Major Studio1 - Final Project - Bias Slot Machine

Time:  Nov. 2018 ~ Dec.2018

Introduction

Till now, 180 human biases have been defined and categorized (decision-making biases, social biases, gender bias, cognitive bias, memory errors, etc.). Everyone is the creator of dissonance in a way we don’t notice or even don’t want. People learn stereotype from a lot of places. The way they brought up, the information they get from TV, movies, social media, the education they receive and the people they team with, all shape the ways they think. In our daily life, we are always viewing others with prejudice, and we are always being examined with prejudice. Studies found even educators have serious biases against their students, and their words and actions directly affect the perceptions of the next generation.

Not only people have the bias, but AI also has the bias. Not only do biases spread from person to person, but data and algorithms spread and create new biases. As Cynthia Dwork said, "Historical biases in the training data will be learned by the algorithm, and past discrimination will lead to future discrimination." This assumes that the algorithmic bias will expand and never be eliminated. Deirdre k. Mulligan says "Any system that is making sorts of predictions is going to have biases." Machine learning models will, therefore, end up give preferences to some particular groups of people over others in a way that is not justified. Machine learning is a means to pursue more accurate input prediction. It has clear inputs and outputs and very clear optimization goals. “To create a model, then, we make choices about what’s important enough to include, simplifying the world into a toy version that can be easily understood and from which we can infer important facts and actions.” 2 For example, when the user has the highest probability of clicking or the where can generate the highest income and so on. The model evolves in that direction during training. The input and output are all defined by people. So the root of algorithm bias is human bias. However, if the original data is the real data, the algorithm just simulates the actual model.

Design Questions

Base on the introduction, the topic of this final project is going to focus on the human bias because the source of algorithm bias is human bias. The research questions are:


1 How bias shapes our ways of thinking and how it influences the person whom we have biases on, both in reality and in the AI world.

2 How to use design to educate people to act without bias in a light and playful way, and what’s the impact of bias education.

Precedents

Group 2.png
Group 2 Copy.png

Concept Description

Based on facial recognition, the result of a slot machine is no longer defined by luck, but by the operator’s gender, racial, and other outside appearances. The user does not know the mechanism of the slot machine, but certain people who always win and others who never earn anything. And when they figure out why they will understand that’s exactly the mechanism of prejudice in life, without fairness, without principle, some people get preferences. This slot machine aims to raise awareness toward the issue of bias, and to some extent to guide people to act without bias. When a kind of persons always wins or always loses, they will have the willingness to explore the mechanics behind the game. The bias slot machine will have a button that will reveal the answer to the question when pressed. Players will be given a corresponding anti-bias card as feedback.

This is also a combination of human bias and AI bias. Human Bias is the back mechanisms, and the rule of this game is the opposite of reality. Algorithm bias is that facial recognition can get wrong because of the bad data. However, if the original data is the real data, the algorithm just simulates the actual model. The mistakes AI makes are also the mistakes humans might make in reality.

Group 4.png
Group 8.png

Demo Video

This Demo is played on an iPad. It's not my ideal device because there are too many limitations on iOS:

1. I use a webcam, but in iOS Camera and Browser are two different applications that cannot run simultaneously. There is no such problem on the PC - After the user clicks the take a spin button, the slot will run and show the result.

2. A touch event is mandatory to load and play audio on iOS, so there is no spinning sound, fail sound or success sound.