Ritwik Sharma — PhD Research Portfolio

Doctoral Research Proposal · September 2026

Ritwik Sharma

Game Designer · Systems Engineer · Researcher

Autonomous Administration of Live Game Worlds

What happens when the administrator of a live game world is an agent, and can the understanding it accumulates while running that world become the world’s own unauthored history?

[email protected] +91 94623 32146 LinkedIn phd.ritwiksharma.com GitHub: Zombita Bikaner, Rajasthan, India

Research Background

MA, Game Design — London College of Communication, University of the Arts London (2019–2021). The dissertation examined branching, narrative illusion, and the conditions under which player choice produces emergent story in role-playing games. Director and Lead Developer at Inter Tech Interact since December 2022. Author of Under The Cherry Blossoms (2024). A paper on the architecture and its emergent-narrative argument was submitted to AIIDE 2026.

Why This Proposal Exists

The question predates the vocabulary I now have for it. It was first written down in 2018 as a note about action-driven non-linear narrative that ended mid-sentence, reappeared in the MA dissertation, and has since been pursued through a live production system, a controlled agent prototype with a formal measurement instrument, and two fully specified game designs. The doctoral work is not to build the system. It is to turn it into a research instrument.

Four Research Questions

RQ1 · Authority

Bounded, auditable, extensible authority

How can administrative authority over a live game world be transferred to an LLM agent so that its power stays bounded, auditable and progressively extensible, and when do its proposals converge with what a competent human administrator would decide?

RQ2 · Understanding

What the administrator comes to know

What does an agent that continuously administers a real community come to understand about it, and how does that interpreted record differ, in content, in structure and in what it makes narratable, from the behavioural logs of simulated storyworlds?

RQ3 · Narrative

Participation versus recovery

Does a season-end history authored by the participant-administrator produce narrative the community recognises as true and meaningful, and how does it compare with an account sifted from the same record by an instrument that reads it from outside?

RQ4 · Transfer

From real people to simulated characters

Does the pattern transfer to a synthetic world whose inhabitants are simulated characters, and what does that comparison reveal about which properties depended on the inhabitants being real?

RQ3 carries most of the weight: the same season record, narrated by the administrator that formed its own reading of these people, and sifted by an instrument that reads it cold, both placed in front of the community whose season it was.