The 37% Rule
How an equation created to hire secretaries ended up explaining dating, relationship apps, career choices, artificial intelligence and human behavior under uncertainty.
Prof. Maurício Pinheiro
Abstract
The 37% Rule is one of the most elegant results in probability theory and decision science. Originally known as the Secretary Problem or the Optimal Stopping Problem, it attempts to answer a universal question: when is the ideal moment to stop exploring possibilities and finally make a decision?
The problem became especially relevant in the digital age, shaped by dating apps, endless scrolling, choice overload and algorithmic anxiety. In a world where there always seems to be someone more attractive, more intelligent or more compatible just one swipe away, the challenge is no longer finding options — but knowing when to stop searching.
The theory demonstrates mathematically that, in irreversible decision processes, the optimal strategy is to reject approximately the first 37% of options solely for observation, learning and calibration. After this initial exploration phase, one should select the first option clearly superior to all previous ones.
This article explores the historical origins of the theory during the rise of applied mathematics in the twentieth century, its probabilistic derivation, and its connections to artificial intelligence, reinforcement learning, behavioral economics and contemporary psychology. It also examines how digital platforms exploit variable reward mechanisms and infinite choice architectures to keep users trapped in endless cycles of search and comparison.
More than a mathematical curiosity, the 37% Rule reveals something profoundly human: under uncertainty, perfection is impossible. The best we can do is optimize our choices before infinite searching destroys our ability to choose at all.
Table of Contents
- Introduction — The Statistical Collapse of the Modern Brain
- The Algorithmic Casino of Modernity
- The Human Brain Was Not Built for Infinite Swiping
- The Mathematics of Indecision
- The Birth of the 37% Rule
- The Mathematical Derivation of the 37% Rule
- The Cruelest Part of the Theory
- When the 37% Rule Fails
- Between Mathematics, Love and Uncertainty
- Recommended Reading
1. Introduction — The Statistical Collapse of the Modern Brain
Imagine the following absurd — and extremely modern — situation.
You install a new dating app promising yourself that this time you will be rational.
“This is it,” you think.
“This time I’m going to choose correctly.”
Then the infinite parade begins.
- One person seems perfect… until you discover she lives 11,000 kilometers away and replies to messages in the emotional time zone of Saturn.
- Another is unbelievably beautiful, but speaks like a CAPTCHA trying to prove it hasn’t been a robot since 2017.
- Another explains quantum astrology, vibrational crystals and Atlantean pyramids with the confidence of someone defending a PhD thesis at MIT.
- Another is intelligent, funny and compatible… until she disappears for three days after a simple “hi :)” and returns with “crazy week lol”.
- Another has perfect chemistry online… but in real life looks like a statistical fusion between Instagram filters and morgue lighting.
- Another appears emotionally mature… until she transforms a casual conversation about coffee into a Freudian analysis of her childhood trauma.
- Another seems sophisticated… until she tries to sell you a “guaranteed” cryptocurrency invented by a visionary cousin named Douglas.
- Another replies exclusively with “lol”, “yeah” and 👍 as if personality were a premium subscription feature.
- Another reappears after thirty years… and you realize getting back together would be like buying back your old car believing nostalgia reduces mileage.
- Another looks so perfect on her profile that you begin to suspect she was trained by an AI specialized in male dopamine manipulation.
- Another casually mentions that all her exes were “narcissistic sociopaths,” which statistically starts raising questions.
- Another reveals a “little surprise” under the skirt that probably should have been detected by the algorithm… but apparently even the app entered psychological denial.
- Another seems like the love of your life… until she says “I’m not really into reading” right after you mention a book.
- Another looks like she stepped out of a Vogue editorial… but makes Botox duck faces even in passport photos.
- Another looks flawless in every picture… because every photo has been taken from exactly the same angle since 2014.
- Another says she “channels spirits”… and you realize the relationship will probably include paranormal technical support.
- Another spends the entire date monitoring who viewed her stories as if the Nasdaq depended on it.
- Another interrupts dinner to film a TikTok because “the lighting is insane right now.”
