IX-SILA — Social Impact Leverage Algorithm

What This Is

A 9-variable scoring engine for evaluating initiatives — "should we build this" comparisons, not individual coding-task risk (that's /alignment-harness:governer's job). The 9 variables are the same ones the governer accepts via alignment-harness score --sila v1,...,v9 — this skill is the initiative-scoring front door to that same engine, walking you through judging each variable with evidence before you submit.

The weak-link idea, which is the point of this whole method: a weak score in any variable caps the whole initiative, no matter how strong the others look. The weakest variable is always the most important finding — not the total score.

The Formula — two numbers, two different jobs

There are two ways to combine the nine 1-10 scores, and they answer different questions. Don't confuse them.

1. The submittable score (0-100): geometric mean. This is what alignment-harness score --sila v1,v2,v3,v4,v5,v6,v7,v8,v9 computes — the same engine the governer uses for task risk. Nine scores of 8 produce 80. Use this number wherever you need a score that's comparable across initiatives on a normal 0-100 scale, and wherever you'd hand a number to a person or another tool.

2. The internal diagnostic (unbounded): straight multiplication. For finding the weakest link and doing sensitivity analysis (below), multiply the nine raw scores straight through instead of averaging — this is what correctly proves that fixing the single weakest variable gives more benefit than nudging a variable that's already strong. This number gets huge fast (nine 8s = 8⁹ ≈ 134 million) and is never the score you submit or report as "the" score — it exists only to do the weak-link math in Step 4-5 below.

The 9 Variables (Evaluate IN THIS ORDER)

General Variables (1-6)

#	Variable	Key Question	10 =	1 =
1	Value/Effort Ratio	How much value vs cost to build?	Costs nothing, massive value	Costs as much as value delivered
2	User Pain Severity	Are users already in pain?	Daily visible crisis, no pitch needed	Requires awareness campaign
3	Actor Readiness	Is someone positioned to execute?	Already doing adjacent work	Must recruit from scratch
4	Platform Readiness	Do systems already exist?	All APIs/models/UI exist	Must build every piece
5	Loop Closure	Does output feed back into input?	Each cycle makes next more likely	Every cycle needs external energy
6	Compounding Rate	Gets better over time on its own?	Accelerates without investment	Flatlines without constant effort

Your Product's Own Variables (7-9)

These three are meant to be tuned to whatever your product actually is — the questions below are a starting template, not fixed:

#	Variable	Key Question	10 =	1 =
7	AI Delegability	Can agents execute autonomously?	Fully end-to-end autonomous	Requires constant human oversight
8	Bottleneck Impact	Targets your most limiting metric right now?	Directly addresses your biggest inhibitor	Targets a metric already performing well
9	Reach Multiplier	Scales toward your growth ceiling?	Each user served makes the next cheaper	Each new user costs the same

Variable 8 needs you to know your own current bottleneck (churn, activation, conversion, whatever it actually is right now) — ask the person if it isn't already established, rather than guessing. Variable 9 needs your own sense of what "reach" means for this product. Without an answer to either, score the other seven, report 8 and 9 as unset, and mark the result partial rather than inventing a number.

Benchmarks

Raw scores mean little without something to compare against.

Example use of this method (opt-in — replace with your own): a "Curitiba" reference initiative scored elsewhere came out to 583,200 on the straight-multiply SILA-6 diagnostic, and a perfect 10×10×10 on the product-specific three gives 1,000, for a combined diagnostic ceiling of about 583 million.

For your own use: once you've scored more than a couple of initiatives, use your own highest-scoring one as the benchmark instead — "this new idea is at 40% of our best one so far" is more useful than comparing to someone else's unrelated program.

Step-by-Step: Score an Initiative

Step 1: Gather Context

Before scoring, understand the initiative deeply. If you have institutional-memory search or an intent database configured (see /alignment-harness:harness-setup), search it for related past decisions and prior scores. Otherwise, search the project's own docs, past sessions, and git log. Confirm the current bottleneck metric (variable 8) with the person if it isn't already established.

Step 2: Score Each Variable (IN ORDER)

The order matters — each step informs the next.

For each of the 9 variables:

1. Score 1-10
2. Write a one-sentence justification that is specific and falsifiable
   • BAD: "Pretty good infrastructure"
   • GOOD: "Platform exists with 500 active users but no enrollment flow"

Step 3: Compute Both Numbers

submittable_score  = geometric_mean(v1..v9) * 10     # via: alignment-harness score --sila v1,...,v9
diagnostic_product = v1 * v2 * v3 * v4 * v5 * v6 * v7 * v8 * v9   # internal only, for steps 4-5

Step 4: Identify Weakest Variable

The weakest variable is the MOST IMPORTANT finding. Not the score itself.

Step 5: Sensitivity Analysis

Using the diagnostic product from Step 3, for each variable compute: total_if_improved = (diagnostic_product / current_score) * min(10, current_score + 1). Sort by gain from +1. The top row is always the weakest variable — this proves mathematically that fixing your weakest link gives the highest ROI.

Step 6: Design Recommendation

Recommend one specific change to raise the weakest variable by at least 2 points.

Step 7: Benchmark Comparison

Compare the diagnostic product to your own running-best initiative (or the labelled example above, if you have nothing else yet), and report the submittable 0-100 score on its own scale.

Output Format

{
  "idea_name": "string",
  "description": "One sentence",
  "variables": {
    "value_effort_ratio": { "score": N, "justification": "..." },
    "user_pain_severity": { "score": N, "justification": "..." },
    "actor_readiness": { "score": N, "justification": "..." },
    "platform_readiness": { "score": N, "justification": "..." },
    "loop_closure": { "score": N, "justification": "..." },
    "compounding_rate": { "score": N, "justification": "..." },
    "ai_delegability": { "score": N, "justification": "..." },
    "bottleneck_impact": { "score": N, "justification": "..." },
    "reach_multiplier": { "score": N, "justification": "..." }
  },
  "submittable_score_0_100": N,
  "diagnostic_product": N,
  "weakest_variable": "string",
  "design_recommendation": "What to fix first and how",
  "benchmark_comparison": "string"
}

Predicting Higher-Leverage Actions

After scoring multiple ideas, synthesize a NOVEL idea by:

1. Identify systemic weakness — which variable is consistently low across all ideas?
2. Find complementary strengths — which ideas score highest on different variables?
3. Fuse — combine the high-scoring elements while patching the systemic weakness
4. Score the synthesis — run it through the same 9 variables to verify it actually scores higher
5. Submit — record it wherever your project tracks leverage ideas (see below)

Recording Scored Initiatives

Submit the score with alignment-harness score --task "<initiative name>" --sila v1,...,v9 so it's recorded the same way the governer records task scores. If your project has its own leverage-ideas tracker (a database, an admin UI), file it there too, tagged as scored by this method, with the submittable 0-100 score (never the diagnostic product) in whatever field expects a comparable score. If you don't have one yet, alignment-harness records leverage-ideas prints a local folder — write a short markdown file there with the full output above.

If the initiative addresses specific UX intents and you're running /alignment-harness:intent-db, link the intent slugs in your record.

Internal leverage vs. IX-SILA — use both

• Internal leverage (impact ÷ effort, however you track it) answers "what to do this week" — quick wins for velocity.
• This 9-variable score answers "what builds durable value over the long run" — systems thinking for direction.

Use both together rather than picking one.

Agent Prompt Template

Copy this to have any AI agent score an initiative:

You are evaluating an initiative using a 9-variable multiplicative scoring model.

Score each variable 1-10 with a one-sentence justification. Evaluate IN ORDER:

General 6:
1. VALUE/EFFORT RATIO
2. USER PAIN SEVERITY
3. ACTOR READINESS
4. PLATFORM READINESS
5. LOOP CLOSURE
6. COMPOUNDING RATE

Product-specific 3:
7. AI DELEGABILITY
8. BOTTLENECK IMPACT (current bottleneck: [ASK IF UNKNOWN])
9. REACH MULTIPLIER (what "reach" means for this product: [ASK IF UNKNOWN])

Compute the submittable score as geometric_mean(all 9) * 10 (0-100 scale).
Separately, multiply all 9 straight through as an internal diagnostic to find the
weakest variable and rank sensitivity — never report this number as "the score."
Identify weakest variable. Recommend one fix to raise it by 2+ points.

The initiative:
[INSERT HERE]

Key Principles

1. Never average for the weak-link diagnostic. Multiply. One weak link kills everything — that's what the diagnostic product is for.
2. Only the geometric-mean (0-100) number is ever submitted or reported as "the score." The diagnostic product is internal only.
3. The weakest variable is the most important finding. Not the score itself.
4. Justifications must be specific and falsifiable. Not "pretty good" — cite evidence.
5. The sequence builds understanding. Don't jump ahead. Each variable contextualizes the next.
6. Always benchmark — against your own best-scored initiative once you have one.
7. Two scoring systems coexist. Internal leverage for velocity, this 9-variable score for direction.