7.1.4 Artificial Intelligence: Applications and Wider Impact
Artificial intelligence is used when a computer system performs a task that would normally require human judgement, recognition, prediction or decision-making. This section focuses on what AI systems do, where they are used, and how their use can affect people, organisations and the environment.
By the end of this section, you should be able to:
- Describe artificial intelligence without assuming that it has human consciousness or understanding.
- Explain several practical applications of AI using appropriate technical vocabulary.
- Describe how data, a model or rule base, and an output are connected in an AI system.
- Evaluate social, economic and environmental consequences of an AI application.
- Give a balanced judgement that identifies stakeholders, benefits, risks and possible safeguards.
What Makes a System Artificially Intelligent?
A conventional program follows instructions written directly by a programmer. An AI system may instead use stored knowledge, patterns learned from data, or a search process to classify, predict, recommend or control an action.
Artificial intelligence: computer-based techniques used to carry out tasks
associated with human intelligence, such as recognising patterns, understanding language,
making predictions or selecting actions.
Calling a system “AI” does not mean that it thinks or understands exactly like a person. A system may perform one specialised task extremely well while being unable to deal with situations outside the conditions for which it was designed.
| Type of system | How the response is produced | Example |
|---|---|---|
| Fixed-rule program | Follows explicitly programmed conditions. | A calculator applies a defined arithmetic operation. |
| Rule-based AI | Uses a knowledge base and rules to reach a conclusion. | A maintenance assistant suggests likely causes of a machine fault. |
| Machine-learning system | Uses patterns learned from training data. | A camera system classifies fruit as damaged or undamaged. |
Common misconception: A system is not intelligent simply because it is automatic.
A timer that switches lights on at 18:00 is automated, but it is not making a prediction or adapting
its behaviour from data.
A Simplified AI Process
AI systems can be complex, but many can be understood through four broad stages.
| Stage | Purpose | Example in a crop-monitoring system |
|---|---|---|
| 1. Input | Collect data about the situation. | Images of leaves are captured by a field camera. |
| 2. Processing | Apply rules or a trained model to the input. | The model searches for visual patterns associated with disease. |
| 3. Output | Produce a classification, prediction, recommendation or action. | The system labels the plant as “possible infection”. |
| 4. Review or feedback | Check the result and improve future use. | An agronomist confirms the result before treatment is applied. |
Exam tip: When explaining an AI application, describe the data it receives,
the task it performs and the output it produces. This is more precise than saying only
that “the AI makes a decision”.
Applications of AI
The syllabus does not require one fixed list of applications. Students should be able to recognise AI in unfamiliar contexts and explain the task being performed.
| Application area | AI task | Original example |
|---|---|---|
| Computer vision | Recognise objects or patterns in images or video. | A recycling plant identifies glass, metal and plastic items on a conveyor belt. |
| Natural language processing | Interpret, generate or translate human language. | A school platform produces captions and translates short classroom instructions. |
| Prediction | Estimate a future value or event from previous data. | A delivery company predicts which vehicles are likely to need maintenance. |
| Recommendation | Rank possible choices for a user. | A learning platform suggests revision questions based on earlier errors. |
| Autonomous control | Select actions using sensor data. | A warehouse robot changes route when an aisle is blocked. |
| Decision support | Use rules or patterns to assist a human decision. | An energy engineer receives likely explanations for unusual power usage. |
Classification: placing an input into a category.
Prediction: estimating an unknown or future outcome.
Recommendation: ranking or suggesting possible choices.
Prediction: estimating an unknown or future outcome.
Recommendation: ranking or suggesting possible choices.
Limits, Risk and Human Oversight
The quality of an AI result depends on the system design, the data, the conditions in which it is used and the way people respond to its output.
| Issue | Why it matters | Possible safeguard |
|---|---|---|
| Unrepresentative training data | The system may work less accurately for situations missing from the data. | Test with varied data and monitor results for different user groups. |
| Incorrect output | A false positive or false negative may cause harm or unnecessary action. | Use confidence thresholds and human review for high-risk decisions. |
| Lack of transparency | Users may not understand why a result was produced. | Record relevant evidence and provide an explanation where possible. |
| Automation bias | People may trust the system even when other evidence suggests it is wrong. | Train users to challenge outputs and keep final responsibility clear. |
| Changing conditions | A model trained on older data may become less reliable. | Review performance and retrain or update the system when needed. |
Common mistake: “The AI is unbiased” is not a safe assumption.
Bias can enter through the data collected, the categories chosen, the objective used
to train the system or the way its output is applied.
