discernment

name

discernment

description

Apply a structured judgment and discernment framework to any high-stakes decision, recommendation, or AI-generated output. Use this skill whenever the user wants to think more carefully before committing to something — a people decision, a strategic call, a piece of writing they're about to send, or an AI output they're not sure whether to trust. Trigger on phrases like "is this right?", "am I confident about this?", "help me think this through", "run the discernment framework", "judgment check", "calibration check", "am I being overconfident", "should I trust this output", "premortem", or any time someone is wrestling with whether their thinking is sound. Also trigger proactively when someone appears to be accepting a claim, recommendation, or AI-generated output without questioning it — especially on high-stakes topics like hiring, restructuring, or communications that will reach many people.

Discernment Framework

Good judgment is not about confidence. It is about calibration — knowing how much your confidence is warranted by the evidence, your actual expertise, and the quality of your thinking process. This framework draws on four bodies of work to help you assess that.

Use this skill by understanding the situation first, defining the root problem using SCQuARE, then walking through the four lenses in order, and synthesising into a clear verdict with a specific next action.

---

Step 0: Understand the situation

Before applying any lens, establish:

• What exactly is being assessed? A decision? A recommendation? A piece of AI-generated text? A course of action someone is about to commit to?
• What are the stakes? Reversible or irreversible? Who is affected? What is the cost of being wrong?
• What has already been decided vs. what is still open? This shapes whether the work is pre-decision quality-checking or post-decision review.

Ask the user for this context if it isn't clear. Don't start the framework on thin air.

---

Step 1: Root problem — SCQuARE

Before applying any judgment lens, establish what problem is actually being solved. This step is easy to skip and almost always worth doing. It is based on the SCQuARE framework (Situation, Complication, Question, Answer, Rationale, Evidence), which forces crisp problem definition before analysis begins.

The risk without this step: excellent thinking applied to the wrong question.

Work through these five elements:

• Situation — What is the stable, agreed-upon context? What do we know to be true and uncontroversial? (2–3 sentences max. If this is hard to state, the problem isn't well-formed yet.)

• Complication — What has changed, or what tension exists, that makes the situation insufficient? This is the reason the decision or question has arisen at all. A good complication is specific — not "things are uncertain" but what specifically has created the need to act or decide.

• Question — Given the situation and complication, what is the precise question that needs answering? This should be a single, crisp question. If there are multiple questions, the scope is too large — name the primary one.

• Answer — What is the hypothesis or proposed answer to that question? This is what is about to be assessed, acted on, or committed to. Making it explicit is the whole point: you cannot pressure-test a vague intent.

• Rationale and Evidence — Why do we believe this answer? What evidence supports it? And what evidence against it has been considered? If the rationale is thin, the rest of the discernment process will be doing more work.

The diagnostic question: Is the question we are actually trying to answer the right question — or is it a symptom of a deeper or different problem?

If the SCQuARE step surfaces a different root question than the one originally presented, name it explicitly and confirm with the user before proceeding. Applying the four lenses to the wrong question is a sophisticated way to get the wrong answer.

---

The Four Lenses

Lens 1 — Dunning-Kruger: Do you know what you don't know?

The Dunning-Kruger effect is not just about being wrong. It is about lacking the metacognitive ability to recognise when you are wrong — which means the most dangerous moment is not when you know nothing, but when you know just enough to feel certain.

Apply this lens by exploring:

• What is the person's actual depth of expertise in this specific domain? Not adjacent experience — this domain.
• Is their confidence rising with familiarity, or with evidence? Familiarity-driven confidence is a warning sign.
• Are there signals of "peak of Mount Stupid" thinking? Easy answers, low uncertainty, dismissal of nuance.
• What would a genuine expert in this area say that this person has not considered?

The question to put to the user: If you found out you were wrong about this, what would most likely be the reason? If they struggle to answer that, the Dunning-Kruger risk is real.

---

Lens 2 — Hogan Judgment: What is your decision-making style and where does it distort?

Hogan's Judgment model identifies three personality dimensions that shape how people approach decisions — and each creates predictable blind spots under pressure:

Dimension	Poles	Blind spot risk
Motivation	Threat-avoidant ↔ Reward-seeking	Reward-seekers underweight downside risk. Threat-avoiders underweight opportunity.
Thinking style	Tactical ↔ Strategic	Tactical thinkers over-optimise the immediate. Strategic thinkers lose the operational detail.
Information processing	Intuitive ↔ Data-driven	Intuitive thinkers move fast on pattern-matching. Data-driven thinkers can stall or over-engineer.

Critically, Hogan adds a fourth dimension most models miss: how a person responds to feedback after a decision goes wrong. Leaders who cannot update on negative feedback repeat the same mistakes.

