devils-advocate

Overview

The Devil's Advocate is an intellectually rigorous stress-test designed to dismantle arguments, expose blind spots, and "steel man" opposing viewpoints. This skill replaces vague skepticism with structured adversarial analysis—using inversion, fragility detection, and probabilistic auditing to ensure a proposal can survive an unforgiving market.

Guiding Principles

Principle 1: Invert, Always Invert (Source: Munger, Poor Charlie's Almanack)

To solve a problem or test a plan, think about what would cause the opposite of the desired result. Listing all the ways a plan could fail is more effective than listing why it will succeed. (Source: Farnam Street, "Inversion")

Principle 2: Anti-Resulting (Source: Duke, Thinking in Bets)

Never judge a decision solely by its outcome. A "good" result can come from a "bad" process due to luck. The Devil's Advocate must audit the decision process, not just the projected result.

Principle 3: Fragility Detection (Source: Taleb, Antifragile)

Any proposal that relies on things "staying the same" or has non-linear downside and capped upside is fragile. Strategy should aim for "Optionality"—the ability to benefit from volatility.

Principle 4: Steel-Manning (Source: Paul Graham, "How to Think for Yourself")

Do not create a "straw man" of the opposition. Instead, build the absolute strongest version of the counter-argument. If the original proposal can't survive the "Steel Man," it is fundamentally weak.

Principle 5: The Certainty Audit (Source: Duke, How to Decide)

We tend to believe we "know" things that are only probable. Force a percentage confidence score on every key assertion. If an assertion is "certain," it is likely a blind spot.

When to Use This Skill

• When a strategic proposal has unanimous "groupthink" support.
• When an analysis relies on a single high-impact assumption (a "linchpin").
• Before presenting a high-stakes recommendation to a board or executive leadership.
• When evaluating a complex project with significant "downside" risk.

When NOT to Use This Skill

• During early-stage creative brainstorming where critique might stifle ideation.
• For minor, low-stakes tactical decisions where speed is more important than rigor.

Core Process

Step 1: Identify the Linchpin Assumption

Determine the single weakest foundational assumption in the overall thesis.

1. The "What Must Be True" Test: What is the one thing that, if proven false, collapses the entire argument? (Source: Lafley, Playing to Win)
2. The "Secret" Check: Is the proposal built on a unique truth, or just conventional wisdom? (Source: Thiel, Zero to One)

Step 2: Perform the Inversion Audit

List all the specific ways the proposal could lead to a catastrophic failure. (Source: Munger, Ch. 2)

1. The Pre-Mortem: Imagine it is 12 months from now and the project has failed. What specifically happened? (Source: Duke, How to Decide)
2. Reverse the Goal: What actions would we take if we wanted this project to fail? Are we currently doing any of those things?

Step 3: Detect Systemic Fragility

Evaluate how the proposal handles volatility and "Black Swan" events. (Source: Taleb, Ch. 1)

1. Non-Linear Downside: Is there a "cliff" where a small change in inputs leads to a massive failure?
2. Capped Upside: Is the reward for success small relative to the cost of failure?

Step 4: Steel-Man the Opposition

Construct the absolute strongest argument against the core thesis.

1. Competitor View: How would a smart, resourced competitor invalidate this plan?
2. The Skeptic's Evidence: What data exists that contradicts the author's primary claims?

Step 5: The "So What?" and Sensitivity Check

Challenge the conclusion's impact and robustness.

1. The Needle Check: Even if the author is 100% right, does it move the needle for the business?
2. The 10% Swing: If key metrics (conversion, cost, timing) swing +/- 10%, does the plan still hold together? (Source: Gil, High Growth Handbook)

Frameworks & Models

The Checklist for Misjudgment (Source: Munger, Poor Charlie's Almanack)

Audit the proposal for the 25 standard psychological biases, including:

1. Reward-Superresponse: Are incentives driving bad logic?
2. Incentive-Caused Bias: Does the author benefit from the proposal being accepted, regardless of its truth?
3. Consistency and Commitment: Is the author stuck in a previous decision?

Via Negativa (Source: Taleb, Antifragile)

1. Subtract to Strengthen: What should we stop doing or remove from this proposal to make it more robust?
2. Remove the "I-Knew-It-All-Along" bias: Document current predictions to prevent future "hindsight bias."

Cross-Skill Invocations

• REQUIRED SUB-SKILL: mental-model-library — Use cross-domain models (Inversion, Second-order thinking) for analysis.
• RECOMMENDED SUB-SKILL: assumption-audit — To validate the evidence for key claims.
• RECOMMENDED SUB-SKILL: decision-frameworks — To structure the final choice after the stress test.

Common Mistakes

1. Straw-Manning: Attacking a weak version of the opposing argument to make the original look better. (Source: Graham, "Think")
2. Resulting: Judging a proposal based on a lucky success in a different context. (Source: Duke, Ch. 1)
3. The Lollapalooza Trap: Ignoring how multiple small risks can combine into a massive failure. (Source: Munger, Ch. 3)

Diagnostic Checklist

• Have we identified the "Linchpin" assumption that must be true for success?
• Have we performed a "Pre-Mortem" identifying 3+ specific failure paths?
• Have we "Steel-Manned" the strongest possible counter-argument?
• Is the proposal "Antifragile" (does it benefit or stay robust under volatility)?
• Have we assigned confidence percentages to all "certain" assertions?

Sources

• Duke, Thinking in Bets, Ch. 1, 2, 3 — Anti-Resulting, Certainty Bias, Cuckolding Luck.
• Munger, Poor Charlie's Almanack, Ch. 2, 4 — Inversion, Psychology of Human Misjudgment.
• Taleb, Antifragile, Ch. 1, 22, 24 — Fragility, Via Negativa, Skin in the Game.
• Farnam Street, "Inversion" & "Probabilistic Thinking" — Applied mental models.
• Paul Graham, "How to Think for Yourself" — Independent thinking vs. consensus.