Incentive Design

Incentive Design

Overview

Behavior follows incentives more reliably than character, intent, or training. Get the incentives right and mediocre operators produce excellent results; get them wrong and talented teams produce dysfunction. This is Charlie Munger's "Reward and Punishment Superresponse Tendency" — his first and most important of 25 psychological tendencies (1995 Harvard Law School lecture). The operational question: when behavior is undesirable, ask "what incentive makes this rational?" before asking "what's wrong with these people?"

Composes with principal-agent, goodharts-law, signaling-games, okr-goal-setting, prisoners-dilemma.

When to Use

• Designing compensation, bonuses, commissions, OKRs, or performance management
• Diagnosing why a team is producing undesirable behavior despite training or management
• Drafting contracts, regulations, or platform rules where behavior must be shaped
• Evaluating an existing system for hidden perverse incentives
• Designing reward signals or pricing in AI-native products (RLHF/reward hacking, usage-based vs outcome-based pricing, scarce AI-talent comp amid heavy AI capex and fast AI adoption)

Not when: clearly individual misconduct unrelated to systemic incentives; using incentive framing to excuse deliberate bad-faith behavior.

Coaching Novices (Adaptive Front Door)

• Engine mode: user has a concrete incentive design challenge → run The Process directly.
• Coach mode: user is new to the framework → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. One-line: rational actors produce the behavior incentives favor, regardless of stated intent — check incentives before character.
2. Check fit. If the behavior is clearly individual misconduct, this framework adds less value. Otherwise, apply.
3. Elicit the goal and current incentives. What behavior do you want? What incentives exist now? What are those incentives producing?

[WAIT — do not advance until user responds]

1. Diagnose and design. What behavior do current incentives make rational? Where's the gap? What would make goal-behavior the rational choice?

[WAIT — do not advance until user responds]

1. Close: redesigned structure + gaming countermeasures + monitoring plan.

[WAIT — do not advance until user responds]

The Process

Step 1 — Goal and actors: desired outcome · required behavior · actors · time horizon. Step 2 — Map current incentives: rewards (financial, status, autonomy) · penalties · timing · observability. Step 3 — Diagnose alignment gap: what behavior do current incentives rationally produce? where's the mismatch (metric, magnitude, timing)? Step 4 — Design new structure (7-item checklist): (1) alignment (2) measurability (3) timing (4) threshold structure (5) anti-gaming predictions (6) long-short balance (7) tampering defense. Step 5 — Anticipate Goodhart's Law: whatever you incentivize will be optimized — map the most-likely gaming pattern and close it. Step 6 — Implement and monitor: pilot first · monitor 3-6 months · review cycle every 6-12 months · build in actor feedback.

Output Template

Incentive Design: 
Goal/actors | Current incentives | Alignment diagnosis
Redesign (7-item) | Pilot scope / Monitoring / Review cycle

→ Method in Action: Munger 1995 + FedEx + Modern Applications → 2026 lens: Incentive Design in the AI Economy — RLHF reward design, scarce-talent comp, usage-based pricing (2024–2026)

Pack: Application Patterns

Applying It Well

• Look at the incentives before you look at the people
• Incentive-caused bias is invisible to the actor — they sincerely believe they are doing the right thing
• The most important incentive design is at the founding moment; embedded structures become hard to change
• "Roughly right" incentives often produce dramatically wrong behavior at scale — precision matters

→ Primary sources: references/sources.md

Common Rationalizations

[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.

Red Flags

• Behavior attributed to character/capability rather than incentives
• New people hired or fired without examining the incentive structure
• Compensation or KPI system not reviewed in 2+ years
• Previously-functioning incentive system starting to produce gaming
• Anti-gaming countermeasures absent from a metric-driven system

Verification

• Goal-behavior explicitly specified
• Current incentives mapped
• Alignment gap diagnosed
• Redesigned incentives address the alignment
• 7-item checklist applied
• Goodhart-style gaming anticipated
• Monitoring and review cycle in place

---

Part of deciqAI Knowledge Skills — 227 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/c/incentive-design · ⭐ Star the repo → https://github.com/deciqAI/knowledge-skills · Contributions welcome.

Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/incentive-design.json