# Generative Agents
Generative agents are software agents powered by large language models that store their experiences in a natural-language memory, reflect on that memory to form higher-level conclusions, and plan their behavior, so that a population of them can produce believable social life inside a simulated environment. The method was introduced in April 2023 by [Joon Sung Park](https://www.edgepedia.org/joon-sung-park) and colleagues at Stanford University and Google in the preprint "Generative Agents: Interactive Simulacra of Human Behavior" (arXiv:2304.03442), which populated a small town of 25 agents in a sandbox inspired by [The Sims](https://www.edgepedia.org/the-sims).[1](https://arxiv.org/pdf/2304.03442)
| Key fact | Detail |
|---|---|
| Origin | Park et al., Stanford and Google, April 2023 preprint; peer-reviewed at ACM UIST 2023[1](https://arxiv.org/pdf/2304.03442) • [2](https://dl.acm.org/doi/fullHtml/10.1145/3586183.3606763) |
| Environment | Smallville, a Sims-inspired town of 25 agents users could interact with in natural language[1](https://arxiv.org/pdf/2304.03442) |
| Core architecture | Memory stream, retrieval by relevance, recency, and importance, reflection, and planning[1](https://arxiv.org/pdf/2304.03442) |
| Believability | Full architecture scored μ=29.89 (σ=0.72) versus μ=22.95 for a crowdworker human baseline; effect size d=8.16 against the prior-work condition[1](https://arxiv.org/pdf/2304.03442) |
| Emergent behavior | Knowledge of a mayoral candidacy spread from 1 to 8 of 25 agents and of a party from 1 to 13 over two simulated days; relationship network density rose from 0.167 to 0.74[1](https://arxiv.org/pdf/2304.03442) |
| Cost | Thousands of dollars in token credits and multiple days of wall-clock time for 25 agents over two game days[1](https://arxiv.org/pdf/2304.03442) |
| Code | Openly released on GitHub (joonspk-research/generative_agents)[3](https://github.com/joonspk-research/generative_agents) |
## What generative agents are
A generative agent is an LLM wrapped in an architecture that gives it memory, self-generated conclusions, and a schedule. The agents live in [Smallville](https://www.edgepedia.org/smallville), an interactive sandbox inspired by The Sims, where end users can interact with a small town of twenty-five agents using natural language.[1](https://arxiv.org/pdf/2304.03442) [Ars Technica](https://www.edgepedia.org/ars-technica), covering the release in April 2023, described it as a miniature RPG-style virtual world where 25 characters, controlled by ChatGPT and custom code, live out their lives independently with a high degree of realistic behavior.[4](https://arstechnica.com/information-technology/2023/04/surprising-things-happen-when-you-put-25-ai-agents-together-in-an-rpg-town/)
The point of the design is that believable behavior comes from more than a good prompt: each agent must remember what happened to it, decide what those events mean, and turn that understanding into concrete plans and moment-to-moment actions.
## Origin and the Park et al. 2023 study
The method was introduced by a Stanford and Google team in the April 2023 preprint arXiv:2304.03442.[1](https://arxiv.org/pdf/2304.03442) The work was peer-reviewed and published at ACM UIST 2023 (DOI 10.1145/3586183.3606763).[2](https://dl.acm.org/doi/fullHtml/10.1145/3586183.3606763) The authors released the code openly on GitHub under joonspk-research, along with practical warnings that simulations as of early 2023 could be costly, especially with many agents, and that OpenAI's API can hang at hourly rate limits, requiring simulation restarts.[3](https://github.com/joonspk-research/generative_agents)
The study's headline claims were emergent social behavior: without user intervention, information diffused through the town, and relationships accumulated, described in the numbers section below. By September 2026 the paper's SciSpace page recorded 929 citations.[5](https://doi.org/10.1145/3586183.3606763)
## How the mechanism works
The architecture has three components.[1](https://arxiv.org/pdf/2304.03442)
**The memory stream** is a long-term memory module that records, in natural language, a comprehensive list of the agent's experiences. A memory retrieval model combines relevance, recency, and importance to surface records that inform moment-to-moment behavior.[1](https://arxiv.org/pdf/2304.03442)
**Reflection** synthesizes memories into higher-level inferences. **Planning** translates those conclusions into high-level plans and then detailed behaviors. Ars Technica's description of the loop matches the paper: based on the agent's perceptions, the architecture retrieves relevant memories, then uses those retrieved memories to determine an action, to form longer-term plans, and to create higher-level reflections, which re-enter the memory stream as new memories.[4](https://arstechnica.com/information-technology/2023/04/surprising-things-happen-when-you-put-25-ai-agents-together-in-an-rpg-town/)
In short, the loop is: perceive, retrieve from memory, act, write the experience back to memory, and periodically reflect so that scattered observations become durable conclusions that shape later plans.
## By the numbers
All figures below are from the original paper's own evaluation; no independent measurements are available in the evidence for this article.
**Believability.** In an ablation (test removing components to measure each one's contribution) evaluation, the full generative agent architecture produced the most believable behavior (μ=29.89; σ=0.72). Removing reflection scored μ=26.88; removing reflection and planning scored μ=25.64; removing memory, planning, and reflection scored μ=21.21. Comparing the condition representing prior work to the full architecture produced a standardized effect size of d=8.16.[1](https://arxiv.org/pdf/2304.03442) A crowdworker human baseline scored μ=22.95 (σ=0.69), which the full architecture exceeded; the authors note this baseline did not represent maximal human believability.[1](https://arxiv.org/pdf/2304.03442) • [2](https://dl.acm.org/doi/fullHtml/10.1145/3586183.3606763)
**Information diffusion and relationships.** Over the two-day simulation, the number of agents who knew about Sam's mayoral candidacy increased from one (4%) to eight (32%), and the number who knew about Isabella's party increased from one (4%) to thirteen (52%). Relationship network density rose from 0.167 to 0.74.[1](https://arxiv.org/pdf/2304.03442)
**Behavioral texture and errors.** Out of 453 agent responses regarding their awareness of other agents, 1.3% (n=6) were hallucinated. In the party scenario, 5 of 12 invited agents showed up at Hobbs cafe for Isabella's Valentine's Day party.[1](https://arxiv.org/pdf/2304.03442)
**Cost.** The authors report the study cost thousands of dollars in token credits and took multiple days to complete for 25 agents over two days of game time.[1](https://arxiv.org/pdf/2304.03442) The code repository adds that running such simulations, at least as of early 2023, could be somewhat costly, especially with many agents.[3](https://github.com/joonspk-research/generative_agents)
## References
1. [Generative Agents: Interactive Simulacra of Human Behavior (arXiv preprint)](https://arxiv.org/pdf/2304.03442)
2. [Generative Agents: Interactive Simulacra of Human Behavior (UIST 2023, ACM)](https://dl.acm.org/doi/fullHtml/10.1145/3586183.3606763)
3. [joonspk-research/generative_agents (official code repository)](https://github.com/joonspk-research/generative_agents)
4. [Surprising things happen when you put 25 AI agents together in an RPG town (Ars Technica, April 2023)](https://arstechnica.com/information-technology/2023/04/surprising-things-happen-when-you-put-25-ai-agents-together-in-an-rpg-town/)
5. [SciSpace paper page: Generative Agents (2023)](https://doi.org/10.1145/3586183.3606763)
---
*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › Foundation-model methods and training › Prompting, reasoning, and agents*
*Initially written Sep 17, 2026 · Edited: Oct 11, 2026; Oct 15, 2026*
*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*
License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license