- Another has hypnotic eyes… and the intellectual depth of a 2014 Facebook comment about cute kittens.
- Another seems mysterious… until you realize the mystery was simply complete absence of internal content.
- Another is so beautiful that you ignore every warning signal from your brain… until she starts explaining how vaccines altered her reptilian DNA.
- Another seems perfect… until she talks to the waiter like the sociopathic CEO of a cyberpunk megacorporation.
- Another transforms every casual conversation into an eight-hour Marxist-Leninist political debate without commercial breaks.
- Another takes medication to wake up, medication to sleep, medication for anxiety, medication for focus and probably medication to organize the other medications.
- Another is magnetic… and emotionally available like a Soviet server in 1983.
- Another arrives with five children, three schools, two litigious exes and a family logistics system worthy of NATO.
- Another has incredible chemistry with you… and apparently with half the app simultaneously.
- Another comes bundled with a mother-in-law so venomous she makes a rattlesnake look like a therapy animal.
- Another listens to voice messages on speaker mode at 2x speed in restaurants as if civilization were optional.
- Another seems incredibly cultured… until you discover she has never finished a single book without a Netflix adaptation.
- Another turns astrology into the official operating system of reality.
- Another seems perfect… until the first kiss reveals an entire anaerobic ecosystem living inside that halitosis.
- Another appears emotionally balanced… until she blocks you during disagreements “to protect her energy and realign her chakras,” as if basic communication were an advanced form of spiritual contamination.
- Another seems absolutely perfect… and that is exactly what scares you, because nobody appears that perfect without some hidden bug somewhere in the source code of reality.
- Another is a widow permanently trapped in emotional mourning, speaking about her late husband with such reverence and frequency that halfway through dinner you begin to suspect you’re not on a date, but accidentally participating in a paranormal tribute documentary.
- Another finally feels like “the one”… until you realize someone probably applied the 37% Rule to you as well.
- …
After hundreds of profiles, dozens of dates, endless conversations and an industrial quantity of swipes to the left and right, something strange begins happening inside the human brain.
Abundance stops feeling like freedom.
And starts feeling like statistical noise.
For the first time in human history, finding options stopped being difficult.
The difficult part became figuring out when to stop searching.
2. The Algorithmic Casino of Modernity
For almost all of human history, the fundamental problem of survival was scarcity.
There was not enough food.
Not enough information.
Not enough opportunities.
Not enough choices.
For thousands of years, most human beings were born, lived and died within a radius of only a few kilometers, knowing relatively few people and having extremely limited access to experiences, careers and lifestyles.
The human drama was simple:
how to obtain options.
The internet destroyed that problem.
And created another one infinitely stranger.
Today billions of people live surrounded by virtually infinite possibilities.
Dating apps offer thousands of potential partners sliding endlessly across a screen.
Streaming platforms offer millions of movies, series and videos that no human being could possibly finish within a single lifetime.
Food delivery apps transform even the choice of a hamburger into an existential crisis involving 247 restaurants, 83 flavors and contradictory reviews written by strangers with anime profile pictures.
Amazon converted consumption into a virtually inexhaustible catalog of objects you did not even know existed five minutes earlier.
LinkedIn transformed careers into permanent markets of professional comparison.
Social media continuously displays apparently more beautiful, more intelligent, happier, more productive and more interesting versions of other people’s lives.
The internet did not eliminate scarcity.
It created excess.
And perhaps this represents one of the greatest evolutionary shocks ever experienced by the human mind.
Our brains evolved inside small tribes, in environments with limited choices, relatively stable relationships and low social complexity.
Suddenly, they were thrown into a digital environment where there is always another option waiting immediately below the screen.
Another profile.
Another opportunity.
Another video.
Another apparently better person.
Another possible life.
Intuitively, this should make people happier.
But the opposite frequently happens.
The greater the number of possibilities, the greater the anxiety.
The harder it becomes to decide.
The greater the fear of choosing incorrectly.
The greater the post-choice regret.
