Base Rate Neglect

Base Rate Neglect is the everyday habit of letting a vivid case crowd out how common or rare something is in general. You hear one story, see one example, or get one striking match, and that specific picture feels more real than the background rate you never pictured.

The classic demonstration is the taxicab problem. Imagine a city where 85 percent of the cabs are green and 15 percent are blue. A cab is in a night-time hit-and-run, and a witness who is right about cab color 80 percent of the time says it was blue. Ask people how likely it is the cab really was blue and most answer around 80 percent, echoing the witness. The correct answer, once you fold in that blue cabs are rare to begin with, is only about 41 percent - closer to a coin flip than to certainty. The witness detail is vivid and specific; the 15 percent base rate is abstract, so it quietly drops out of the calculation.

This is a gut-weighting error, not a classroom Bayes problem. You are not failing a formula on paper; you are skipping the question "how often does this happen?" because the case already feels decisive. The denominator never gets a seat at the table.

The same pull shows up whenever a stereotype or prototype feels compelling. When people are told a group held 70 engineers and 30 lawyers, a short personality sketch can make them ignore that split entirely and guess by resemblance instead. A quiet reader feels like a librarian even when engineers vastly outnumber librarians in the city, and a plane crash on the news makes flying feel dangerous even when driving kills far more people per mile.

Modern feeds make the gap worse. Viral cases, dramatic headlines, and personal anecdotes arrive constantly, while base rates sit in footnotes, reports, or nowhere at all. The more vivid the story, the harder it is to hold the boring statistical backdrop in mind.

Using case details is not the mistake - doctors need symptoms, investors need company facts, and jurors need testimony, and specific evidence should update your estimate when it is genuinely strong. The bug is narrower than "trust numbers over stories." It is letting one memorable case set the odds before you ask how common the outcome is in the population you are actually in, so that the base rate is neglected exactly when it would have changed the answer.

• "I saw a plane crash on the news, so flying feels too dangerous to risk."
• "My friend got rich on this stock, so it must be a good bet for everyone."
• "One scary crime story means my neighborhood is getting unsafe."
• "This resume looks exactly like our top performers, so hire probability is high."
• "That quiet, bookish stranger fits the profile, so they must work in a library."
• "This startup founder is charismatic and smart, so success is almost guaranteed."
• "Everyone on my feed got laid off - the whole industry is collapsing."
• "The viral scam story means this happens to people all the time."

The scary screening result: a routine test comes back positive for a rare condition, and the result feels like a verdict. But when a disease affects one person in a thousand and the test throws occasional false alarms, most positive results are false positives - the true odds of illness can still be modest. Skipping the prevalence math turns a piece of evidence into a diagnosis and triggers days of fear before any confirmatory test.

The star-employee lookalike: a candidate matches your mental model of the best hire you ever made - same school, same easy confidence, same way of talking about the work. You rate their fit as near-certain while never asking how often people who look like that actually succeed in your specific pipeline, where the quiet reference-checkable performers may outperform the type you keep falling for.

The proven-strategy trap: a hot sector and a few loud winners make an investing approach feel like a sure thing. What you cannot see are the many similar bets that failed quietly and never got a headline, so the survivors set your estimate of the odds. The crowded, late-stage version of the trade feels as promising as the early one because the base rate of failure was never in the frame.

The one-incident rule: a single dramatic event - a breach, an accident, a headline crime - drives a new policy because the case is unforgettable. Broader data on how often the event occurs, what it costs, and what the new rule trades away gets far less airtime, so a rare occurrence reshapes procedures that thousands of ordinary cases never touched.

The feed-as-frequency error: a viral post about a scam, a side effect, or an overnight success feels like the norm because it is recent, emotional, and everywhere in your timeline. The algorithm surfaces the vivid outlier precisely because it performs, and you mistake how often you see the story for how often the thing happens.

Base Rate Neglect pushes you toward confident judgments built on unrepresentative stories. Medical worry, legal errors, bad hires, and costly bets can follow when rare events get treated as typical and common ones get overlooked because they lack a poster case.

