Branchoria Sharper Thinking Sharper Thinking

Practical ways to reason better, test ideas, and make decisions under uncertainty.

Within Sharper Thinking

How Confident Should You Be?

Confidence becomes more useful when it is stated clearly, checked against evidence, and updated after outcomes.

On this page

  • Why confidence needs a scale
  • Calibration checks for forecasts and decisions
  • Updating without overcorrecting
Preview for How Confident Should You Be?

Introduction

Confidence is useful only when it means something. Saying “I’m sure” or “I think so” gives little help unless the listener knows how that confidence relates to reality. Confidence calibration is the habit of matching your stated certainty to the evidence: when you say 70%, outcomes like that should happen about seven times in ten; when you give a 90% range, the true answer should fall inside it about nine times in ten. This matters because better thinking is not just about reaching the right answer. It is about knowing how much weight to place on an answer before events prove it right or wrong.

Overview image for Calibration Calibration turns uncertainty from an embarrassment into a tool. It helps you separate strong evidence from a strong feeling, make forecasts that can be checked, learn from outcomes without rewriting history, and update beliefs without swinging wildly after every new clue. Research on forecasting, overconfidence and judgement shows the same basic lesson in different settings: confidence improves when it is expressed on a scale, tested against feedback, and revised with attention to both prior evidence and new information. [iarpa.gov+2learnmoore.org]iarpa.govACEThe goal of the ACE Program is to dramatically enhance the accuracy, precision, and timeliness of intelligence forecasts for a broad r…

Why confidence needs a scale

The first step in calibration is replacing vague confidence words with usable degrees of belief. “Likely”, “possible” and “almost certain” sound precise in conversation, but people use them differently. One person’s “likely” may mean 55%; another’s may mean 80%. A numerical probability is not magic, but it forces a clearer claim. “I’m 70% confident this supplier will deliver by Friday” is easier to test than “they should probably deliver”.

A calibrated confidence statement has two parts: the answer and the uncertainty around it. For a yes-or-no forecast, the statement may be a probability: “There is a 30% chance this feature will miss the deadline.” For a quantity, it may be an interval: “I am 80% confident the final cost will be between £40,000 and £55,000.” The point is not to pretend that judgement is mathematical in every detail. The point is to make the uncertainty visible enough that it can be compared with later outcomes.

The classic calibration test is simple. Gather many predictions made at the same confidence level and ask whether the outcomes match. If 100 of your forecasts were made at 80% confidence, roughly 80 should turn out right. If only 60 do, you are overconfident at that level. If 95 do, you may be underconfident or making your forecasts too cautious. For interval estimates, the same logic applies: if your 90% confidence intervals contain the true value only half the time, your ranges are far too narrow.

This is why calibration is different from ordinary accuracy. A person can be accurate on easy questions while still being poorly calibrated if they express too much certainty on hard ones. A person can also be modestly accurate but well calibrated if they know when the evidence is weak. In decision-making, that second skill is often more valuable than it sounds. A well-calibrated thinker can say, “I am probably right, but not right enough to bet the project on it.”

Research on overconfidence helps clarify the target. Don Moore and Paul Healy distinguish three forms of overconfidence: overestimating your actual performance, overplacing yourself relative to others, and overprecision, which is excessive certainty that your belief is correct. Calibration mainly attacks the third problem. It does not ask, “Are you clever?” It asks, “Does the certainty you expressed match the track record of claims like this?” [healy.econ.ohio-state.edu]healy.econ.ohio-state.eduThe Trouble With Overconfidenceby DA Moore · 2008 · Cited by 3888 — The authors present a reconciliation of 3 distinct ways in defined overconfidence: (a) overesti…

What calibration checks reveal that intuition misses

Calibration becomes powerful when predictions are recorded before the result is known. Without that record, feedback is easily distorted. After an outcome arrives, people often remember their earlier view as more accurate than it was, or judge the quality of a decision mainly by whether it worked out. That is dangerous because a good decision can fail in an unlucky world, and a poor decision can succeed by chance. [Springer]link.springer.comOpen source on springer.com.

