Decision-making under uncertainty is the central problem of practical reasoning. Every significant decision - personal, professional, or organisational - involves acting on incomplete information, competing values, and unpredictable consequences. The question is not how to eliminate uncertainty but how to reason well within it.

This guide covers the core frameworks, cognitive biases, and practical tools that research identifies as most relevant to improving decision quality. It links to more detailed treatments of each concept and includes practical tools - a decision journal template, premortem protocol, and bias checklist - that can be applied immediately.

What Makes a Decision Good or Bad

A good decision is not the same as a good outcome. A decision made with sound reasoning, appropriate information, and honest acknowledgment of uncertainty can still produce a bad outcome - because outcomes involve factors outside the decision-maker's control. Equally, a poor decision made on flawed reasoning can produce a good outcome through luck.

This distinction matters because it defines what you can actually improve. You cannot directly control outcomes. You can improve the quality of your reasoning process. The goal of decision science is to produce better average outcomes over time by systematically improving the process - not to guarantee any individual outcome.

A good decision process typically includes:

  • Clearly defined problem or choice
  • Honest inventory of what is known and unknown
  • Identified options, including doing nothing
  • Explicit consideration of second-order consequences
  • Appropriate calibration of confidence to evidence
  • Acknowledgment of which biases are most likely to be active

The Core Cognitive Biases That Distort Decisions

Cognitive biases are systematic errors in thinking that affect judgments and decisions in predictable ways. They are not random mistakes - they have consistent patterns, consistent triggers, and consistent directions of distortion. Understanding them does not eliminate them, but awareness creates opportunities for correction.

Availability Bias

We judge the probability of events partly by how easily examples come to mind. Events that are vivid, recent, or emotionally significant feel more probable than they statistically are. Flying feels more dangerous than driving because plane crashes are dramatic and heavily covered. The correct comparison is per-mile fatality rates - not ease of recall.

Decision implication: Slow down risk assessments that rely heavily on memorable examples. Ask: what does the base rate show? What would be the representative distribution of outcomes, not the salient ones?

Sunk Cost Fallacy

Past investment - time, money, or effort already spent - should not influence decisions about future action. The resources are gone regardless of what you do next. But humans systematically weight sunk costs in forward-looking decisions, continuing projects, relationships, or investments because of what has already been committed rather than what future returns are likely.

Decision implication: When evaluating whether to continue something, ask: if I were starting fresh today, with full knowledge of current conditions, would I choose this? If no, sunk cost reasoning may be driving the decision.

Anchoring

Initial information disproportionately influences subsequent judgments. In salary negotiations, the first number stated strongly influences the final settlement. In probability estimation, starting from a number and adjusting produces estimates that cluster too close to the anchor. Even arbitrary, irrelevant anchors (like a randomly spun wheel) measurably shift numerical estimates.

Decision implication: Generate your own estimate independently before hearing others. In negotiations, be aware of anchoring effects when counterparties make the first offer. Make explicit adjustments when you identify that an anchor may be distorting your estimate.

Overconfidence

People are systematically overconfident in the accuracy of their predictions, the precision of their knowledge, and the superiority of their judgment relative to others. Calibration studies consistently find that when people say they are "90% confident" they are right, they are actually right roughly 70% of the time. Experts are often overconfident in their areas of expertise.

Decision implication: Widen your confidence intervals. Distinguish between what you know and what you believe. Treat confidence as a quantity to be checked against track record, not a self-report.

Confirmation Bias

We preferentially seek, interpret, and remember information that confirms existing beliefs. When evaluating a new business idea, we tend to find examples of similar businesses that succeeded rather than systematically searching for failure rates. This is not deliberate self-deception - it is a pervasive pattern in how memory retrieval and attention operate.

Decision implication: Actively seek disconfirming evidence. Assign someone the explicit role of devil's advocate. Ask: what would change my mind? If nothing could change your mind, the belief is not responding to evidence.

Probabilistic Thinking: Reasoning With Numbers

Probabilistic thinking is the practice of reasoning in terms of probability distributions rather than point estimates or binary outcomes. Instead of asking "will this succeed?" ask "what is the distribution of likely outcomes, and what does each scenario require?"

Several specific tools support this:

Base Rates

Base rates are the statistical frequency of outcomes across a reference class of similar cases. Before estimating how likely your specific project is to succeed, find the base rate for projects of that type. This is reference class forecasting - anchoring estimates in empirical distributions rather than in the details of the specific case, which tend to produce optimistic biases.

Expected Value

Expected value is probability multiplied by outcome magnitude, summed across scenarios. A 10% chance of winning $1,000 has an expected value of $100. A 90% chance of losing $10 has an expected value of -$9. Comparing expected values across options provides a systematic basis for choices under uncertainty when outcomes are quantifiable.

Expected value calculations are useful tools, not complete decision guides. They assume outcomes are well-quantified, probabilities are reliable estimates, and the decision-maker is risk-neutral. Real decisions often violate one or more of these assumptions - but the discipline of making probabilities and outcomes explicit is valuable even when the arithmetic is approximate.