Choice overload creates a strange psychological state in which nothing ever feels sufficiently good, because the mind remains obsessed with the possibility that something even better may still be hidden just ahead.
The result is a modern form of existential paralysis.
Human beings have never had so many options.
And perhaps they have never had such difficulty deciding what to do with them.
3. The Human Brain Was Not Built for Infinite Swiping
Dating apps amplify this problem brutally.
The user enters an algorithmic dopamine casino built on variable rewards and infinite novelty.
The next swipe might reveal someone:
- more attractive;
- more intelligent;
- more compatible;
- more interesting;
- less emotionally catastrophic.
And because the search never truly ends, the brain gradually shifts into a permanent optimization mode.
The person stops searching for connection.
And starts searching for statistical maximization.
The problem is that infinite systems deliberately destroy any feeling of completion.
Digital platforms are designed precisely to prevent you from feeling finished.
The goal is not for you to quickly find what you are looking for.
The goal is to keep you permanently trapped inside the loop:
scrolling screens,
swiping profiles,
watching videos,
comparing people,
consuming content,
generating data,
feeding algorithms.
Economically, your indecision is extraordinarily profitable.
You never feel like you have seen “enough.”
You never feel like you have explored all possibilities.
You never feel like you have found the best possible choice.
There is always the disturbing suspicion that perhaps the next profile might have been better.
The scroll never ends.
The catalog never ends.
The possibilities never end.
And precisely because of that, the search itself also never ends.
The human mind enters a permanent state of anxious optimization, trapped inside the illusion that the perfect choice might exist just one more swipe away.
And it is exactly here that mathematics enters the story.
4. The Mathematics of Indecision
There is an actual mathematical equation — rigorous and surprisingly powerful — created precisely for problems like this.
It belongs to a branch of mathematics known as Optimal Stopping Theory.
Decades before Tinder, Bumble or Hinge existed, mathematicians were already trying to formally solve a deeply human dilemma:
how do you make the best possible decision when options appear sequentially and rejected choices cannot be recovered?
The mathematical answer to that problem would become one of the most elegant and counterintuitive ideas ever produced by decision theory.
The theory suggests that if you want to maximize your chances of finding the best possible option within an irreversible sequence of choices, the optimal strategy is surprisingly simple:
reject approximately the first 37% of options,
learn from them,
and then choose the first option better than all previous ones.
Yes.
There is literally a mathematical formula for optimizing choices in dating markets, hiring processes, investments, career decisions, real estate purchases and even modern artificial intelligence systems.
The idea gained enormous contemporary popularity especially after the brilliant book Algorithms to Live By: The Computer Science of Human Decisions (2016), written by Brian Christian and Tom Griffiths.
The book became a modern classic precisely because it revealed something fascinating: many of the psychological, emotional and existential dilemmas of everyday life had already been studied decades earlier by mathematicians, computer scientists and decision theorists.
Algorithms originally designed to solve computational problems unexpectedly turned out to describe profoundly human behavior with remarkable precision.
Suddenly, seemingly chaotic questions such as choosing romantic partners, changing careers, answering emails, investing money or deciding when to stop searching began to resemble formal optimization problems.
And perhaps the most disturbing aspect is exactly this:
many of our modern emotional dramas possess the same logical structure as classical problems in computer science.
What we call romantic anxiety, professional indecision or “choice overload paralysis” is often not merely a psychological problem.
It is also an algorithmic one.
5. The Birth of the 37% Rule
The 37% Rule did not emerge from dating self-help books, motivational podcasts or productivity coaches.
It emerged from an extremely serious mathematical problem — and one almost cruel in its logic.
The problem began appearing formally during the 1940s and 1950s, amid the explosion of applied mathematics driven by the Second World War.
Governments massively invested in mathematicians, statisticians and physicists to solve concrete problems involving logistics, recruitment, cryptography, military planning and decision-making under uncertainty.
It was within this context that researchers began studying a deceptively simple question:
how do you make the best possible decision when options appear one at a time and cannot be recovered once rejected?
