Artificial Tears

Artificial Tears: a video game about developing empathetic AI

Minh-Quan Nguyen, College of Engineering

Collaborators: Emily Sulkey, UMSI; Chani Kohtz, LSA; Hilary Kohtz, Stamps

Faculty Advisor: Austin Yarger, Lecturer II, CoE

The impending Age of Artificial Intelligence has only just begun to shape how we as a society understand and interact with the world around us. Art has played, and will continue to play an active role in developing the public’s consciousness and understanding of AI technology.

Questions not long ago deemed theoretical and reserved for the most philosophical of academics have now fallen upon the shoulders of the average individual to grapple with. What are we building, and how do we know that what we’re creating and using has the best interests of humanity at heart?

In Artificial Tears, players will assume the role of an AI researcher tasked with developing an empathetic Artificial Intelligence to understand the depths of the human condition. Gameplay will involve abstracted versions of the procedures used by real AI researchers to train Reinforcement Learning (RL) agents, and the general experience can be most likened to a game from the “Pet Simulator” genre (with experimental research lab theming and much more serious tone).

Actual machine learning methods and technology will of course be integrated under the hood to facilitate gameplay, as will published computational Cognitive Science models of human emotions (see Supporting Materials). While 3D graphics and art will be essential in building out the “training sandbox” for training the in-game AI machination, 2D artwork will be necessary for UI and HUD elements.

The intended impact of this game is to both offer a suggestion on how we can develop AI that are truly empathetic at their core, as well as provide players with educational insight into how AI is developed. This will offer those with no technical background a jargon-free, handson opportunity to learn more about the nuances and implications of AI technology.

This project, like any other video game project, is inherently interdisciplinary. Expertise in both machine learning as well as the principles of game design will be required to develop an engaging and meaningful core gameplay loop. In addition, cohesive and fitting art and audio work will be critical to immersing the player in the themes and emotions that this work is meant to convey. The intended deliverable is a playable experience to be publishable online via a web-build, as well as through desktop clients.

The project is to be finished by the end of Winter 2024 semester, with the last 2 weeks of February dedicated to design/pre-production and plans to have a minimum viable product by end of March, and fully polished product by end of April.

The collaborators on this project have worked with each other before and shipped finished games as part of both WolverineSoft Studio and 48-hr game jams. Their skillsets and thus roles on this project are as follows:

Minh-Quan

  • Game Design and Development (i.e. programming, technical art implementation, etc)
  • Game Project Management
  • Machine Learning (Including Reinforcement Learning) backend

Emily

  • UX Design
  • Video Game Music Composition

Hilary

  • Game Art (2D Assets Only)

Chani

  • Game Design and Development
  • Machine Learning
  • Computational/scientific modeling & data analysis