Communication Silo Guide

The Communication Silo manages signaling and language evolution: signal repertoires, protocol evolution, coordination messages, and deception detection. Multi-agent systems need communication to coordinate, and evolved communication is more robust than hard-coded protocols.

Overview

Traditional multi-agent systems use fixed communication protocols. The Communication Silo enables evolved communication:

• Signal invention - Agents create new signals with meanings
• Signal adoption - Successful signals spread through population
• Coordination - Agents coordinate through messaging
• Honesty/deception - Honest signaling evolves under pressure
• Dialect formation - Subgroups develop distinct languages

Why Evolved Communication?

Architecture

The Communication Silo uses TWEANN controllers at three levels:

Sensors (10 inputs)

The L0 controller receives 10 communication measurements:

Actuators (8 outputs)

The controller adjusts 8 communication parameters:

Communication Dynamics

Signal Evolution

Signals evolve meanings through invention and adoption:

Vocabulary growth:

%% New signals created per generation
new_signals = vocabulary_growth_rate * population_size
%% Signal stabilizes when adopted by >50% of population

Coordination

Agents coordinate through message exchange:

Coordination reward:

reward = coordination_reward * success_rate * num_participants
%% Cost of coordination:
cost = communication_cost * messages_exchanged

Honesty and Deception

Evolutionary pressure shapes honest signaling:

Equilibrium:

• High lying penalty + high detection = honest population
• Low lying penalty + low detection = deceptive population
• Balance emerges from evolutionary dynamics

Dialect Formation

Subgroups develop distinct languages:

Dialect isolation:

%% Shared vocabulary between dialects
shared_vocab = total_vocab * (1 - dialect_isolation)
%% Higher isolation = more distinct dialects

Integration with Other Silos

Outgoing Signals

Incoming Signals

Signal Examples

%% Receive social network density from social silo
handle_cast({cross_silo, social_network_density, Density}, State) ->
    %% Dense networks increase communication baseline
    NewFrequency = State#state.base_frequency * (1 + Density * 0.5),
    {noreply, State#state{communication_frequency = NewFrequency}};

%% Send coordination capability to task silo
send_cross_silo_signals(SensorValues, State) ->
    PopId = State#communication_state.population_id,
    CoordSuccess = maps:get(coordination_success_rate, SensorValues, 0.5),
    task_silo:receive_signal(PopId, coordination_capability, CoordSuccess).

Events Emitted

Practical Examples

Example 1: Cooperative Task

Configure for high coordination and honest signaling:

Config = #{
    coordination_reward => 0.4,          % High coordination reward
    lying_penalty => 0.4,                % Strong penalty for deception
    deception_detection_bonus => 0.15,   % Reward catching liars
    communication_cost => 0.02,          % Low message cost
    dialect_isolation => 0.1             % Encourage shared language
}.

Expected outcomes:

• High coordination success rate
• Honest population
• Unified vocabulary
• Efficient team actions

Example 2: Competitive Task

Configure for strategic signaling in competition:

Config = #{
    coordination_reward => 0.1,          % Low coordination emphasis
    lying_penalty => 0.1,                % Allow strategic deception
    deception_detection_bonus => 0.2,    % Reward detecting enemy lies
    communication_cost => 0.05,          % Medium message cost
    dialect_isolation => 0.7             % Allow secret codes
}.

Expected outcomes:

• Strategic deception emerges
• Coalition-specific languages
• Arms race between lying and detection
• Information warfare dynamics

Example 3: Mixed Environment

Configure for adaptive communication:

Config = #{
    coordination_reward => 0.25,
    lying_penalty => 0.2,
    deception_detection_bonus => 0.1,
    communication_cost => 0.03,
    dialect_isolation => 0.4,
    vocabulary_growth_rate => 0.03,      % Moderate invention
    message_compression_pressure => 0.15 % Some efficiency pressure
}.

Expected outcomes:

• Context-dependent behavior
• Honest within groups, strategic across groups
• Moderate dialect formation
• Balanced vocabulary growth

Tuning Guide

Trade-offs

Common Issues

Recommended Starting Point

DefaultConfig = #{
    vocabulary_growth_rate => 0.02,
    communication_cost => 0.05,
    lying_penalty => 0.2,
    deception_detection_bonus => 0.1,
    coordination_reward => 0.2,
    message_compression_pressure => 0.1,
    dialect_isolation => 0.3,
    language_mutation_rate => 0.02
}.

Control Loop

The Communication Silo executes per generation:

1. Collect messages/signals - Gather communication data from agents
2. Process signal inventions - Add new signals to registry
3. Track coordination attempts - Record success/failure
4. Record deception events - Track lying and detection
5. Update dialect membership - Manage subgroup languages
6. Apply actuator outputs - Set communication parameters
7. Emit communication events - Notify listeners
8. Send cross-silo signals - Update dependent silos

Configuration Reference

Full Configuration Record

-record(communication_config, {
    %% Enable/disable
    enabled = true :: boolean(),

    %% Vocabulary limits
    max_vocabulary_size = 1000 :: pos_integer(),
    max_message_complexity = 100 :: pos_integer(),

    %% Dialect limits
    max_dialects = 10 :: pos_integer(),

    %% Frequency tracking
    frequency_window_ms = 60000 :: pos_integer(),
    max_frequency = 100 :: pos_integer(),

    %% History limits
    history_limit = 1000 :: pos_integer(),

    %% L2 update interval
    l2_update_interval_ms = 30000 :: pos_integer()
}).

API Functions

%% Start the communication silo
communication_silo:start_link(Config)

%% Record a message
communication_silo:record_message(PopId, Message)

%% Record signal invention
communication_silo:record_signal_invention(PopId, Signal)

%% Record coordination attempt
communication_silo:record_coordination_attempt(PopId, Attempt)

%% Record deception event
communication_silo:record_deception_event(PopId, Event)

%% Get signal registry
Registry = communication_silo:get_signal_registry(PopId)

%% Get dialect info
Dialects = communication_silo:get_dialect_info(PopId)

%% Enable/disable silo
communication_silo:enable(PopId)
communication_silo:disable(PopId)

Biological Inspiration

The Communication Silo mirrors biological communication:

Source Code Reference

Core implementation files:

References

• PLAN_COMMUNICATION_SILO.md - Full specification
• "The Evolution of Language" - Nowak & Krakauer
• "Signals: Evolution, Learning, and Information" - Skyrms
• "Modeling the Cultural Evolution of Language" - Steels

See Also

• Social Silo Guide - Social networks affect communication
• Cultural Silo Guide - Cultural learning and transmission
• Task Silo Guide - Coordination capability for tasks
• Competitive Silo Guide - Strategic signaling in competition