Exploring the Use of Machine Learning as Game Mechanic – Demonstrative Learning Multiplayer Game Prototype
Jan Dornig, Changbai Li · 2020
Games and Machine Learning (ML) most commonly intersect in the areas: 1) ML testing in Games, 2) ML use for Game-AI and 3) ML use for game assets. These areas have seen strong interest in the past, but the use of ML as main game mechanic has been less explored so far. Most likely due to a lack of real-time learning capabilities and inaccessibility of algorithms to the gaming community in the past. While some game-like experiences like the "Game of life"-type simulations exist and versions of it use ML, these usually provide minimal interaction and confine the player to initial parameter setup and subsequent watching of a simulation. This paper describes a digital game prototype which explores using Demonstrative Learning in a multi-player setting as main game mechanic, to achieve interaction levels similar to Interactive Machine Learning. We describe the process the considerations that have led to using this type of ML among other decisions related to finding suitable compromises between capabilities and needs of ML and ideal player experience. The prototype is used to discuss game design choices and shows how a focus on the ML-mechanics can inspire game design and facilitate creative design and future research and design possibilities.