Repository is archived
No release in over 3 years
Models attention as a scarce resource for brain-modeled agentic AI; allocates a limited attention budget across competing demands using economic principles
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
 Dependencies

Development

 Project Readme

lex-attention-economy

Models attention as a scarce resource for brain-modeled agentic AI; allocates a limited attention budget across competing demands using economic principles.

What It Does

Treats cognitive attention as a finite budget (default 1.0) that must be rationed across competing demands. Demands have a priority, a cost, and a return-on-investment score. The agent allocates budget to high-priority, high-ROI demands and can rebalance when over-committed. Budget recovers gradually over time. This models the well-documented human limitation that sustained cognitive engagement depletes attention resources.

Core Concept: Attention Budget

# Budget starts at 1.0, demands consume it
add_demand(label: :security_audit, demand_type: :cognitive, priority: 0.9, cost: 0.3, roi: 0.8)
add_demand(label: :log_monitoring, demand_type: :maintenance, priority: 0.3, cost: 0.1, roi: 0.4)

# Allocate budget to demands
allocate_demand(demand_id: security_audit_id)  # spends 0.3
allocate_demand(demand_id: log_monitoring_id)  # spends 0.1
# remaining budget: 0.6

# Recover some budget (e.g., after a rest period)
recover_budget  # adds 0.05

Usage

client = Legion::Extensions::AttentionEconomy::Client.new

# See where attention should go
client.best_roi_demands(limit: 3)
# => { demands: [{ label: :security_audit, roi: 0.8, priority: 0.9 }], count: 1 }

# Free up budget by dropping low-ROI allocations
client.rebalance_budget

# Check utilization
client.attention_status
# => { budget: 0.9, spent: 0.4, remaining: 0.5, utilization: 0.44 }

Integration

Combine with lex-attention's signal filtering: filtered signals generate demands, and budget allocation determines which signals receive full cognitive processing. Wire into lex-tick to model the finite cognitive resources available per processing cycle.

Development

bundle install
bundle exec rspec
bundle exec rubocop

License

MIT