05.12.2023 | Warm Up | 3.10 Pt 2

Machine learning algorithms are never biased – it is impossible for them to make unfair or inaccurate predictions.

• True

• False

Which of the following describes algorithmic bias?

• A human being who loves the color yellow, always buys yellow clothing

• An AI algorithm leads to outcomes which are unfair in a systematic way

• A calculator makes a lot of mistakes in seemingly random ways

• An AI algorithm makes 100% correct predictions.

One way of minimizing the risk of algorithmic bias is to use diverse and representative datasets for AI systems.

• True

• False

Which of these is an example of algorithmic bias?

• An algorithm to identify farm animals makes lots of mistakes but an equal number of mistakes predicting each animal.

• An algorithm to identify whether a person is cheating on a test wrongly accuses boys of cheating more often than girls

• An algorithm to identify flowers works perfectly accurately for many different types of flowers.

• A store owner decides to use AI to stock products instead of making the decisions herself.

What is a potential negative consequence of an AI system designed to help blind or visionimpaired people cross a street, if it fails to work properly? (Select all correct answers.)

• A vision-impaired person could be told it’s safe to cross a road which is actually unsafe, putting them in danger of an accident

• A vision-impaired person could be told it’s unsafe to cross roads which are actually safe – which prevents them from crossing any streets at all.

• A vision-impaired person uses an AI app that works 100% perfectly and helps them cross safely all the time.

• A vision-impaired person decides to turn off the AI system and not cross any streets.

An AI used to classify dolphins seems to be biased towards incorrectly identifying sharks as dolphins. Which of the following is most likely to have caused this?

• The training dataset didn’t contain enough diverse examples of sharks

• The training dataset didn’t contain any images of dolphins.

• The training dataset contained only images of trucks.

• The training dataset contained many diverse examples of both sharks and dolphins

In most apps that use AI – like social apps – the AI does not exist on its own. Most apps also include a user interface and many different additional functions that help make it useful.

• True

• False

. When you design an app that uses AI, you should first find and research a real-world problem you are trying to solve.

• True

• False

Which of the following are steps in designing an app using the iterative design process (Select all correct answers.)

• Research a problem

• Ideate and brainstorm possible solutions

• Do an inspirational dance

• Build an app without any planning

• Test your solution and get feedback

When you design an app that uses AI, you should think about the users - or stakeholders - of an app, and think about how your app impacts them.

• True

• False