# Generative Agents (Smallville)

*Game* · `generative-agents-smallville` · status: **sourced**

A research sandbox in which 25 language-model agents in a Sims-like town keep natural-language memory streams, reflect, and plan their days, producing unscripted information spread and social coordination.

| field | value |
|---|---|
| aliases | Smallville, Generative Agents: Interactive Simulacra of Human Behavior |
| year | 2023 |
| developer | Stanford University and Google Research |
| platforms | Web |
| subgenre | AI-native |
| subgenre secondary | Social sim, Life sim |
| sim domain | Social / interpersonal, Domestic / personal |
| player role | Observer / gardener |
| agency model | Conversational, Indirect incentive |
| time model | Run on demand |
| sim scale | Settlement |
| fidelity | Stylized |
| perspective | Top-down / side 2D |
| win structure | Open sandbox |
| failure model | None |
| narrative mode | Generative |
| multiplayer | Single-player |
| modding | Content |
| production tier | Academic / research |
| agent count order | ~100 |
| emergence | 3 |
| influence | 4 |
| influenced by | [The Sims](../game/the-sims.md) |
| cluster *(computed)* | C08 |
| core count *(computed)* | 3 |
| mechanic count *(computed)* | 7 |

## Summary

A research sandbox in which 25 language-model agents in a Sims-like town keep natural-language memory streams, reflect, and plan their days, producing unscripted information spread and social coordination.

## Core Loop

agents perceive nearby events and objects -> observations enter the memory stream -> retrieve memories by recency, importance, and relevance -> plan the day and react -> converse and act -> reflections summarize memories -> plans and relationships change

## Design Notes

Park and colleagues replace authored behavior trees with a loop of language-model calls on gpt-3.5-turbo. Every observation is stored as text with timestamps; retrieval scores memories by recency (decay 0.995 per game hour), importance (rated 1-10 by the model when stored), and embedding relevance, each normalized and weighted equally. When the summed importance of recent events passes 150 (two or three times a day) the agent writes reflections, and plans are decomposed from a day outline into hour and 5-15 minute chunks. Seeding one agent with the intent to host a Valentine's Day party led to 13 of 25 agents hearing about it and five attending over two game days, at a cost of thousands of dollars in API credits. The open-source release steps the town 10 game seconds per step.

## Relations

**influenced by**

- [The Sims](../game/the-sims.md)

**uses**

- [Memories and thoughts](../mechanic/memories-thoughts.md) (core): Each agent's memory stream stores observations, plans, and reflections as text; retrieval ranks them by recency (0.995 decay per game hour), model-rated importance (1-10), and embedding relevance, and reflections fire when summed importance exceeds 150.
- [Daily schedules](../mechanic/daily-schedules.md) (core): Each morning the model drafts a day plan of about five to eight broad chunks, decomposed into hour-long and then 5-15 minute actions that agents revise when they react to events.
- [Gossip propagation](../mechanic/gossip-propagation.md) (core): Information spreads only through agent conversations; over two game days awareness of a mayoral candidacy rose from 1 to 8 agents and of a party from 1 to 13, with 6 of 453 awareness answers hallucinated.
- [Generative content on demand](../mechanic/generative-content.md) (supporting): Plans, dialogue, importance ratings, and reflections are all generated by gpt-3.5-turbo calls at runtime.
- [Natural-language interaction](../mechanic/natural-language-interaction.md) (supporting): Users address agents in natural language as a persona or as the agent's inner voice, and can rewrite the status of objects in natural language.
- [Agent inspection](../mechanic/agent-inspection.md) (supporting): Researchers interview agents and read their memory streams to evaluate believability.
- [Relationship values](../mechanic/relationship-values.md) (supporting): Agents form acquaintances through conversation; network density rose from 0.167 to 0.74 over two game days.

## Backlinks (computed)

**deploys:** [Generative agents (memory stream, reflection, planning)](../technique/generative-agents.md)

**influenced:** [Project Sid](../game/project-sid.md)

**precedent for:** [Language-agent society terrarium](../opportunity/llm-society-terrarium.md)

## Claims

- In the two-day end-to-end run, relationship network density rose from 0.167 to 0.74 and five agents attended the Valentine's Day party. *(confidence: high)* [[34](../source/src-park-2023-generative-agents.md)]

## Sources

- [Generative Agents: Interactive Simulacra of Human Behavior](../source/src-park-2023-generative-agents.md)
- [joonspk-research/generative_agents](../source/src-park-2023-generative-agents-repository.md)

---
Provenance: created in run `2026-10-24-catalog-gd`; reviewed in `2026-10-25-contributor-readiness`.