Definition
The Delhi Metro map shows straight lines, neat angles and evenly spaced stations. The real tracks curve and stations are far apart or close together. Yet the simplified map is exactly what a passenger needs.
Abstraction means focusing on the details that matter for a problem and ignoring the ones that do not. It produces a simpler model of a situation that is easier to think about and solve, while keeping everything needed for the task.
In simple words
When you describe your house to a friend coming over, you say "blue gate, second building after the temple, third floor". You do not mention the number of bricks or the colour of the curtains. You kept only what helps them find it. That is abstraction.
Why it matters
Real situations contain endless detail. Without abstraction, every problem would be too complicated to solve. Every AI model is an abstraction: it keeps some features of the world and ignores the rest, and choosing what to keep decides how well it works and who it works for.
Key terms
- Abstraction
- Keeping the details that matter and removing the rest.
- Model
- A simplified representation of something real, made for a purpose.
- Relevant detail
- Information that affects the answer to the problem.
- Irrelevant detail
- Information that does not change the answer and can be left out.
- Generalisation
- Describing many specific cases with one general idea, such as "a vehicle" for cars, buses and bikes.
- Level of detail
- How much is kept in a model. Different tasks need different levels.
Explained in detail
Maps: the classic abstraction
The London Underground map, designed by Harry Beck in 1933, became famous for throwing away geography. It ignored real distances and curves and kept only what passengers needed: which lines exist, which stations they stop at, and where you can change. Metro maps in Delhi, Mumbai, Bengaluru and Kolkata follow the same idea. A road map for drivers keeps different details: road widths, one-way streets, petrol pumps. A map for trekkers keeps contour lines and water sources. Each map is an abstraction designed for one purpose.
Choosing what to keep
| Task | Keep | Ignore |
|---|---|---|
| Planning a school bus route | Students' stops, road connections, travel times | House colours, students' favourite subjects |
| Predicting if a mango is ripe | Colour, softness, smell | Which tree it grew on, the shop's name |
| A game of chess on a computer | Positions of the pieces, whose turn it is | The wood of the board, the shape of the knight |
Abstraction in AI
When people design an AI system, they decide which pieces of information, called features, the model will use. A model to predict crop yield might use rainfall, soil type, seed variety and fertiliser, and ignore the farmer's name. These choices are abstractions, and they can go wrong in two ways. Leave out something important, such as irrigation, and predictions will be poor. Keep something that should not matter, such as a person's religion or caste in a loan decision, and the model may become unfair.
Too much or too little
Too little abstraction leaves you drowning in detail; too much leaves out what matters. A weather forecast that says only "India will have weather tomorrow" is abstract but useless. One that lists the temperature of every street is detailed but unreadable. The right level depends on who will use it and what they need to decide. A good habit is to ask: if I remove this detail, would the answer change? If not, it can go.
Abstraction also lets people work together. A driver does not need to understand how the engine burns fuel; they only need the steering wheel, pedals and gears. The engineer who designed the engine does not need to know where the driver is going. Each person works with a simpler view of the system, and the details are hidden behind a clear interface.
How it works
- State the task clearly: what decision or answer is needed?
- List all the details you know about the situation.
- For each detail, ask whether removing it would change the answer.
- Keep the ones that matter; set the rest aside.
- Check your simplified model still works on a real example.
Try it yourself
Drawing a map for a new student
Aarav has to draw a map to help a new student walk from the gate to the science lab.
Keep or ignore? Predicting if a train will be late
Which details matter for this prediction?
Fog on the route
Colour of the engine
How late the train was at the previous station
Track repair work ahead
The driver's favourite song
Common mistakes
- Keeping every detail "just in case".
- Throwing away a detail that actually changes the answer.
- Using one abstraction for every purpose.
Check yourself
Summary
- Abstraction keeps the details that matter and removes the rest.
- Every map and every AI model is an abstraction made for a purpose.
- Ask whether removing a detail would change the answer.
Finished reading?
Sources: BBC Bitesize: Introduction to computational thinking · CS Unplugged