Project Proposal – Procedurally Generated Games with AI-Motivated Agency: Blurring the Line between Humans and AI


In this paper (via the long format – no project), I would like to talk about and research the direction in which game design and development appears be going in, which is open-story procedural game generation, and how AI-controlled and generated game agents will be used to represent players and player personas – and what type of impact this will have on our culture.

Computer games have had over 50 years of time within our culture and have metamorphosed from simple block-shapes offering rudimentary game mechanics, all the way to lush 3D open-world environments with complex AI-controlled Non-playable Characters (NPCs) with physics engines that simulate real-world action, interaction, and dynamic environments.

If we look at two driving forces that are pushing games in the above stated direction, we can articulate them as: (1) procedurally generated worlds (i.e., dynamic content generated based on rulesets) and (2) AI-driven agents that support varying degrees of autonimity while the player is away.

Looking at the former, we see large game titles such as 2009, Call of Duty: Modern Warfare 2, 2013 Grand Theft Auto V, where literally hundreds of millions of dollars were spent to produce the blockbusters (List of most expensive video games to develop, 2017). It therefor becomes apparent that the complexity of such large-scale titles pushes the cost of content production exponentially. This is due to all of the game assets and what is entailed to realize them within the game environment starting with concept artwork, then 3D modeling, all of the image maps (texture, specular, transparency, ambient, etc.) that are projected onto the 3D meshes, the physics properties that are assigned to the objects, AI driven NPCs, and finally any sort of interactive routines that are required to allow the asset to be interacted with (e.g., moving, exploding, smoking, blowing in the wind, etc.) by or because of the player(s).

This cost is thus leaning on and creating pressure to dynamically generate most – if not all – of the content and assets that go into such an epic game.  From NPCs and variance in their faces, walks, skin color, clothing, and overall behaviors, to buildings’ shapes, sizes, color, placement, and other environmental settings such as forests, lakes, streams, and rivers.

Addressing the latter, AI driven and managed personas of player characters truly blurs lines between computer and person. Currently, player character customization means setting characteristics such as name, age, gender, likes, dislikes, hair color, and simple wardrobe choices. However, as AI (via deep learning) becomes more powerful (and customizable – even personalizable), players will be able to set complex and seemingly abstract choices such as behaviors that the AI should do in stead while the player is away. Things such as purchasing power-ups, how to defend a base, and whether or not to form an alliance with others will become commonplace. Actor-network theory (Banks, 2011) becomes the center of the discussion for this and new questions will need to be addressed.

Questions such as who owns the story? Who’s legally liable for illegal activity during player absence? How will this new paradigm affect culture? How will this affect human behavior? Will this debase the human condition (Chagani, 2014)? These are questions that I wish to explore and attempt to address.



Banks, D. (2011, December 2). A Brief Summary of Actor Network Theory. Retrieved from The Society Pages:

Chagani, F. (2014). Critical political ecology and the seductions of posthumanism. Journal of Political Ecology, 13.

List of most expensive video games to develop. (2017, Feb 22). Retrieved from Wikipedia:

Leave me a Comment