Organizations repeat the mistake at scale. Training programs, policies, and product bets get shaped by memorable outliers instead of measured frequencies, so resources flow toward dramatic risks and away from quieter ones that actually cause more harm. The louder the anecdote in the room, the more likely it is to set direction.

On a personal level, anxiety and avoidance can spike when you treat low-probability threats as likely because one example stuck, and you may quietly underestimate real risks that never produced a vivid story. Both errors run in the same direction: attention, not frequency, is setting your sense of danger.

The social cost is polarization about what is "obvious." Two people can look at the same headline case and draw opposite lessons because neither anchored on how common the event actually is, and each walks away certain the other is ignoring reality.

Brains prefer concrete narratives over abstract frequencies. Representativeness - how much a case fits a stereotype - feels like probability even when it is not, especially under time pressure or emotional load, and a story arrives with a face and stakes while a base rate arrives as a bare number.

Relevance and framing matter too. Bar-Hillel's account is that people use base rates when the rate seems causally relevant to the case and discard it when the specific evidence feels more directly tied to the outcome. Without practice translating both the story and the frequency into one estimate, the case in front of you wins by default.

Kahneman and Tversky's "On the Psychology of Prediction" (1973) showed that clearly stated base rates - such as a 70/30 split of engineers to lawyers - were largely ignored once a personality description invited a similarity judgment, and were used mainly when no individuating story was available. Prototype fit was treated as probability even when the prototype was rare in the population.

Maya Bar-Hillel's "The Base-Rate Fallacy in Probability Judgments" (1980) mapped when the neglect is strongest, using problems like the taxicab case to show that people lean on specific evidence and drop base rates they judge to be less relevant, even when those rates should dominate the answer. Her relevance account explains why the same person will use a base rate in one framing and neglect it in another.

• A single story or example feels more convincing than any statistic you have seen.
• You can describe the case in detail but cannot state how often it happens in the population.
• Your estimate would barely move if someone told you the base rate was ten times lower.
• A positive test or a stereotype match feels nearly decisive without any of the math.
• Base rates get dismissed as "just numbers" while one anecdote gets treated as ground truth.
• After a vivid news event, your sense of risk shifts sharply though nothing changed in the overall data.

When stakes are high, name how common the outcome is before you let the vivid case run the show. The point is not to ignore the details - it is to stop the story from becoming the whole denominator.

• Ask "how common is this in general?" before "how well does this case fit?"
• Seek a second number when you only have a story: incidence, failure rate, or historical frequency.
• Compare your gut estimate to an outside source or a simple head-count, not only to the case in front of you.
• Notice when representativeness feels like proof and pause for the denominator.
• After a vivid news event, look up base rates before you change behavior.
• When teaching or deciding for others, lead with frequency context, not only the dramatic example.

When should the vivid case override the base rate?

When you have strong, reliable individuating evidence that truly splits this case from the class - a diagnostic test with excellent specificity, not a stereotype or a story. Most "this time is different" claims fail that bar.

If I accept base rates, do I erase personal responsibility or unique talent?

Base rates are starting points, not destinies. Update with real evidence of skill or effort. Neglect is skipping the start line; respect is beginning there.

Can focusing on base rates make me cruel in hiring or medicine?

Used bluntly, yes. Use them to set priors, then apply fair individuating tests. Cruelty is freezing people at the group average without a chance to show signal.

What if the published base rate feels politically inconvenient?

Inconvenience is not a refutation. Check measurement quality, then update. Motivated dismissal of rates is neglect with a flag.

Isn't every individual a special case?

Everyone is a case; not every case has information that beats the class average. Specialness must be earned by data, not by narrative preference.

For Base Rate Neglect, put the background odds on the table before the vivid description gets the last word.

Positive test

Perfect fit candidate

News-driven fear

Practice this pattern in the Reframing App - capture the trigger, label it (like Base Rate Neglect), check evidence, and write a more balanced thought.