Overview

Bias in AI is a common but difficult problem to solve. Oftentimes, it can go unnoticed and affect real world outcomes such as the quality of a person's medical treatment, whether or not someone is hired for a job, and much more. As AI becomes more present in society, bias will become a bigger problem which needs to be addressed. To help students become more aware of AI and bias, we propose the design of an interactive web application which can give students both technical and ethical insights into AI. By teaching students to recognize the implicit biases which exist in modern AI models, they will be better equipped to navigate the world of AI. Students will learn how the training data directly affects the model behavior. Students will begin by being provided data samples for a classification task. Upon reviewing the samples, students will be asked to complete a timed classification task where they must best use their given samples to classify new samples. This could be classifying written digits, fruits, etc. Once the student is done, a model trained on the same data will repeat the task. Students can then see the difference in their preformance compared to hte models. After seeing bias in the models preformance, students will have a chance to changeFrom this, students should learn about how models learn from data and how bias can be hidden in samples. 

What type of AI Ideas will be taught?

Students will learn about how computers learn from data as well as the ethics surrounding AI.

How will students interact with the system?

Students will learn from the training data, classify a set of samples, and gain an understanding of bias in AI. Additionally, students will observe a bias model while it classifies samples and review real world cases of AI bias.

How will the project engage students with the ethics of AI?

By engaging with the idea of bias through a game setting, students will have a better understanding of how bias becomes a problem in the first place. By applying this new awareness to real life scenarios of bias, students will be able to see the ethical implications of AI in the modern world. 

AI Model Use and Awareness

Goal

The goal is to bring awareness about the applicability of ml algorithms in the real world.

Platform

Cognimates or machinelarningforkids +  Scratch 

What is the problem

One of  the reasons why AI has been taught to kids is so that they understand the applicability and behave responsibly while they are dealing with it. At the level of K-12 ,  Kids are not expected to learn the more technical details of machine learning, instead they can  be focused on general intuition of the algorithms and how they are applied. 

How are we addressing the problem

We address this by providing an interactive interface where we can give accessibility of these algorithms in the form of blocks and now kids can understand how input is given to specific algorithms and what kind of output the algorithms give in return . Later we build the real application scenarios on scratch so that kids can now relate to these algorithms . This might enforce the kids to think about which algorithms are behind these applications they use in daily life . Once they can conceptualize about the algorithm they would be interact with AI better 

The idea

We would create machine learning algorithms on cognimates or machine learningforkids , and visualize them as blocks on scratch platforms. Kids interact with these blocks by video, voice, or drawing  and understand the intuitions of how these blocks work. Now after they get comfortable with a few machine learning algorithms blocks in Scratch then we introduce them to scenario based questioning, where kids now see visual scenarios and have to guess the machine learning algorithm  block that is best suited in the scenario.

What type of AI Ideas will be taught?

Awareness + Learning 

Interface design ideas

How can students interact with the system?

How different algorithms will do different things , image recognition involves webcam/uploading sprite, vs voice to speech recognition - involves kids talking about something using a microphone or drawing the sprites .

What data can be gathered?

• qualitative (interview, survey)

 Trigger their thinking capability , how kids would use these recognition item in other applications. Ask a student scenario , how a combination of these algorithms are used .

For example in the case of recognition what algorithms among can be used , now kids should identify the algorithms image recognition + voice recognition 

How will the project engage students with the ethics of AI?

They deal responsibly with these applications and can build their thinking capability to use different combinations of these algorithms . Better Awareness of the model the students can understand the limitations of the model, saying these models don't detect everything . 

Things we intend to add to the idea

We would like to visually show the limitations of these algorithms, especially Bias . However we might be much more conclusive on this idea in future .