CS: Machine Learning and Artificial Intelligence Basics
Understanding data, models, training, and responsible AI
CS: Machine Learning and Artificial Intelligence Basics
Understanding data, models, training, and responsible AI
CS - Grade 9-12
- 1
In your own words, explain the difference between artificial intelligence and machine learning.
Think of artificial intelligence as the larger category and machine learning as one method inside it.
Artificial intelligence is the broader field of creating computer systems that can perform tasks that seem intelligent, such as understanding language or making decisions. Machine learning is a part of artificial intelligence where a system learns patterns from data instead of being programmed with every rule by hand. - 2
A music app recommends songs based on songs a user has liked in the past. Identify the input data, the model's task, and the output.
The input data includes the songs the user liked, skipped, played often, or rated. The model's task is to find patterns in the user's preferences. The output is a list of recommended songs the user may enjoy. - 3
Classify each task as supervised learning, unsupervised learning, or reinforcement learning: a spam filter trained with emails labeled spam or not spam; a program grouping customers by shopping behavior without labels; a robot learning to move by receiving rewards for reaching a goal.
Labels point to supervised learning, hidden groups point to unsupervised learning, and rewards point to reinforcement learning.
The spam filter is supervised learning because it uses labeled examples. Grouping customers without labels is unsupervised learning because the system finds patterns on its own. The robot learning from rewards is reinforcement learning because it improves by receiving feedback from actions. - 4
A model is trained to predict whether an image contains a dog or a cat. The training set has 9,000 dog images and 1,000 cat images. Explain one problem this may cause.
Compare how many examples the model sees for each class.
The model may become biased toward predicting dog because most of the training examples are dog images. It may perform poorly on cat images because it has seen far fewer examples of cats during training. - 5
A machine learning model is tested on 200 images. It correctly classifies 170 of them. Calculate the model's accuracy as a percentage.
The accuracy is 170 divided by 200, which equals 0.85. As a percentage, the model's accuracy is 85%. - 6
Explain why it is important to test a machine learning model on data that was not used during training.
A useful model should work on new data, not just old examples it has already seen.
It is important because the test data shows how well the model works on new examples. If the model is only checked on training data, it may look accurate because it memorized the training examples instead of learning patterns that generalize. - 7
Define overfitting and give a simple example of what it might look like in a machine learning project.
Overfitting happens when a model learns the training data too closely, including noise or unusual details, so it performs poorly on new data. For example, a model that gets nearly every training image correct but makes many mistakes on new images is likely overfitting. - 8
A school wants to use an AI system to help decide which students may need tutoring. Name two types of data that might be useful and one type of data that should be handled carefully for privacy or fairness reasons.
Think about data that helps identify academic need and data that could reveal sensitive personal information.
Useful data might include recent quiz scores and missing assignment counts. Data such as disability status, family income, race, or home address should be handled carefully because it can affect privacy and may lead to unfair decisions if used improperly. - 9
A chatbot gives a confident but incorrect answer about a historical event. Explain why users should not automatically trust every answer from an AI system.
Users should not automatically trust every answer because AI systems can generate incorrect information, misunderstand context, or repeat errors from their training data. A confident tone does not prove that the answer is true, so important claims should be checked with reliable sources. - 10
A model predicts whether a loan application should be approved. It uses past approval data from a bank that historically approved fewer loans for some neighborhoods. Explain how bias could enter the model.
Machine learning models can learn patterns that already exist in the data, including unfair patterns.
Bias could enter the model because the training data reflects past unfair decisions. If some neighborhoods were historically treated unfairly, the model may learn that pattern and continue making unfair predictions unless the data and system are carefully checked. - 11
In a machine learning project, put these steps in a reasonable order: train the model; collect data; evaluate the model; clean and prepare the data; use the model for predictions.
A reasonable order is to collect data, clean and prepare the data, train the model, evaluate the model, and then use the model for predictions. Evaluation should happen before real use so problems can be found and fixed. - 12
Explain the role of features in a machine learning model. Then give two possible features for a model that predicts the selling price of a used car.
Features are measurable pieces of information used as inputs.
Features are the input details a model uses to find patterns and make predictions. For predicting the selling price of a used car, possible features include the car's age, mileage, brand, model, condition, accident history, and location.