A practical calibration check needs three ingredients:

  1. A clear forecast. The statement must be specific enough to resolve. “This campaign will do well” is too vague; “This campaign will generate at least 500 qualified leads by 30 September” is checkable.
  2. A confidence level. The prediction should include a probability or interval. “70% chance” is better than “I expect”, because it lets you compare similar forecasts over time.
  3. A resolution rule. You need to know what counts as right. For complex decisions, this may mean defining the metric, deadline, data source and threshold in advance.

Forecasting tournaments show why this discipline matters. The Intelligence Advanced Research Projects Activity’s Aggregative Contingent Estimation programme was designed to improve intelligence forecasting by eliciting probabilistic judgements, combining forecasts, and testing them against real events. The Good Judgment Project, led by Philip Tetlock and Barbara Mellers, used behavioural interventions such as probability training, collaboration and tracking of high performers; its research found that these interventions improved both calibration and resolution, meaning forecasters became better at matching probabilities to reality and at distinguishing more likely from less likely events. [iarpa.gov]iarpa.govACEThe goal of the ACE Program is to dramatically enhance the accuracy, precision, and timeliness of intelligence forecasts for a broad r…

The lesson is not that every personal or workplace decision should become a formal tournament. It is that calibration requires a feedback loop. A sales manager, doctor, analyst, engineer or student can all benefit from the same basic routine: write the prediction down, include the confidence, decide how it will be checked, and later compare the confidence with the outcome.

The Brier score is one common way to score probabilistic forecasts. It measures the squared difference between the probability assigned to an event and the event’s actual outcome, so lower scores are better. It rewards both being right and being appropriately confident. However, it should not be treated as a complete verdict on judgement by itself: recent methodological work stresses that Brier scores can reflect not only calibration but also the underlying difficulty and distribution of the prediction problem. [PMC]pmc.ncbi.nlm.nih.govOpen source on nih.gov.

Calibration illustration 1

How to state uncertainty without sounding evasive

Many people avoid calibrated language because they fear it will make them look indecisive. In practice, the opposite is often true. A precise uncertainty statement can be more decision-ready than a confident slogan because it tells others how much risk remains.

Compare these two statements:

“The launch will be fine.”

“I put the chance of launching on time at about 65%. The main risk is supplier testing, and I would raise my confidence to 80% if the test report arrives by Wednesday.”

The second statement is less macho but more useful. It identifies the confidence level, the evidence behind it, and the condition that would change the estimate. That is exactly what good analytical thinking needs.

A calibrated statement usually does four things:

  • Names the claim. “The renewal rate will stay above 85% this quarter.”
  • Gives a probability or range. “I’m about 75% confident.”
  • States the evidence. “The last three quarters were stable, but the price rise affects a third of customers.”
  • Names the update trigger. “If cancellations exceed 5% in the first two weeks, I would revise down sharply.”

This structure also reduces false certainty in group settings. Teams often hear the loudest or most senior voice as the most confident voice, then mistake confidence for evidence. Asking everyone to give a probability before discussion can reveal hidden disagreement. A team that appears aligned around “probably yes” may discover that one person means 55%, another means 85%, and a third means “I have not really thought about it”.

Calibration does not require pretending that all uncertainty is measurable with scientific precision. A probability can be a disciplined estimate rather than a laboratory result. The benefit is that it makes uncertainty comparable. “I am 60% confident” and “I am 90% confident” should lead to different actions, different levels of checking, and different willingness to commit resources.

The Science of Making Better Decisions

Channel: Janus Talks

Open on YouTube

Calibration checks for forecasts and decisions

Forecasts are the cleanest training ground for calibration because they resolve. Decisions are messier, because their outcomes depend on action, luck, changing circumstances and the quality of execution. Still, most serious decisions contain forecasts inside them. “We should hire this person” includes forecasts about performance, fit, retention and opportunity cost. “We should buy this tool” includes forecasts about adoption, savings, integration and risk.

A useful decision journal therefore records not just what you chose, but what you expected to happen. This protects learning from outcome bias. If a risky decision succeeds, the journal can show whether it was a good risk or a lucky escape. If a careful decision fails, it can show whether the reasoning was weak or the world simply broke against you.

