework with phased implementation; readers should consult the current official text and guidance for applicable duties. AI literacy requires continuous learning as tools, standards and regulations evolve.

10. Detailed Case Study: Preparing a Public Learning Guide

Scenario: An educator asks a generative AI tool to draft a public guide about misinformation. The draft looks professional but contains an invented journal article, an old statistic and a potentially stigmatising example.

Actions: She checks every citation with publishers and institutions, removes the fabricated article, replaces or omits the outdated statistic, revises the example, and adds a clear explanation of how claims were verified. She keeps responsibility for the final publication and discloses AI assistance where appropriate.

Outcome: The revised guide becomes more trustworthy because human review and evidence checking were integral to the workflow.

Lesson: AI can speed up drafting, but reliable educational work still depends on independent research, ethical awareness and accountable editorial decisions.

11. Reflection Questions and Answers

1. How is machine learning related to AI? AI is the broad field; machine learning is one family of methods that learn from data.

2. Why can AI sound confident while being wrong? Fluent generation is not the same as checking facts against evidence.

3. What makes a prompt useful? A clear goal, context, audience, constraints and requested output improve relevance.

4. How should AI-generated statistics be handled? Check original sources, dates, definitions and calculations before using them.

5. Why does bias matter? Systematic errors may unfairly affect particular groups and must be evaluated in context.

6. What is responsible data sharing? Provide only necessary information, protect confidential details and review service policies.

7. Should AI make consequential decisions alone? Human oversight, professional expertise and avenues for correction are important.

8. Does AI replace the need to learn? No; subject knowledge enables better questions, error detection and informed judgement.

12. Final Lesson and Reflection

AI literacy is not merely knowing which buttons to press. It is understanding capabilities, limitations and consequences. Before relying on an AI output, ask: What evidence supports this? What could be missing? Who may be affected? Who is responsible for the decision?

13. Conclusion

Artificial intelligence offers substantial opportunities for learning and creativity, but it also introduces risks involving accuracy, fairness, privacy and accountability. Understanding AI foundations and developing habits of verification allow people to benefit from these tools while preserving independent thought. The goal is not blind acceptance or blanket rejection: it is informed, responsible and critically engaged use.

14. Harvard-Style References and Useful Links

Turing, A.M. (1950) “Computing Machinery and Intelligence”, Mind, 59(236), pp. 433–460. https://doi.org/10.1093/mind/LIX.236.433

Vaswani, A. et al. (2017) “Attention Is All You Need”, Advances in Neural Information Processing Systems, 30. https://arxiv.org/abs/1706.03762

UNESCO (2021) Recommendation on the Ethics of Artificial Intelligence. https://www.unesco.org/en/artificial-intelligence/recommendation-ethics

OECD (2019, updated 2024) OECD AI Principles. https://oecd.ai/en/ai-principles

NIST (2023) Artificial Intelligence Risk Management Framework (AI RMF 1.0). https://doi.org/10.6028/NIST.AI.100-1

European Union (2024) Regulation (EU) 2024/1689 (AI Act). https://eur-lex.europa.eu/eli/reg/2024/1689/oj

Official Closing and Copyright

Thank you for reading. Continue exploring, questioning and learning.

© 2026 Mary Lourdes Bonnici MBA. All rights reserved. Learning • Knowledge • Growth. This original educational content is provided for learning and information. Reproduction or republication requires prior written permission, except where permitted by applicable law.