Among the researchers associated with the earliest formulations were Merrill Flood (1908–1991), Herbert Robbins (1915–2001) and John von Neumann (1903–1957), one of the central figures behind modern computing and Game Theory.
The problem would later become far more famous thanks to Martin Gardner (1914–2010), the legendary author of the Mathematical Games column in Scientific American.
Its most famous formulation became known as the Secretary Problem.
Imagine a company attempting to hire an executive secretary.
Candidates appear in random order.
You interview them one at a time.
Once rejected, a candidate disappears forever.
The moment you accept someone, the process immediately ends.
And the goal is not merely to hire someone “good.”
The goal is to maximize the probability of selecting the single best candidate among all possibilities.
That final condition makes the problem elegantly cruel.
It is not enough to find something satisfactory.
The objective is to find the best possible option under conditions of incomplete information and irreversible decisions.
Very quickly, mathematicians realized that this structure described something far more universal than corporate hiring processes.
It also modeled human relationships.
Thus emerged the so-called Marriage Problem.
Long before dating apps existed, mathematicians had already formalized the mathematics of infinite swiping.
6. The Mathematical Derivation of the 37% Rule
Friendly warning for readers traumatized by mathematics: the next sections contain derivatives, integrals, logarithms and potentially dangerous quantities of probabilistic suffering. If equations cause your brain to enter emergency mode, feel free to skip this part. For the survivors, however, we are now officially entering the moment where relationships begin to resemble a problem in differential calculus.
Now comes the most fascinating part of the entire theory:
understanding why the mathematics leads exactly to the number 37%.
The genius of the 37% Rule lies in the fact that it does not emerge from psychological intuition, dating advice or “app wisdom.”
It emerges directly from a rigorous mathematical structure connected to probability theory, decision theory and the so-called Optimal Stopping Problem.
What appears to be romantic anxiety, professional indecision or paralysis caused by excessive options slowly transforms into a formal optimization problem under uncertainty.
The central question is surprisingly simple:
how do we make the best possible decision when options appear sequentially and rejected choices cannot be recovered?
Imagine that there are N possible options.
This variable N represents the total number of available choices.
It may represent:
- job candidates;
- apartments visited;
- startups evaluated by investors;
- career opportunities;
- or dating app profiles.
And here an important detail emerges.
The larger the value of N, the harder the problem becomes.
Choosing between 5 options is relatively simple.
Choosing between 5,000 options begins generating cognitive overload, anxiety, permanent comparison and decision paralysis.
In other words:
an excess of possibilities dramatically increases the complexity of the problem.
The strategy behind the 37% Rule operates in two stages.
First, we deliberately reject the first r candidates.
This variable r represents the initial phase of observation and exploration.
But these first candidates do not exist for selection.
They exist to calibrate the brain.
During this stage, no decision is allowed.
You only observe, compare, learn and construct an internal statistical reference for what counts as:
- average;
- good;
- rare;
- exceptional.
After that, the second phase begins.
The rule changes completely.
The algorithm now works as follows:
“Choose the first option better than all previous ones.”
The mathematical question therefore becomes extremely precise:
what should the optimal value of r be?
In other words:
how many options should we reject before beginning to choose?
Now suppose that the absolute best candidate appears at position k.
Since we assume the order is completely random, every position has exactly the same probability of containing the best option.
Therefore:
P(k) = 1 / N
This means that the best option may appear at any point in the sequence with equal probability.
A useful analogy is a thoroughly shuffled deck of cards.
After shuffling, every position has exactly the same probability of containing the highest card.
But now comes the most important condition in the entire theory.
If the absolute best candidate appears too early, it will be lost.
Not because it is bad.
But because the system still lacks sufficient information to recognize its quality.
For the algorithm to work correctly, the best option must appear after the initial observation phase.
In other words:
k > r
During the initial phase, no choice is allowed — regardless of the quality of the observed option.
This is crucial.
The first candidates function as a statistical sample used to calibrate perception.
Without this initial learning phase, any decision would be based on insufficient information.
