Artificial Intelligence – Core Concepts, Project Cycle & Ethics

Vocabulary flashcards covering definitions and key concepts from AI introduction, domains, project cycle, evaluation metrics and ethics for Grade 9 Artificial Intelligence.

Artificial Intelligence (AI)

The ability of machines to simulate or enhance human intelligence, including reasoning, learning from experience, perception and decision-making.

Core Abilities of AI

Perception (acquiring information), Reasoning (using rules to reach conclusions) and Learning (self-correction and improvement).

Big Data

Extremely large, high-volume datasets that AI systems analyse to discover patterns, trends and make predictions.

Data (AI Domain)

Raw facts—numbers, letters or symbols—used to train or test an AI model; the fundamental base of all AI work.

Natural Language Processing (NLP)

AI domain that enables computers to understand, interpret and generate human language.

Natural Language Understanding (NLU)

NLP component that converts language input into machine-usable representations and analyses its meaning.

Natural Language Generation (NLG)

NLP component that produces meaningful phrases or sentences from machine data.

Lexical Analysis

NLP step that identifies and analyses word structures, creating a list of tokens (words and phrases).

Syntactic Analysis (Parsing)

NLP step that arranges words using grammar rules to reveal relationships within a sentence.

Semantic Analysis

NLP step that maps syntactic structures to a task domain to ensure the sentence makes sense.

Discourse Integration

NLP step that interprets meaning based on information from preceding sentences.

Pragmatic Analysis

NLP step that derives intended meaning by applying real-world knowledge and context.

Computer Vision (CV)

AI domain that enables computers to ‘see’, interpret and understand the content of digital images and videos.

Pattern Recognition

Technique (often used in CV) that detects regularities in data or images to classify or interpret them.

AI Project Cycle

Structured workflow for building AI solutions: Problem Scoping, Data Acquisition, Data Exploration, Modelling, Evaluation and Deployment.

Problem Scoping

First stage of the AI Project Cycle that defines the challenge the AI must solve and frames it for machines.

4Ws Problem Canvas

Tool for problem scoping—Who (stakeholders), What (problem), Where (location) and Why (value of solution).

Data Acquisition

Process of collecting comprehensive, accurate data from surveys, sensors, web scraping, APIs, etc., for an AI project.

System Map

Visual diagram showing relationships among elements in a scoped problem, highlighting data interdependencies.

Data Exploration

Stage where collected data is inspected for patterns, trends and insights before modelling.

Data Visualization

Graphical representation of data (e.g., bar charts, scatter plots) to aid understanding during exploration.

Modelling (AI)

Stage of selecting strategies and algorithms to build an AI model that proposes solutions.

Rule-based Approach

AI model that relies on explicit ‘if-then’ rules; no learning from new data.

Machine Learning (ML)

Learning-based approach where algorithms improve performance on a task through experience with data.

Deep Learning (DL)

Subset of ML using multi-layered neural networks to learn from vast amounts of data automatically.

Evaluation (AI Model)

Testing a model’s predictions against actual outcomes to measure performance and completeness.

Scenario (Evaluation)

Real-world problem area where the AI model is applied and evaluated.

Prediction

Output generated by an AI model for a given input.

Reality

Actual observed outcome used to validate an AI model’s prediction.

True Positive (TP)

Case where the model correctly predicts a positive outcome (e.g., predicts an earthquake and one occurs).

True Negative (TN)

Case where the model correctly predicts a negative outcome (e.g., predicts no earthquake and none occurs).

False Positive (FP)

Case where the model incorrectly predicts a positive outcome; also called Type I error.

False Negative (FN)

Case where the model incorrectly predicts a negative outcome; also called Type II error.

Confusion Matrix

Table summarising TP, FP, TN and FN counts to visualise a classification model’s performance.

Deployment (AI)

Final stage of the AI Project Cycle where the validated model is integrated into real-world systems for use.

AI Ethics

Field ensuring AI development aligns with principles such as fairness, transparency, accountability and human welfare.

Human Rights (AI Ethics)

Principle that AI must respect freedoms, equality and dignity without infringing on people’s fundamental rights.

Bias (AI Ethics)

Discriminatory outcome produced by an algorithm due to skewed data or programmer assumptions.

Privacy (AI Ethics)

Requirement that AI systems minimise personal data collection, secure stored information and give users control.

Inclusion (AI Ethics)

Ensuring AI systems are accessible and fair to all groups, avoiding exclusion or disadvantage.

AI Bias

Manifestation of unfairness in AI results caused by biased training data or embedded programmer preferences.

AI Access

Disparity where only those who can afford AI-enabled tools benefit, creating a digital divide.

Advantages of AI (Boon)

Reduced human error, tireless repetition, rational decisions, speed and applicability across diverse sectors.

Disadvantages of AI (Bane)

High cost, lack of true creativity or emotion, limited experiential learning, privacy concerns and legal risks.