GOD: Govern, Observe, and Direct - A Real-Time Control Room for Agent Societies

Yige Luo; Ran Guan

cs.AIAug 28, 2026

GOD: Govern, Observe, and Direct - A Real-Time Control Room for Agent Societies

Authors: Yige Luo, Ran Guan

Organizations: Laboratories, Huawei

Abstract

Generative-agent systems are easier to start than to inspect. A run can contain many agents, locations, messages, commands, and model calls, yet the operator often gets either a finished replay or raw logs. That makes it hard to ask why an agent moved, test a small intervention, or package a run for another researcher. GOD is a local-first control room for agent societies. From the same browser workflow, an operator can issue targeted questions or interventions and inspect the resulting replay state. The system combines a setup wizard, Agent Studio, Map Studio, a spatial replay interface, Ask and Intervene commands, and portable experiment, map, and agent packs. Its technical contribution is the command and artifact loop: live controls and replay evidence share the same operator command model, while package contracts separate scenario, map, and profile data from local runtime state. The public release includes hosted Smallville-style and PKU replays, the open-source repository, and downloadable packs. We evaluate this path on 15 completed run slots. Across the 14 intervention runs, 78 of 84 target-agent checks recorded the commanded destination, and 169 of 182 state answers matched a saved location or action string.

Figures & tables

System	Spatial replay	Targeted ask	Live intervention	Browser/no-code	Portable artifacts
Generative Agents	R	R	P	NR	P
AgentSociety	R	R	R	P	P
AutoGen Studio	NR	P	P	R	R
AgentScope	NR	R	P	R	R
GOD	R	R	R	R	R

Table 1: Feature-level comparison based on the cited primary papers and released GOD artifacts. R denotes an explicitly reported capability, P a related but non-equivalent affordance, and NR that the capability was not reported in the cited source; NR is not evidence of absence. Portable artifacts may be sessions, workflow configurations, replays, or scenario packages.

Figure 1: GOD control room for The Ville. The spatial state, replay and live-step controls, run summary, and shared command input remain visible in one browser view.

Figure 2: PKU campus map used by the 22-agent benchmark and public replay. Labels identify agents at their recorded positions in this frame.

Measure	Value
Completed planned run slots	15 / 15
Event commands avoiding movement routing	14 / 14
Event-specific trace@2	212 / 308 agent-runs
Target destination recorded@1	78 / 84 targets
Non-target at commanded destination	23 / 224 non-targets
Replay-state string match	169 / 182 answers

Table 2: Command-to-artifact results. Counts pool the 14 completed intervention runs; four scenario templates contribute two runs. JSD is the mean over four pairs of final agent-location distributions.

AI Agent Monitoring

AI agent monitoring tracks actions, tool use, and execution traces to detect failures, misuse, drift, and policy violations during operation.

See all 285 papers

Computational Multi-Agents Society Experiments: Social Modeling Framework Based on Generative Agents
Aug 24, 2025Hanzhong Zhang, Muhua Huang, Jindong WangComputational Social Science Agent-Based Modeling
William & Mary, Williamsburg, VA, USA · Stanford University, Stanford, CA, USA
What Could the Agent See at 19:05? Generating Temporal Enterprise Scenarios from Real Research and Replaying Them to Evaluate Agents
Aug 2, 2026Tezan Sahu, Himani AroraAI Agent Evaluation AI Agent Benchmarks
Microsoft Hyderabad, India
AgentSociety 2: An Integrated Research Environment for Executable Social Science
Jun 11, 2026Jinghua Piao, Jun Zhang, Haoyu Huang +16Multi-Agent LLM Systems AI Agents for Scientific Discovery
Department of Electronic Engineering, Tsinghua University · School of Environment, Tsinghua University · School of Public Policy and Management, Tsinghua University +2

Aug 24, 2025J/K move · Enter open · S save

Computational Multi-Agents Society Experiments: Social Modeling Framework Based on Generative Agents

Hanzhong Zhang, Muhua Huang, Jindong Wang

William & Mary, Williamsburg, VA, USA · Stanford University, Stanford, CA, USA

Computational Social Science Agent-Based Modeling Large Language Model-Based Social Simulation Human Behavior Simulation

Open paper AskPDF arXiv

Computational social science has developed descriptive, interpretative, and predictive modeling, but comprehensive modeling combining in-world human intervention with outcome forecasting remains underexplored. Generative agent-based platforms scale populations or reproduce complex scenarios, but researchers typically act through setup, external controls, or post-hoc analysis rather than in-world roles. We introduce Computational Multi-Agent Society Experiments (CMASE), a generative agent-based framework for in-world researcher participation and traceable interventions at runtime. Treating simulation as a simulated ethnographic field, CMASE supports (1) real-time human-computer interaction to study intervention processes; (2) controlled experiments with ethnography-inspired observation, interviewing, and interpretation; and (3) forward exploration of intervention trajectories and downstream responses. CMASE combines LLM-driven perception, memory, and emotion; a composable environment built from texture and area maps; and researcher roles spanning observation, interviewing, participation, and intervention, so that interventions enter as social events rather than exogenous parameters. We replicate two published field studies from different disciplines: urban greenery and social fragmentation, and cooperation in a public goods game, using two substantially different language models. CMASE reproduces reported statistical patterns and provides process-level records for mechanism-oriented interpretation. A controlled comparison shows that embedded social interventions leave observable, auditable processes that parametric manipulation does not. These findings highlight CMASE's methodological value for studying interventions in small- and medium-scale social settings and its interdisciplinary potential. The official code is available at: https://github.com/armihia/CMASE.

Abstract

AI Agent Monitoring

Explore similar work

Computational Multi-Agents Society Experiments: Social Modeling Framework Based on Generative Agents

What Could the Agent See at 19:05? Generating Temporal Enterprise Scenarios from Real Research and Replaying Them to Evaluate Agents

AgentSociety 2: An Integrated Research Environment for Executable Social Science