Skip to content
Oak National Academy
Oak National Academy
View my library
0
Teachers
Pupils
View unit
New
New
Lesson 5 of 8
  • Year 10

Bias and accuracy in machine learning

I can describe the impact of data on ML models and explain bias in ML model predictions.

Download all resources
Share activities with pupils
View unit
Lesson 5 of 8
New
New
  • Year 10

Bias and accuracy in machine learning

I can describe the impact of data on ML models and explain bias in ML model predictions.

Download all resources
Share activities with pupils

These resources will be removed by end of Summer Term 2025.

Switch to our new teaching resources now - designed by teachers and leading subject experts, and tested in classrooms.

These resources were created for remote use during the pandemic and are not designed for classroom teaching.

View new resources

Lesson details

Key learning points

  1. The type, quality and amount of data used significantly affect how accurate a machine learning (ML) model is.
  2. ML models require both training data and separate test data to ensure reliability.
  3. Bias is introduced into ML models when the data is unrepresentative or contains stereotypes.
  4. Use large, representative data sets and diverse perspectives to reduce bias and improve ML model fairness and accuracy.

Keywords

  • Fair - when something is free from bias and gives equal consideration and treatment to all parts of a group or situation

  • Unrepresentative - when the data does not properly reflect the whole group or situation it is meant to describe

  • Diverse - when data includes a wide range of examples that fairly represent different parts of the whole group or situation

Common misconception

ML models are neutral and always give fair or accurate information.

ML models can reflect or amplify bias in their training data. If the data contains stereotypes or imbalances, the ML model can repeat or reinforce them.


To help you plan your year 10 computing lesson on: Bias and accuracy in machine learning, download all teaching resources for free and adapt to suit your pupils' needs...

Use everyday examples students know, like music playlists or sports scores, to show how data affects results. Give them a quick activity with incomplete or uneven data so they can see for themselves how it can lead to unfair or wrong outcomes.
Teacher tip

Equipment

Licence

This content is © Oak National Academy Limited (2025), licensed on Open Government Licence version 3.0 except where otherwise stated. See Oak's terms & conditions (Collection 2).

Lesson video

Skip lesson video

Loading...

Prior knowledge starter quiz

Download quiz pdf

6 Questions

Q1.
What do we call the message or instruction you give to a large language model to get a response?

Correct Answer: prompt

Q2.
Why is it important to make your prompt specific when using an LLM?

It makes the model run faster.
It saves computer memory.
Correct answer: It helps the model to determine exactly what you want.
It changes the training data.

Q3.
What can happen if your prompt to an LLM is too vague?

Correct answer: The model may give a general or unrelated response.
The model will always give a perfect answer.
The model will stop working.
The model will ask you to try again.

Q4.
Match the keyword to its definition.

Correct Answer:prompt,the input or question given to the LLM

the input or question given to the LLM

Correct Answer:vague,unclear or too general

unclear or too general

Correct Answer:context,extra information to help the LLM understand your prompt

extra information to help the LLM understand your prompt

Correct Answer:engineering,using knowledge to design or improve things

using knowledge to design or improve things

Q5.
Why is it important to check LLM responses for bias?

It makes the answers longer.
It improves the graphics.
It changes the model’s training data.
Correct answer: Bias can make answers unreliable or unfair.

Q6.
Why should you critically evaluate the responses from an LLM?

because the model always repeats the same answer
Correct answer: because the responses might reflect bias from the training data
because the LLM is always correct
because it is required by law

Assessment exit quiz

Download quiz pdf

6 Questions

Q1.
Which statement about machine learning (ML) models is correct?

ML models are always neutral and never make mistakes.
ML models automatically correct any unfairness in data.
ML models automatically correct any unfairness in data.
Correct answer: ML models can reflect or repeat bias found in their training data.

Q2.
Put these steps in order for training an ML model fairly.

1 - Collect a large, diverse dataset.
2 - Train the model on the data.
3 - Test the model with separate data.
4 - Check for bias in the results.

Q3.
Why do ML models need both training data and separate test data?

to speed up the computer
to use more storage
Correct answer: to make sure the model works reliably on new data
to repeat the same results

Q4.
What does it mean if data is “unrepresentative”?

Correct answer: It does not reflect the whole group or situation.
It covers all possible groups equally.
It is always fair.
It is the same as diverse data.

Q5.
What is one effect of bias in an ML model?

The model always gives the right answer.
The model is always neutral.
The model ignores its training data.
Correct answer: The model favours some groups unfairly.

Q6.
Match the keyword to its definition.

Correct Answer:bias,when results unfairly favour or exclude certain groups

when results unfairly favour or exclude certain groups

Correct Answer:context,extra information that helps a model interpret your input

extra information that helps a model interpret your input

Correct Answer:fair,treating all groups equally, without favouritism

treating all groups equally, without favouritism

Correct Answer:machine learning,when data is used to help a model get better at recognising patterns

when data is used to help a model get better at recognising patterns

Correct Answer:training data,data used to check if a model works well on new information

data used to check if a model works well on new information

Correct Answer:test data,data used to check if a model works well on new information

data used to check if a model works well on new information