Understanding AI with online games
There is no better way to learn than to put your hands in the dough
I have always admired the experts who are able to adapt their discourse to the listener's level — especially their ability to synthesize and convey complex information in an intuitive way. For example, WIRED’s 5-level video series is a perfect demonstration of specialists discussing incredibly complex theories to schoolchildren, college students, adults, and fellow researchers.
During my Masters, I got interested in Machine Learning, and I started collecting interactive games explaining AI concepts in an innovative way. In this article, I will share 3 free online games (+ a bonus) to understand how AI works, to grasp the complexity of building a “fair AI”, and hopefully make you think about the implications of such a technology on our society.
Survival of the Best Fit — Get into the shoes of a hiring manager (play here)
In this game, you take the role of a hiring manager, and you will train a machine learning algorithm to hire candidates. Over the game, you will see that such algorithms are infused with human bias. The goal of this game is to prove that machine bias is pervasive, and that there is no such thing as “neutral software”. The creators wish to reframe the current discourse: AI is presented to us as a threat, but in reality, we are contributing to those biases. We should take accountability when building such software systems.
MonsterMatch — Learn how the algorithms figure you out in a dating app (play here)
Whether you are using a dating app, or you promised yourself never to do so, the way they work influences your life. Dating apps (as well as your Facebook feed, YouTube recommendations, Google searches…) use -among others- an algorithm called collaborative filtering. Early users influence what later users will see. For example, Juliet swipes right on Romeo, but left on Clyde. Later, when Bonnie joins the app and exhibits similar behavior as Juliet by swiping right on Romeo, the algorithm will learn from its experience with Juliet and assume she won’t like Clyde, and not show his profile. As you swipe right or left, you will get an intuition about the way those algorithms work.
Courtroom algorithm game — Can you make AI fairer than a judge? (play here)
This game’s format is different: it is an article with interactive data visualizations. It explores COMPAS, a risk-assessment tool used by the US legal system, which attributes a “risk score” to each defendant, thus helping judges with their sentence decisions. During the game, you will set thresholds to determine whether an individual must be released or jailed, and the exercise gets increasingly complicated as new factors come into play. You will realize it is difficult to define “fairness”, and that the problem is not merely algorithmic.
BONUS:
Bingo — A bingo game to teach children about AI
This game has been developed by Blakeley H. Payne, a researcher at MIT, for schoolchildren. The goal is to expose them to technology, foster interest in STEM, and help them realize how AI is incorporated into different aspects of our lives. After players are given an introductory course to AI and machine learning, they have to match the correct Datasets (past songs that you’ve listened to) and Predictions (new songs you may like) to a Task (listen to a recommended song on Spotify). Just as in traditional Bingo, the first team having correctly completed 5 tiles (diagonally, in a row, in a column) wins.
There is an ongoing conversation around ethics and AI, but I feel like a lot of people take a stand in this debate with a simplistic understanding of the technology (see ultracrepidarianism). I think those games are a great help to grasp the real-life implications of AI. I hope you learned something after reading this article and playing those games !
