Non-player characters that learn, grow up, and pass it on: NPCs as independent entities

Game characters usually share one behavior and memory policy, so they differ in parameters and voice, not in kind. We build each character as an independent entity whose behavior is its own data. In Village Sim, retrieval over natural-language instructions drives every action, word, and memory, and runs the world. Each villager forms its own memories and, growing up, writes new instructions from them. Villagers pass on news, learn at school and in a trade, marry, and have children who inherit their parents' traits. Over six forty-day runs of thirty villagers, news reached a median of 29 in 20 wordings, drifting toward danger. Decisions only knowledge made possible rose from 3.4% on day one to 24.8% on the last. Swapping temperament in replays changed 8.2% of actions and how news was worded and kept. Three ten-year villages grew from thirty to 46–59 people, and 300 villagers ran the same way.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-15
DOI
https://doi.org/10.5281/zenodo.24238881
Primary Topic
Artificial Intelligence in Games
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Non-player characters that learn, grow up, and pass it on: NPCs as independent entities

Scott N. Hwang
Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence in Games
preprint

Non-player characters that learn, grow up, and pass it on: NPCs as independent entities

Scott N. Hwang
preprint en

Abstract

Game characters usually share one behavior and memory policy, so they differ in parameters and voice, not in kind. We build each character as an independent entity whose behavior is its own data. In Village Sim, retrieval over natural-language instructions drives every action, word, and memory, and runs the world. Each villager forms its own memories and, growing up, writes new instructions from them. Villagers pass on news, learn at school and in a trade, marry, and have children who inherit their parents' traits. Over six forty-day runs of thirty villagers, news reached a median of 29 in 20 wordings, drifting toward danger. Decisions only knowledge made possible rose from 3.4% on day one to 24.8% on the last. Swapping temperament in replays changed 8.2% of actions and how news was worded and kept. Three ten-year villages grew from thirty to 46–59 people, and 300 villagers ran the same way.

Zenodo (CERN European Organization for Nuclear Research)
Pennsylvania State University (US)
Artificial Intelligence in Games
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon