Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Apply the Elaboration Likelihood Model to design persuasion strategies by matching message type to audience elaboration level. Use this skill when the user needs to craft persuasive communications, choose between argument-based and cue-based messaging, predict attitude durability after a campaign, or when they ask 'why did the message fail to persuade', 'how to change resistant attitudes', or 'should we use data or endorsements'.
.claude/skills/asgard-ai-platform-grad-elm/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 61% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 25% | 0% |
The Elaboration Likelihood Model (Petty & Cacioppo, 1986) proposes that persuasion occurs through two distinct routes depending on the audience's motivation and ability to process a message. The central route relies on careful evaluation of argument quality, producing durable attitude change. The peripheral route relies on heuristic cues (source attractiveness, number of arguments), producing temporary and fragile attitude shifts.
IRON LAW: Attitude change via the central route is MORE durable
and resistant to counter-persuasion — but requires motivation
AND ability to process. When either is absent, only the
peripheral route is available, and its effects decay.Key assumptions:
| Factor | High Elaboration | Low Elaboration | |--------|-----------------|-----------------| | Personal relevance | High (topic matters to them) | Low (distant from self) | | Need for cognition | High (enjoys thinking) | Low (avoids effortful thought) | | Prior knowledge | Sufficient to evaluate arguments | Insufficient to engage deeply | | Distraction level | Low (can focus) | High (divided attention) | | Time pressure | Low (can deliberate) | High (must decide quickly) |
Central route elements:
Peripheral route elements:
| Route | Durability | Behavior Prediction | Counter-Persuasion Resistance | |-------|-----------|--------------------|-----------------------------| | Central | High | Strong | High | | Peripheral | Low | Weak | Low |
markdown## ELM Persuasion Strategy: [Context] ### Audience Elaboration Assessment | Segment | Motivation | Ability | Elaboration Level | |---------|-----------|---------|-------------------| | [segment] | [High/Low] | [High/Low] | [High/Moderate/Low] | ### Route Selection: [Central / Peripheral / Mixed] ### Message Design - Primary arguments: [if central route] - Peripheral cues: [if peripheral route] - Source selection: [credibility/attractiveness rationale] ### Predicted Outcomes - Attitude durability: [High/Medium/Low] - Behavioral impact: [Strong/Moderate/Weak] - Counter-persuasion vulnerability: [High/Medium/Low] ### Risk Mitigation - [What happens if audience elaboration is misjudged]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 54,796 | 22,385 | -59% | 1 | 1 | 0% | 6,936 | 4,067 | -41% | 0 | 0 | — |
case-02 | fail→pass | 27,603 | 19,068 | -31% | 1 | 1 | 0% | 4,289 | 4,210 | -2% | 0 | 0 | — |
case-03 | fail→pass | 29,199 | 21,934 | -25% | 1 | 1 | 0% | 4,522 | 3,838 | -15% | 0 | 0 | — |
case-04 | fail→pass | 19,611 | 23,465 | +20% | 1 | 1 | 0% | 2,675 | 4,297 | +61% | 0 | 0 | — |
case-05 | pass→pass | 22,928 | 28,191 | +23% | 1 | 1 | 0% | 3,430 | 4,899 | +43% | 0 | 0 | — |
case-06 | fail→pass | 23,050 | 15,373 | -33% | 1 | 1 | 0% | 2,836 | 3,555 | +25% | 0 | 0 | — |
case-07 | pass→pass | 19,918 | 22,030 | +11% | 1 | 1 | 0% | 2,893 | 4,061 | +40% | 0 | 0 | — |
case-08 | pass→pass | 22,134 | 23,578 | +7% | 1 | 1 | 0% | 3,061 | 3,502 | +14% | 0 | 0 | — |
case-09 | pass→pass | 15,484 | 13,614 | -12% | 1 | 1 | 0% | 2,517 | 3,513 | +40% | 0 | 0 | — |
case-10 | fail→pass | 27,989 | 23,345 | -17% | 1 | 1 | 0% | 2,885 | 4,746 | +65% | 0 | 0 | — |
case-11 | fail→pass | 20,602 | 15,384 | -25% | 1 | 1 | 0% | 3,097 | 3,490 | +13% | 0 | 0 | — |
case-12 | fail→pass | 15,956 | 29,309 | +84% | 1 | 1 | 0% | 2,626 | 3,953 | +51% | 0 | 0 | — |
case-13 | pass→pass | 20,287 | 13,225 | -35% | 1 | 1 | 0% | 2,574 | 3,236 | +26% | 0 | 0 | — |
case-14 | fail→pass | 20,944 | 21,933 | +5% | 1 | 1 | 0% | 2,757 | 4,071 | +48% | 0 | 0 | — |
case-15 | fail→fail | 26,769 | 20,104 | -25% | 1 | 1 | 0% | 3,835 | 4,319 | +13% | 0 | 0 | — |
case-16 | fail→fail | 19,780 | 24,397 | +23% | 1 | 1 | 0% | 2,660 | 4,373 | +64% | 0 | 0 | — |
case-17 | pass→pass | 35,191 | 18,935 | -46% | 1 | 1 | 0% | 5,554 | 4,016 | -28% | 0 | 0 | — |
case-18 | fail→pass | 19,174 | 19,937 | +4% | 1 | 1 | 0% | 2,822 | 3,547 | +26% | 0 | 0 | — |
case-19 | fail→pass | 20,947 | 15,961 | -24% | 1 | 1 | 0% | 3,168 | 3,556 | +12% | 0 | 0 | — |
case-20 | pass→pass | 13,847 | 17,308 | +25% | 1 | 1 | 0% | 2,035 | 3,561 | +75% | 0 | 0 | — |
case-21 | pass→pass | 23,085 | 30,499 | +32% | 1 | 1 | 0% | 3,243 | 4,967 | +53% | 0 | 0 | — |
case-22 | pass→pass | 23,553 | 22,172 | -6% | 1 | 1 | 0% | 3,743 | 4,393 | +17% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +50 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.