How to make better decisions by dismantling hidden cognitive biases

Executive overview

Most decisions fail not because of bad logic, but because of bad inputs — beliefs shaped by cognitive bias, luck misread as skill, and information filtered through ego. Annie Duke's two books map this problem: Thinking in Bets diagnoses how luck and uncertainty corrupt our judgment; How to Decide gives practical tools to improve the information quality that drives every decision.

The core fix is to separate decision quality from outcome quality — and then build processes that let you actually see the difference.

You cannot control luck, but you can control the quality of the information you put into a decision.

Why outcomes mislead us

• Resulting: judging decision quality by outcome quality — Pete Carroll's Super Bowl pass was mathematically sound; the interception was a bad outcome, not a bad call
• Self-serving bias: attributing good outcomes to our own skill, bad outcomes to external forces
• When thinking about others' decisions we do resulting; when thinking about our own, we do self-serving bias
• Cognitive biases — confirmation bias, availability bias, overconfidence, hindsight bias — thrive precisely because there is so much uncertainty; remove uncertainty and the bias has nowhere to hide
• In single-car accidents, close to 40% of drivers report the crash was not their fault — the bias to protect self-image is that strong
• Hindsight bias: we cannot accurately recall what we knew at the time of a decision, so retrospective clarity is an illusion

The information quality problem

• Luck is uncontrollable; information quality is not — improving inputs is the lever available to every decision-maker
• Amateur poker players make decisions with less information than professionals — not because the information doesn't exist, but because they don't know how to see or process it
• Beliefs ≠ truth: most reasoning is done to confirm existing beliefs, not to discover what is actually true
• Junk inputs produce bad decisions regardless of how good the decision process is
• The goal: reduce the chance you will later say "if I only knew then what I know now"

Tracking knowledge to close feedback loops

• Write down what you knew at the time of a decision, what you predicted, and your rationale — before the outcome is known
• This creates a record that makes it possible to separate good process from lucky results
• Example: an investor who bought Zoom pre-pandemic for fundamentals-based reasons deserves less credit than they think — the pandemic outcome was not in their prediction set
• Prospective tracking lets you check whether the world unfolded the way you expected, independently of whether the outcome was good
• Good process applied consistently will produce good outcomes over time; without a record, you cannot tell when you got lucky

Grit vs. quit: a calibration problem

• Perseverance and quitting are complements, not opposites — one cannot exist without the other
• Staying on a losing course to avoid admitting failure is using ego for harm; outside observers can see this clearly; we cannot when it is us
• Opportunity cost is chronically neglected — saying yes to one path means saying no to everything else
• Omission/commission bias and status quo bias cause us to overweight what we are already doing vs. alternatives
• Sunk cost fallacy, identity-protective cognition, and loss aversion all make it harder to pivot
• The landscape metaphor: you need to see the whole terrain to know whether the hill you are climbing is worth climbing — or whether there is a bigger mountain nearby
• Calibration, not conviction, is the goal — same principle applies to confidence: underconfidence is as damaging as overconfidence

Making faster decisions through process

• Building a robust deliberate process first makes fast decisions possible later — you cannot skip the slow to get to the fast
• Two-way door decisions (easily reversible, low impact) can be made quickly; irreversible or high-impact decisions cannot
• Knowing what a full deliberate process looks like lets you decide when to abbreviate it
• "Move fast and break things" applies selectively — not to signing a two-year lease or releasing a large undifferentiated batch to all customers

Entry point: asynchronous independent elicitation

The single highest-leverage change for teams is shifting from in-room group discussion to asynchronous independent collection of opinions before any meeting.

Why group meetings fail at information discovery:

• First speaker anchors the conversation; subsequent speakers are influenced by what has already been said
• Status, seniority, and extroversion determine who is heard
• Persuasion happens in real time — innovative ideas get suppressed before they surface
• Consensus is manufactured, not discovered; outlier views feel socially risky

How to run asynchronous elicitation:

• Define the decision and the feedback you need before contacting the team
• Distribute the same facts and questions to each team member independently
• Require specific answers: numeric estimates with upper and lower bounds, ratings on defined scales — not "pretty quick" or "good market"
• Require a brief rationale (2–3 sentences) for each opinion
• Collect responses in a way that hides individual answers from other participants until all have responded (email without reply-all, a shared form, Airtable, Coda, Google Sheets)
• Collate and share all responses before the meeting

What this produces:

• Full information from every team member, uncontaminated by others' views
• Visibility into where real disagreement exists — the meeting focuses on dispersion, not areas of consensus
• Psychological safety to hold non-consensus opinions — everyone can see that different views are normal
• A written record of beliefs and predictions at the time of the decision, usable for future debriasing
• Shorter, more efficient meetings focused on what actually matters
• Clarity about what information the decision requires (you must define this in order to design the elicitation)