CUBES Simulator Architecture

Summary

CUBES (CUstomer BEhavior Simulator) is a multi-agent simulation platform built on the Swarm engine that simulates populations of consumer agents interacting concurrently in a virtual market with competing brands. The architecture separates social dynamics (imitation, conditioning processes) from reactive modulators (personality traits: mistrust, opportunism, innovativeness) and uses behavioral primitives as the core decision mechanism.

Overview

CUBES was developed as part of the CUBES project at FTR&D / LIP6 to simulate consumer behaviors in competing markets including several brands and virtual populations of several thousand consumers. The software includes a simulation engine, parameterization tools, and observation tools.

Main Content

System Components

The CUBES architecture (Figure 1 in the paper) consists of:

1. Brand agents: Each brand has:

   • A marketing strategy (publicity actions, brand image)
   • The ability to emit stimuli to the consumer population

2. Consumer agents: Each consumer has:

   • A set of behavioral attitudes (BA)
   • A socio-demographic profile
   • Opinions about each brand
   • A decision mechanism based on behavioral primitives

3. Simulation engine: Built on the Swarm simulation engine (http://www.swarm.org), providing concurrent agent execution

Architecture Diagram

The system follows a two-layer architecture:

Social Dynamics Layer:

• Conditioning process: Repeated exposure to stimuli reinforces attitudes
• Imitation process: Agents copy behaviors of connected agents (See Imitation and Conditioning Processes for the formal mechanics.)

Reactive Modulators Layer (Personality Traits):

• Mistrust personality trait: Filters incoming stimuli with skepticism
• Opportunism personality trait: Amplifies deal-seeking behavior
• Innovativeness personality trait: Modulates openness to new products

These two layers interact to produce:

• Instantiated Behavioral Attitudes (from social dynamics)
• Derived Behavioral Attitudes (modified by personality traits)

Simulation Flow

1. Brands emit external stimuli (promotions, rumors, innovations, recommendations)
2. Each stimulus is characterized by type, color (brand identifier), and intensity
3. Consumer agents perceive stimuli through their behavioral primitives (BP)
4. BPs filter stimuli through inhibiting and triggering thresholds
5. Accepted stimuli modify the consumer’s opinions about brands
6. Purchase decisions are made based on accumulated opinions
7. Market-level outcomes (market shares, diffusion curves) are observed

Agent Communication

Each consumer agent has a perception field — a spatial radius limiting communication with other agents. This field is a function of the agent’s socio-demographic profile and behavioral attitudes (particularly innovativeness). The communication gradient models the fact that WOM influence weakens with social distance. This spatial radius acts as a local network topology; see Network Topology Effects on Diffusion and Social Network Formation in Consumer Markets for how network structure shapes diffusion outcomes.

Population Scale

Simulations typically include:

• Several thousand consumer agents (5000-7000 in experiments)
• 3 competing brands
• 60-120 simulation time steps

Connections

• CUBES implements ABM Methodology and Principles with a psychology-inspired architecture
• The behavioral layer is detailed in Behavioral Attitudes in CUBES and Behavioral Primitives and Thresholds
• Population calibration uses Genetic Algorithm Calibration for ABM
• The architecture enables the emergent phenomena observed in Market Share Equilibrium and Lock-In

See Also

• Ben Said et al 2002 - Overview — paper context
• Behavioral Attitudes in CUBES — the five BA types
• Behavioral Primitives and Thresholds — the core decision mechanism
• Imitation and Conditioning Processes — the two social dynamics implemented in CUBES
• Emergent Phenomena in ABM — market share equilibria emerge from individual agent interactions
• Network Topology Effects on Diffusion — how perception field radius shapes information diffusion
• Social Network Formation in Consumer Markets — social distance and WOM propagation