Second-Order Thinking

Most people reason about immediate first-order effects. Second-order thinking explicitly asks: if this first-order outcome occurs, what happens next? What are the downstream consequences of the response to the first-order effect? Many important failures in business, policy, and personal decisions involve not the first-order effects that were intended but the second-order effects that were not anticipated.

The Premortem: Finding Failure Before It Happens

The premortem is a structured technique developed by psychologist Gary Klein for identifying risks before committing to a course of action. It inverts the usual planning process by imagining that the project has already failed and asking: what went wrong?

The technique works because prospective hindsight - imagining a specific past failure - is psychologically easier than prospective risk identification. It is harder to say "this might fail" before committing than to say "this failed" after imagining the failure. The premortem exploits this asymmetry.

Premortem Protocol

Use this protocol before committing to any significant decision or project:

  1. State the decision clearly. Write down what you are about to do in one or two sentences.
  2. Set the scene. "It is [12 months from now]. We went ahead with this decision. It has failed badly. Not a minor setback - a significant failure. We are sitting here trying to understand what happened."
  3. Generate failure causes independently. Each participant writes their own list of plausible failure causes without discussion. This prevents anchoring on the first idea raised and surfaces independent perspectives.
  4. Share and aggregate. Each person shares one item in rotation until all causes are on the table.
  5. Evaluate and mitigate. For the most plausible and consequential failure modes, ask: is there anything we can do now to reduce this risk? Adjust the plan if so.
  6. Identify the single most likely killer. If one cause stands out as both highly plausible and highly consequential, make it explicit and decide whether it changes the decision.

Decision Journal Template

A decision journal is a record of significant decisions, the reasoning behind them, and the outcomes. It serves two purposes: it forces more explicit reasoning at decision time, and it provides a feedback mechanism for improving calibration over time.

Most decisions go unrecorded. This means most decision-makers never get systematic feedback on the quality of their reasoning - only on outcomes, which are influenced by luck as well as judgment. A journal separates process from outcome, enabling genuine learning.

Decision Journal Entry Template

DECISION JOURNAL - WhenNotesFly Editorial
whennotesfly.com/concepts/decision-making/better-decisions-under-uncertainty

DATE:
DECISION:
   What exactly am I deciding?
   What are the main options (including doing nothing)?

WHAT I KNOW:
   What evidence is most relevant?
   What do I know that argues for each option?
   What do I not know that would change the decision?

WHAT I BELIEVE:
   What is my overall assessment?
   How confident am I? (0-100%)
   What would change my mind?

BIASES TO WATCH:
   Which cognitive biases are most likely to be active here?
   (e.g., sunk cost, availability, overconfidence, confirmation)

SECOND-ORDER EFFECTS:
   If this works as intended, what happens next?
   If this fails, what happens next?

DECISION:
   What am I deciding to do?
   Why this over the alternatives?

EXPECTED OUTCOME:
   What result do I expect in [timeframe]?
   How will I know if this worked?

----
[Fill in after outcome is known]

ACTUAL OUTCOME:
DATE OF OUTCOME:
WAS MY REASONING SOUND, IN HINDSIGHT?
WHAT WOULD I DO DIFFERENTLY?

Reversible vs. Irreversible Decisions

Jeff Bezos's two-way door framework distinguishes between decisions that can be easily undone (Type 2 decisions) and those that are difficult or impossible to reverse (Type 1 decisions). The implication is that different amounts of deliberation are appropriate for each type.

Type 2 decisions - which cover the vast majority of everyday choices - should be made quickly with good-enough information, with the understanding that course correction is available. Over-deliberating reversible decisions wastes time and slows action without improving outcomes.

Type 1 decisions - hiring, major capital commitments, irreversible structural changes, health decisions - warrant more deliberation because the cost of a bad outcome cannot be easily mitigated by subsequent correction. The asymmetry in decision cost justifies asymmetric investment in decision quality.

The practical skill is correctly categorising which type of decision you face. Many people treat Type 2 decisions as if they were Type 1, which produces paralysis and excessive deliberation on choices that do not warrant it.

Mental Models That Improve Decision Framing

Mental models are representations of how things work that inform how we understand situations and evaluate options. Several are especially useful for decision-making:

Inversion: Instead of asking how to achieve a goal, ask what would guarantee failure. Remove the causes of failure as a path to success. Munger's formulation: "Tell me where I'm going to die, so I'll never go there."

Opportunity cost: Every resource - time, money, attention - used for one purpose is unavailable for alternatives. The true cost of any choice includes what you give up by not choosing the alternatives. People systematically underestimate opportunity costs by thinking only about what they are gaining, not what they are forgoing.

First principles: Break a problem down to its fundamental components and reason up from there, rather than by analogy to how things have always been done. This is most useful when analogies are misleading or when genuinely novel solutions are needed.