Apply this lens by asking:

• Which pole of each dimension does this person (or this reasoning) tend toward?
• Given those tendencies, what is likely being underweighted in this decision?
• If this decision turned out badly, what would they do with that information? Would they update, or explain it away?

---

Lens 3 — Mark Egan / Behavioural Insights Team: Are you well-calibrated?

Dr. Mark Egan's research at BIT shows that overconfidence is the dominant bias — 80%+ of UK adults are more confident than they are correct. The goal is not to become less confident; it is to become well-calibrated, meaning your confidence reliably tracks your accuracy.

The BIT framework distinguishes three states:

• Overconfident: confidence exceeds correctness — the most common and dangerous state
• Underconfident: correctness exceeds confidence — often in experienced professionals with impostor syndrome
• Well-calibrated: confidence and accuracy track each other — the goal

Key debiasing tools from Egan's "good form for thinking" toolkit:

Premortem: Before committing, imagine it is 12 months later and the decision has failed. What went wrong? This surfaces risks that forward-looking optimism hides.

Thinkgroup: Deliberately include a dissenting voice before the decision is finalised. Not a devil's advocate for show — someone who genuinely sees it differently.

Confidence diagnostic: Ask — on similar decisions in the past, how often was I right when I felt this certain? If you don't track that, your confidence numbers are noise.

"Cognitive hygiene" check: Like eyesight, judgment degrades without regular calibration checks. Ask: when did I last get feedback that updated my thinking in this domain?

Apply this lens by asking:

• How confident is the person, on a scale of 0–100%?
• What is their actual track record on similar judgments?
• Has a premortem been run? Has a thinkgroup been consulted?
• Is there evidence of "loud certainty" overriding calibrated thinking?

---

Lens 4 — Tetlock / Superforecasting: Is the reasoning structured and updatable?

Tetlock's research on superforecasters shows that well-calibrated judgment is learnable. Superforecasters are not smarter — they have better process. The key habits:

Outside view first: Before reasoning from inside the specific situation, ask — what is the base rate? Of all similar decisions/situations/predictions, how often does it go this way? This counteracts the inside-view bias where our specific situation always seems exceptional.

Precise probabilities, not vague language: "It's likely" is not a forecast. "I'd put this at 70%" is. Precision forces honesty and makes errors detectable.

What would change your mind? A well-reasoned position has explicit conditions under which it would shift. If there is nothing that would change the person's view, it is not a belief — it is an identity.

Update proportionally to evidence: New information should move the needle in proportion to its actual strength. Overreacting and underreacting are both errors.

Track your predictions: Confidence calibration only improves if you score yourself. The question is not just "was I right?" but "how right was I relative to how sure I was?"

Apply this lens by asking:

• What is the outside-view base rate for this situation?
• What specific evidence would raise or lower confidence in this view?
• Has the person stated a precise confidence level, or used vague language?
• Is there a mechanism to check back and score this decision against the outcome?

---

Step 5: Synthesise

After walking through the four lenses, produce a judgment verdict that covers:

1. Calibration verdict: Is the confidence level warranted? State whether the reasoning is well-calibrated, overconfident, or underconfident — and which lens most clearly reveals the risk.

2. The dominant bias risk: Name the one cognitive pattern most likely to undermine this decision. Be specific — not "confirmation bias" as a catch-all, but why this specific bias matters in this specific context.

3. The quality of the reasoning process: Not whether the conclusion is right — that's often unknowable — but whether the process is sound. Has the outside view been taken? Has a dissenting voice been heard? Has the premortem been run?

4. One concrete next action: Before the person commits, what is the single most valuable thing they should do? Options include: run a premortem, seek a thinkgroup, assign a specific probability and commit to tracking it, identify what would change their mind, or simply wait 48 hours.

---

Tone and approach

This framework is not a checklist to be pedantically completed. It is a thinking tool. Use judgment about which lenses are most relevant to the specific situation — sometimes one lens is doing all the work, sometimes all four.

The goal is not to make the person less confident. It is to make them appropriately confident — someone who acts decisively when the evidence warrants it and holds back when it doesn't.

Avoid abstract diagnosis. Name the specific thing. Be direct about the risk. The value of good discernment is that it can be spoken plainly.

---

Quick reference: Five questions

If the context is time-pressured, cut to these five questions:

0. SCQuARE: What is the precise question we are trying to answer — and is that actually the right question?
1. Dunning-Kruger: If you turned out to be wrong, what would be the most likely reason?
2. Hogan: What does your natural decision style tend to underweight — and is that happening here?
3. Egan / BIT: How often have you been right when you felt this certain? Have you run a premortem?
4. Tetlock: What is the base rate? What specific evidence would change your view?

---

Reference material

See references/frameworks.md for deeper notes on each of the four source frameworks, including key publications and direct quotes.