For practical use, calibration checks can be kept light:

  • Use prediction buckets. Review predictions at 50%, 60%, 70%, 80% and 90% confidence. Over time, ask whether each bucket resolves at roughly the stated rate.
  • Track interval coverage. For estimates such as cost, duration or demand, check whether your 80% or 90% ranges contain the true value as often as they should.
  • Separate calibration from discrimination. Calibration asks whether your probabilities match frequencies. Discrimination asks whether you successfully rank high-risk and low-risk cases. A model or person can be good at one and weaker at the other.
  • Review by category. You may be calibrated on technical estimates but overconfident on timelines, people judgements or market reactions.

The category point matters. Studies of confidence and accuracy repeatedly show that calibration is not a single global trait. A person may be well calibrated in a familiar domain with repeated feedback and poorly calibrated in a noisy domain where feedback is delayed or ambiguous. In eyewitness research, for example, confidence can be informative under some “pristine” conditions, but contamination, delay and other case features can weaken the confidence-accuracy link. [psych.utah.edu]psych.utah.eduEyewitness Confidence Does Not Necessarily IndicateEyewitness Confidence Does Not Necessarily Indicate

This is why calibration should be local. Instead of asking, “Am I overconfident?”, ask, “Where does my confidence outrun my evidence?” Common high-risk zones include long timelines, rare events, adversarial situations, personal performance, hiring, investment, medical diagnosis, political predictions and any setting where feedback is slow or ambiguous.

Calibration illustration 2

What good forecasters do differently

Good forecasters are not simply more cautious. If they were, they would assign middling probabilities to everything and avoid embarrassment. That is not good calibration; it is fog. Strong forecasters learn when to move away from 50% and when not to.

The Good Judgment Project evidence is useful because it treated forecasting as a skill that could be measured and improved rather than as a mysterious talent. In the ACE tournament, participants made probabilistic forecasts on real geopolitical questions. Good Judgment’s published research reported that probability training, team collaboration and tracking of high performers improved forecasting performance, while IARPA described the broader programme as an effort to enhance the accuracy, precision and timeliness of intelligence forecasts through elicitation, aggregation and empirical testing against real events. [iarpa.gov]iarpa.govACEThe goal of the ACE Program is to dramatically enhance the accuracy, precision, and timeliness of intelligence forecasts for a broad r…

The habits associated with better forecasting are directly relevant to everyday analytical thinking:

  • Start with the base rate. Before asking what makes this case special, ask how often similar cases succeed, fail, overrun or reverse.
  • Break the question into parts. A launch forecast might depend on engineering completion, legal approval, customer readiness and supplier reliability.
  • Update gradually but meaningfully. Do not ignore new evidence, but do not let one vivid clue erase a stronger prior record.
  • Look for disconfirming evidence. Ask what you would expect to see if your current view were wrong.
  • Keep score. Calibration cannot improve reliably if forecasts disappear after the meeting.

This last habit is often the missing one. People receive feedback constantly, but not all feedback teaches. If the original forecast was vague, if the confidence level was never written down, or if the outcome is judged through hindsight, the mind can protect its self-image while learning very little.

Toward Superforecasting®: Lessons from the Eli Lilly Probability Assessment Panel | SDG

Channel: Strategic Decisions Group

Open on YouTube

Updating without overcorrecting

Good calibration is not stubbornness. When evidence changes, confidence should change too. The difficulty is that people can err in both directions: they may cling to an old view despite strong new information, or overreact to a recent, vivid event and abandon a well-supported prior.

Bayesian reasoning offers a useful ideal: combine prior beliefs with new evidence according to how reliable each is. In plain English, ask two questions. First, what did similar cases usually look like before this new information arrived? Second, how diagnostic is the new information? Research on probabilistic reasoning finds that people often underweight prior information, known as base-rate neglect, and may also underweight new evidence, known as conservatism, depending on the situation and how beliefs are elicited. [Cambridge University Press & Assessment]cambridge.orgOpen source on cambridge.org.

A balanced update has three movements:

  1. Anchor in the prior. Start with the outside view. If nine out of ten similar projects overran, your project should not begin at “almost certain to be on time” just because the plan looks tidy.
  2. Estimate evidence strength. A signed contract, failed test, customer cancellation or audited result should move confidence more than a rumour, anecdote or single emotional meeting.
  3. Move by degrees. A 70% belief might become 55% or 85% after new evidence. It should not automatically become 5% or 99% unless the evidence is decisive.