Economic Impact
Economic effects concern costs, productivity, employment, skills and the distribution of financial benefits.
| Possible benefit | Possible concern |
|---|---|
| Repetitive work can be completed more quickly. | Some existing roles may shrink or change. |
| Prediction can reduce waste, delays and equipment downtime. | Development, training, specialist staff and computing infrastructure can be expensive. |
| New products and specialist occupations may be created. | Workers may need retraining before they can move into new roles. |
| Smaller organisations may access advanced services through shared platforms. | Dependence on a small number of powerful providers may reduce competition or control. |
Exam tip: Do not reduce the economic impact to “AI removes jobs”.
Explain which tasks may be automated, which new tasks may appear and why retraining or
investment may be needed.
Environmental Impact
AI can support environmental goals, but it also requires physical hardware, electricity, cooling and replacement equipment.
| Possible benefit | Possible concern |
|---|---|
| Smart irrigation can reduce unnecessary water use. | Training and operating large models can require substantial electricity. |
| Route optimisation can reduce distance travelled and fuel consumed. | Data centres require cooling as well as electrical power. |
| Predictive maintenance can extend the useful life of equipment. | Specialist processors require raw materials and eventually become electronic waste. |
| Computer vision can support recycling and environmental monitoring. | Frequent replacement of devices can increase manufacturing and disposal impacts. |
A balanced environmental judgement should consider both the resources used by the AI system
and any resources saved because of the decisions it supports.
Interactive: AI Application and Impact Explorer
Choose an application and then select an impact lens. Follow the sequence from input to output before evaluating who benefits, who may be harmed and what safeguard could help.
Worked Example: AI for School Transport Planning
A school group is considering an AI system that predicts demand for its afternoon buses. The system uses attendance records, activity sign-ups and earlier passenger counts.
| Evaluation step | Reasoned response |
|---|---|
| Application | The system predicts how many seats may be needed on each route. |
| Social effect | Students may experience shorter waits, but incorrect predictions could leave some without suitable transport. |
| Economic effect | Better planning may reduce the number of nearly empty journeys, although the system has development and maintenance costs. |
| Environmental effect | Fewer unnecessary journeys could reduce fuel use, but the computing service also consumes energy. |
| Safeguard | A transport coordinator should review the prediction and keep spare capacity for unexpected demand. |
| Judgement | The system could be useful as decision support, but it should not automatically cancel a route without human review. |
Useful structure:
application → stakeholder → benefit → concern → safeguard → judgement.
Common Mistakes and Misconceptions
- Assuming every automated system uses AI.
- Describing an application without stating its input, task or output.
- Saying an AI result is always correct because it was produced by a computer.
- Discussing privacy as the only social concern and ignoring fairness, safety, accountability or accessibility.
- Writing only that AI “creates jobs” or “removes jobs” without explaining which tasks and skills are affected.
- Giving only negative environmental effects and ignoring uses that may reduce waste or resource consumption.
- Listing advantages and disadvantages without linking them to the named application.
Practice
Core questions
- Describe artificial intelligence in a way that does not imply human consciousness.
- Explain the difference between classification, prediction and recommendation.
- Describe the four broad stages in the simplified AI process.
- Explain why a machine-learning system may become less reliable when conditions change.
- Give one reason why human oversight may be needed in a high-risk AI application.
Application questions
- A camera system identifies empty parking spaces. Describe its likely input, processing task and output.
- An AI tool ranks job applications. Explain one social benefit, one social risk and one safeguard.
- A supermarket predicts demand for fresh food. Explain one economic effect and one environmental effect.
- A factory replaces some manual inspection with computer vision. Explain how employment may change rather than simply disappear.
- Evaluate the use of an AI assistant that suggests revision activities to students. Your answer should include social, economic and environmental considerations.
Review
| Question | Strong answer should include |
|---|---|
| What is AI? | Computer-based techniques for tasks such as recognition, prediction, language processing or decision support. |
| What should be explained for an AI application? | The input data, AI task, output and how a person or device uses the result. |
| Why can AI outputs be unreliable? | Data may be incomplete or unrepresentative, conditions may change, or the model may make classification or prediction errors. |
| Which wider impacts are required? | Social, economic and environmental effects. |
| What makes an evaluation balanced? | A named stakeholder, a benefit, a concern, a safeguard and a justified conclusion. |
Final exam tip: Avoid generic statements such as “AI is faster”.
State what becomes faster, who benefits, what risk remains and how that risk could be managed.
Social Impact
Social effects concern people, communities and the way decisions are made. The same application can create both benefits and risks.