lex-arousal

Yerkes-Dodson arousal regulation for brain-modeled agentic AI.

Models the inverted-U relationship between arousal and performance. An agent that is too calm underperforms due to disengagement; an agent that is too aroused underperforms due to anxiety and cognitive overload. The optimal arousal level depends on task complexity — simple tasks benefit from higher arousal, complex tasks require lower arousal for peak performance.

Performance is computed as a Gaussian centered on the optimal arousal for the given task:

performance = exp(-PERFORMANCE_SENSITIVITY * (arousal - optimal)^2)

Optimal arousal: simple=0.7, moderate=0.5, complex=0.4, extreme=0.3.

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

# Stimulate (e.g., due to an urgent alert)
client.stimulate(amount: 0.3, source: :threat_detected)
# => { arousal: 0.6, label: :high }

# Check if current arousal is optimal for the task
client.check_performance(task_complexity: :complex)
# => { performance: 0.71, arousal: 0.6, optimal_arousal: 0.4 }

# Get actionable guidance
client.arousal_guidance(task_complexity: :complex)
# => { guidance: :throttle, arousal: 0.6, optimal_arousal: 0.4 }

# Calm down
client.calm(amount: 0.15)

# Decay tick (call periodically)
client.update_arousal

Wire arousal_guidance into lex-tick mode selection: :boost suggests moving toward full_active mode; :throttle suggests moving toward dormant. High arousal feeds into lex-emotion's intensity signals.

bundle install
bundle exec rspec
bundle exec rubocop

MIT