Weather forecasting gives a clear public example of why updating can support trust when done well. Research on probabilistic weather communication found that explicit uncertainty information and newer, more reliable forecast updates can help users make decisions and maintain trust, especially when forecasts change as better information arrives. The principle transfers: changing your mind is not a weakness if the change is tied to better evidence and explained clearly. [repository.library.noaa.gov]repository.library.noaa.govThe Impact of Forecast Inconsistency and ProbabilisticThe Impact of Forecast Inconsistency and Probabilistic

Overcorrection often happens after emotionally salient outcomes. A failed hire makes a manager distrust all candidates from the same background. A successful product launch makes a team believe its process is better than it was. A missed deadline leads to absurdly padded future estimates. Calibration asks for a quieter review: what probability did we assign, what happened, what evidence did we miss, and how much should similar forecasts move next time?

Calibration illustration 3

The traps that make confidence feel better than it is

Miscalibration often feels like clear thinking from the inside. That is what makes it hard to catch. Several traps are especially common.

Fluency feels like truth. If an explanation is easy to tell, people often feel more confident in it. But a smooth story can be built from selective evidence. Calibration improves when you ask, “What would I expect to see if the opposite were true?”

Detail feels like evidence. A plan with many steps can seem more credible because it is vivid. Yet each extra dependency may add failure points. For timelines and budgets, calibrated thinkers often widen rather than narrow their uncertainty when they inspect the details.

Expertise leaks across boundaries. Skill in one domain can create excessive confidence in another. A person with deep technical knowledge may still be poorly calibrated about customer behaviour, regulation, hiring or politics.

Consensus can hide uncertainty. A group may agree on a decision while disagreeing sharply on likelihoods. Asking for private probability estimates before discussion helps reveal whether apparent agreement is real.

Outcomes rewrite memories. Once the result is known, people tend to see it as more predictable than it was. That weakens learning unless the original prediction is recorded. [MPG.PuRe]pure.mpg.deOpen source on mpg.de.

The practical answer is not to distrust every confident thought. It is to treat confidence as a claim that needs calibration. When the stakes are low, rough confidence is fine. When the stakes are high, confidence should be written down, attached to evidence, and checked later.

You Don't Lose Money by Being Wrong. You Lose It by Being Certain | Superforecasting

Channel: Playbooks de Caja

Open on YouTube

A practical calibration routine

Calibration improves through repeated, low-friction practice. The routine does not need to be elaborate. It needs to be consistent enough to expose patterns.

Start with ten to twenty predictions a week in areas where outcomes will resolve within days or months. They can be work estimates, personal forecasts, project risks, meeting outcomes, learning goals or public events. Each prediction should include a probability and a resolution date. Avoid trick questions and avoid predictions that depend entirely on your own future choice unless that is what you are trying to study.

For each prediction, record:

  • the claim;
  • the probability or confidence interval; [researchgate.net]researchgate.netSource details in endnotes.
  • the main reason for the estimate;
  • what evidence would move the estimate up or down;
  • the eventual outcome.

After enough predictions resolve, look for patterns rather than single embarrassments. Are your 80% predictions right only 60% of the time? Are your 60% predictions actually closer to coin flips? Are your time estimates consistently too narrow? Are you better calibrated after writing down base rates? This is where calibration becomes a thinking skill rather than a personality judgement.

Commercial and professional calibration training shows both promise and limits. A 2024 study of intelligence analysts found that calibration training improved overall calibration and bias, especially for interval estimation, but the effects varied by task; on a binary-choice task, the shift looked more like reduced confidence than a general improvement in metacognitive monitoring. That is an important caution: calibration is not simply “be less confident”. The goal is to be confident at the right times, doubtful at the right times, and explicit enough that reality can correct you. [Wiley Online Library]onlinelibrary.wiley.comOpen source on wiley.com.

A useful personal benchmark is interval estimation. Try giving 80% ranges for factual quantities you do not know exactly: the population of a city, the cost of a past project, the number of customers affected by an issue, or the time required for a task. If fewer than eight out of ten true values fall inside your ranges, widen them. If nearly all do, sharpen them. This exercise teaches the feel of honest uncertainty better than abstract advice does.

What calibrated confidence changes

Calibrated confidence changes the way you argue, decide and learn. Arguments become less about defending identity and more about locating uncertainty. Decisions become less dependent on the most forceful voice. Reviews become less about blame and more about whether the evidence available at the time justified the confidence expressed.

