1. Input from the Conscience
The Ethical Ledger (L_t)
2. The Spirit Integrates the Ledger
The Spirit (S)
Calculate Spirit Score (S_t)
Creates a single score for the turn's overall alignment.
S_t = σ( Σ [ w_i * s_i * φ(c_i) ] )
Update Long-Term Memory (\mu_t)
Blends the current result with past performance to evolve.
μ_t = (β*μ_{t-1}) + ((1-β)*p_t)
Measure Ethical Drift (d_t)
Calculates how "out of character" the response was.
d_t = 1 - cos_sim(p_t, μ_{t-1})
3. The Outputs: Memory and Insights
Spirit Score
9/10
A single metric of alignment for the turn.
Ethical Drift
0.15
A measure of how unusual the response was.
Updated Memory (\mu_t)
The new, evolved state of SAFi's ethical memory.
4. Closing the Loop
The updated memory (\mu_t) is used to generate natural-language feedback, providing personalized coaching to the **Intellect** for the next turn. This is how SAFi learns and self-corrects over time.