Regret minimisation: Bezos's framework for long-term decisions: imagine yourself at 80, looking back. Which choice would you most regret? This shifts the time horizon away from immediate discomfort toward durable values, which is useful for decisions about risk-taking, career changes, and major life choices.

Decision Fatigue: When the Brain Runs Low

Decision quality degrades as cognitive resources are depleted by repeated decisions. This is decision fatigue - a documented phenomenon showing that judges issue harsher sentences later in the day, physicians order more unnecessary tests late in a shift, and people make worse consumer decisions after making many prior choices.

The implication is not that willpower is finite (the "ego depletion" model is contested), but that decision quality is influenced by cognitive state, and that important decisions should be made when cognitive resources are not depleted by prior demands.

Practical responses include: make high-stakes decisions early in the day or after recovery. Reduce trivial decisions through defaults and routines. Avoid committing to major decisions when tired, hungry, or stressed unless absolutely necessary.

Satisficing vs. Maximising

Herbert Simon's concept of satisficing - choosing an option that is good enough rather than searching exhaustively for the optimal option - is not a cognitive failure. It is a rational adaptation to the fact that search has costs and perfect information is rarely available.

Barry Schwartz's research on the "paradox of choice" shows that maximisers - people who insist on exploring all options before deciding - report lower satisfaction with their choices than satisficers, even when they make objectively better choices. The additional information gathered in extensive search increases regret about foregone alternatives without proportionally improving decision quality.

The practical implication: define minimum acceptable standards before searching, and stop when an option meets those standards. This is most applicable to consumer decisions and routine choices. High-stakes irreversible decisions warrant more thorough search - but most decisions are not high-stakes and irreversible.

Bias Checklist for Important Decisions

Before finalising any significant decision, work through this quick checklist:

DECISION BIAS CHECKLIST

? Availability bias: Am I overweighting vivid, recent, or emotionally memorable examples rather than base rates?

? Sunk cost: Am I factoring in past investment that is already gone and cannot be recovered?

? Anchoring: Is my estimate being anchored by an initial number or framing that may not be reliable?

? Overconfidence: Am I more confident than my track record on similar decisions justifies?

? Confirmation bias: Have I actively sought disconfirming evidence, or only information that supports what I already believe?

? Optimism bias: Am I assuming things will go well because I want them to, rather than because the evidence supports it?

? Status quo bias: Am I keeping the current course partly because changing feels risky, beyond what rational risk assessment justifies?

? Planning fallacy: Are my time and cost estimates based on similar past projects, or on optimistic assumptions about this one?

? Second-order effects: Have I thought through what happens after the first-order outcome, good or bad?

? Reversibility: Is this decision reversible if it turns out to be wrong? Am I applying appropriate deliberation given that?

Source Library

  • Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.
  • Klein, G. (2007). Performing a project premortem. Harvard Business Review, 85(9), 18-19.
  • Simon, H. A. (1956). Rational choice and the structure of the environment. Psychological Review, 63(2), 129-138. doi.org/10.1037/h0042769
  • Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124-1131. doi.org/10.1126/science.185.4157.1124
  • Schwartz, B. (2004). The Paradox of Choice: Why More Is Less. HarperCollins.
  • Tetlock, P. E., & Gardner, D. (2015). Superforecasting: The Art and Science of Prediction. Crown.
  • Lovallo, D., & Kahneman, D. (2003). Delusions of success: How optimism undermines executives' decisions. Harvard Business Review, 81(7), 56-63.

Related Reading on WhenNotesFly

Last reviewed: May 2026. Written and reviewed by the WhenNotesFly editorial team. For corrections: editorial@whennotesfly.com

Frequently Asked Questions

What is the difference between a good decision and a good outcome?

A good decision is made with sound reasoning, appropriate information, and honest acknowledgment of uncertainty. A good outcome is a favourable result. Because outcomes are influenced by factors outside your control, a good decision can produce a bad outcome and a poor decision can produce a good outcome through luck. Decision science focuses on improving the process, which improves average outcomes over time.

What is a premortem and how do I run one?

A premortem is a structured technique where you imagine a decision has already failed and work backward to identify what went wrong, before committing. State the decision, imagine it has failed badly 12 months from now, have each participant independently list plausible failure causes, aggregate them, and mitigate the most plausible and consequential risks.

What is a decision journal?

A decision journal is a record of significant decisions, the reasoning behind them, and their eventual outcomes. It forces more explicit reasoning at decision time and provides feedback to improve calibration over time, separating the reasoning process from the outcome.

What is the difference between reversible and irreversible decisions?

Reversible (Type 2) decisions can be easily undone and should be made quickly with good-enough information. Irreversible (Type 1) decisions are difficult or impossible to reverse and warrant more deliberation. A common mistake is treating reversible decisions as if they were irreversible, which causes paralysis.

Which cognitive biases most affect decisions?

The most consequential for everyday decisions are availability bias, sunk cost fallacy, anchoring, overconfidence, and confirmation bias. Awareness does not eliminate them, but it creates opportunities for correction.

Share this article