It also makes intellectual humility more concrete. Humility is often misunderstood as low confidence. Calibration offers a better version: confidence proportional to evidence. A calibrated thinker can be bold when the evidence is strong, cautious when the evidence is thin, and ready to update when the world supplies new information.

The wider benefit is compounding learning. Every clear forecast becomes a small experiment in judgement. Every resolved outcome becomes feedback. Every update becomes a chance to distinguish “I was wrong” from “I was unlucky”, and “I was right” from “I got away with it”. Over time, this builds a more reliable sense of when to trust your judgement, when to seek more evidence, and when to leave room for surprise.

Amazon book picks

Further Reading

Books and field guides related to How Confident Should You Be?. Use these as the next step if you want deeper reading beyond the article.

Browse more on Amazon: Superforecasting Thinking, Fast and Slow How to Measure Anything

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Live-tested eBay searches with available results related to this page.

Using USA
Browse more on eBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

  1. Source: iarpa.gov
    Link: https://www.iarpa.gov/research-programs/ace
    Source snippet

    ACEThe goal of the ACE Program is to dramatically enhance the accuracy, precision, and timeliness of intelligence forecasts for a broad r...

  2. Source: learnmoore.org
    Link: https://learnmoore.org/papers/Mellers%20et%20al%202014.pdf
    Source snippet

    Mellers et al 2014.pdfby B Mellers · 2014 · Cited by 434 — Results showed that probability training, team collaboration, and tracking imp...

  3. Source: healy.econ.ohio-state.edu
    Title: The Trouble With Overconfidence
    Link: https://healy.econ.ohio-state.edu/papers/Moore_Healy-TroubleWithOverconfidence.pdf
    Source snippet

    by DA Moore · 2008 · Cited by 3888 — The authors present a reconciliation of 3 distinct ways in defined overconfidence: (a) overesti...

  4. Source: learnmoore.org
    Link: https://learnmoore.org/mooredata/HOC.pdf
    Source snippet

    Overprecision in Judgmentby DA Moore · Cited by 192 — Moore and Healy (2008) distinguish three varieties of overconfidence: 1) Overestima...

  5. Source: link.springer.com
    Link: https://link.springer.com/article/10.1007/s13164-023-00672-2

  6. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC7333631/

  7. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12818272/

  8. Source: psych.utah.edu
    Title: Eyewitness Confidence Does Not Necessarily Indicate
    Link: https://psych.utah.edu/_resources/documents/people/committee-docs/fall-2025/2024%20-%20Moore%20et%20al%202024%20Eyewitness%20Confidence%20Does%20Not%20Necessarily%20Indicate%20Accuracy.pdf

  9. Source: cambridge.org
    Link: https://www.cambridge.org/core/journals/judgment-and-decision-making/article/base-rate-neglect-and-conservatism-in-probabilistic-reasoning-insights-from-eliciting-full-distributions/57619E3572DB3A6035101546DB147F7E

  10. Source: repository.library.noaa.gov
    Title: The Impact of Forecast Inconsistency and Probabilistic
    Link: https://repository.library.noaa.gov/view/noaa/51145/noaa_51145_DS1.pdf

  11. Source: pure.mpg.de
    Link: https://pure.mpg.de/rest/items/item_2102632_11/component/file_2102631/content

  12. Source: onlinelibrary.wiley.com
    Link: https://onlinelibrary.wiley.com/doi/10.1002/acp.4236

  13. Source: link.springer.com
    Link: https://link.springer.com/article/10.1186/s41235-026-00717-x

  14. Source: link.springer.com
    Link: https://link.springer.com/article/10.3758/BF03205299

  15. Source: link.springer.com
    Link: https://link.springer.com/rwe/10.1007/978-3-319-28099-8_1157-1

  16. Source: assets.publishing.service.gov.uk
    Link: https://assets.publishing.service.gov.uk/media/6093d448e90e0726f52fc54c/updating-the-evidence-behind-the-optimism-bias-uplifts-for-transport-appraisals.pdf

