Award Won!

Best Social Media Crowdsourcing Game

Award Info

Becoming Fei

Organization:

Institute for Creative Technologies, University of Southern California

Dev Category:

Release Year:

Becoming Fei is an educational game aims to teach AI and data literacy for novice learners. The game places players in a foreign planet with the mission to rescue a missing scientist. When things start to go wrong, Fei (played by the learners) must figure out what’s going wrong with the AIs she uses to assist her in the mission, and how to fix them. The game utilizes the main character’s monologue to guide the learners through the basic process of machine learning, and familiarize them with key issues and concepts that impact the performance of AI.

Game Overview

Becoming Fei aims to help novice learners without computer science or engineering background learn the basic process and components of machine learning, in particular supervised learning. Given the availability of user friendly machine learning tools, novice learners can often train a model using readily available datasets. The game aims to help novice learners gain understanding of the impact of data and the choice of algorithm on trained models, how to evaluate a trained model, and the ethical considerations. For example:

1. Supervised learning involves data, algorithm, and a trained model;
2. Data can impact the choice of algorithms, resources required for training, and model performances;
3. Evaluation of a trained model: accuracy, confusion matrix, precision, and recall;
4. Evaluation of a trained model: What is overfitting and how to prevent it.
5. Ethics of AI: what are the real-world impact of of Precision and Recall of a trained model;
6. Ethics of AI: what are the real-world implications of explainability of AI, and the explainability of different machine learning models.

The narrative places the player on a foreign planet for a mission to search and rescue a missing scientist. The primary goals the player faces are: searching for the target and survive on the foreign planet, including fending off enemy attacks, avoid poisonous plants in the environment, and collect resources for survival. Some examples are:

(1) Maintaining RECA, the reconnaissance AI, to identify and alert the player of hazard in the environment;
(2) Maintaining EMA, the AI for predictive maintenance of the player’s equipment;
(3) Maintaining MEDAI, the AI for medical diagnosis and treatment.

The AIs assisting the player (i.e., RECA, EMA, and MEDAI) are trained using data not collected from the planet the player is deployed in. Thus the AI models are prone to error. The player must familiarize themselves with the machine learning process in order to retrain the model and obtain one with acceptable performances for the new environment.

Additionally, when the AI’s recommendations defies the players expectations in critical decisions (e.g., medical treatment), the player must find ways to examine the AI models’ decision processes.

High school students and those with high school education. No prior experience with computer science or engineering required.

We have designed a pre-survey and a post-survey to assess the player’s knowledge skills and attitudinal change.

The pre-survey is at: https://presurvey.becomingfei.org

The post-survey is at: https://postsurvey.becomingfei.org

Game Specs

Game Engine
Operating System
Platform

Game Video

Play the video below to learn more about Becoming Fei

Play