  17. Source: onlinelibrary.wiley.com
    Link: https://onlinelibrary.wiley.com/doi/full/10.1002/acp.4258

  18. Source: sms.onlinelibrary.wiley.com
    Link: https://sms.onlinelibrary.wiley.com/doi/10.1002/smj.3450

  19. Source: healy.econ.ohio-state.edu
    Title: Moore Healy Trouble With Overconfidence WP
    Link: https://healy.econ.ohio-state.edu/papers/Moore_Healy-TroubleWithOverconfidence_WP.pdf

  20. Source: cambridge.org
    Link: https://www.cambridge.org/core/journals/judgment-and-decision-making/article/weighted-brier-score-decompositions-for-topically-heterogenous-forecasting-[tournaments

  21. Source: cambridge.org
    Title: recalibrating probabilistic forecasts to improve theiraccuracy
    Link: https://www.cambridge.org/core/services/aop-cambridge-core/content/view/5D8F0DFC90FC1E9B5C38C12992A6C6A7/S1930297500009049a.pdf/recalibrating_probabilistic_forecasts_to_improve_theiraccuracy.pdf

  22. Source: researchnow.flinders.edu.au
    Link: https://researchnow.flinders.edu.au/en/publications/the-effect-of-judgement-type-and-confidence-scale-on-confidence-a/

  23. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12101119/

  24. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC7923594/

  25. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12523994/

  26. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12730000/

  27. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC10189590/

  28. Source: pubmed.ncbi.nlm.nih.gov
    Link: https://pubmed.ncbi.nlm.nih.gov/18426301/

  29. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC8763848/

  30. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12366471/

  31. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC8074796/

  32. Source: Wikipedia
    Link: https://en.wikipedia.org/wiki/Probability

  33. Source: Wikipedia
    Title: Brier score
    Link: https://en.wikipedia.org/wiki/Brier_score

  34. Source: Wikipedia
    Title: The Good Judgment Project
    Link: https://en.wikipedia.org/wiki/The_Good_Judgment_Project

  35. Source: emergentmind.com
    Title: brier score
    Link: https://www.emergentmind.com/topics/brier-score

  36. Source: dataopsschool.com
    Title: brier score
    Link: https://dataopsschool.com/blog/brier-score/

Additional References

  1. Source: youtube.com
    Link: https://www.youtube.com/watch?v=cac76CPLc-M
    Source snippet

    Toward Superforecasting®: Lessons from the Eli Lilly Probability [Assessment]({{ 'assessment/' | relative_url }}) Panel | SDG...

  2. Source: youtube.com
    Title: The Science of Making Better Decisions
    Link: https://www.youtube.com/watch?v=yDKBHSCV6Bk
    Source snippet

    Confidence Calibration forecasting Superforecasting You Don't Lose Money by Being Wrong. You Lose It by Being Certain | Superforecasting...

  3. Source: youtube.com
    Link: https://www.youtube.com/watch?v=mBDh5yivq7M
    Source snippet

    Decision-Making Under Miscalibration...

  4. Source: researchgate.net
    Link: https://www.researchgate.net/publication/364692195_Base_rate_neglect_and_conservatism_in_probabilistic_reasoning_Insights_from_eliciting_full_distributions

  5. Source: researchgate.net
    Link: https://www.researchgate.net/publication/320911494_Confidence_Calibration_in_a_Multiyear_Geopolitical_Forecasting_Competition

  6. Source: researchgate.net
    Link: https://www.researchgate.net/publication/8685888_Overconfidence_in_Interval_Estimates

  7. Source: researchgate.net
    Link: https://www.researchgate.net/publication/5305238_The_Trouble_With_Overconfidence

  8. Source: researchgate.net
    Link: https://www.researchgate.net/publication/264682051_Confidence-Accuracy_Calibration_with_General_Knowledge_and_Eyewitness_Memory_Cued_Recall_Questions

  9. Source: researchgate.net
    Link: https://www.researchgate.net/publication/277087515_Identifying_and_Cultivating_Superforecasters_as_a_Method_of_Improving_Probabilistic_Predictions

  10. Source: medium.com
    Link: https://medium.com/%40eskandar.sahel/applying-calibration-techniques-to-improve-probabilistic-predictions-in-machine-learning-models-c175c2e38ffc

Topic Tree

Follow this branch

Parent topic

Sharper Thinking

Related pages 29

More on this topic 6