Now we analyze all candidates before position k:
1, 2, 3, …, k − 1
Among them exists a “best so far.”
And here appears one of the most elegant parts of the theory.
For the algorithm to successfully reach the true best candidate at position k, the previously best candidate must lie inside the rejected observation phase.
Otherwise, the algorithm would already have stopped earlier.
Imagine a concrete example.
Suppose:
N = 100;
r = 37;
and the absolute best candidate appears at position 50.
Now we examine the previous candidates:
1 through 49.
Among them there exists a “best so far.”
For the algorithm to reach candidate 50 without stopping earlier, that previous best candidate must belong to the first 37 rejected candidates.
Otherwise, if the previous best candidate appeared for example at position 42, the algorithm would already have stopped at candidate 42 and would never reach the true best candidate at position 50.
The probability that this previous best candidate lies inside the rejected observation phase is:
r / (k − 1)
At the same time, the probability that the absolute best candidate appears exactly at position k is:
1 / N
Multiplying these probabilities gives the probability of success specifically associated with position k:
(1 / N) * (r / (k − 1))
But the best option may appear in many possible positions after the observation phase.
It may appear at:
- r + 1;
- r + 2;
- r + 3;
- …
- all the way to N.
And each of those positions carries a small associated probability of success.
Therefore, the total probability that the strategy succeeds is the sum of all these individual probabilities.
And this is exactly where a mathematically profound structure emerges.
The resulting expression is not just any sum.
It belongs to one of the most famous families in all mathematics:
the Harmonic Series.
P(r) = (r / N) * ∑ [1 / (k − 1)]
k=r+1,…,N
At first glance, this appears to be merely a complicated sum of fractions.
But hidden inside this expression lies an extraordinarily elegant mathematical structure.
The symbol:
∑
represents a summation.
In other words, we are adding all possible success probabilities for every position in which the absolute best candidate may appear after the observation phase.
Expanding the sum explicitly, we obtain:
P(r) = (r / N) * [1/r + 1/(r+1) + 1/(r+2) + … + 1/(N−1)]
Now the structure begins to reveal itself clearly.
The term inside the brackets behaves exactly like a Harmonic Series.
A Harmonic Series has the general form:
1 + 1/2 + 1/3 + 1/4 + 1/5 + …
The fascinating aspect of the Harmonic Series is that it never stops growing — but it grows increasingly slowly.
Each new term contributes less than the previous one.
The growth continuously decelerates.
And this is precisely where one of the deepest and most elegant transitions in mathematics occurs.
For large values of n, the Harmonic Series behaves approximately like a natural logarithm.
More specifically:
1 + 1/2 + 1/3 + … + 1/n ≈ ln(n)
In other words:
a seemingly complicated sum of countless small terms can be approximated by an extraordinarily simple logarithmic function.
And this completely transforms the problem.
Because now we can replace a difficult discrete sum with a far more elegant continuous approximation.
Observe again the specific portion of the sum appearing in our problem:
1/r + 1/(r+1) + 1/(r+2) + … + 1/(N−1)
As N grows, this expression contains increasingly more terms while each individual term becomes progressively smaller.
The “steps” of the sum become so small and numerous that they begin behaving almost like a continuous curve.
And it is exactly at this point that the integral emerges.
Instead of manually summing thousands of tiny discrete blocks, we calculate directly the continuous area under the curve (the integral) of the function:
1/x
over the interval:
r ≤ x ≤ N
That is:
∫ (1/x) dx = ln(N) − ln(r)
Now we use a fundamental logarithmic property:
ln(a) − ln(b) = ln(a/b)
Therefore:
ln(N) − ln(r) = ln(N/r)
Substituting this result into the original expression, we finally arrive at the compact form of the probability function:
P(r) = (r / N) * [1/r + 1/(r+1) + 1/(r+2) + … + 1/(N−1)] ≈ (r / N) * ln(N / r)
And here something truly extraordinary happens.
All the combinatorial complexity of the problem — possible positions, conditional probabilities, sequences and irreversible decisions — collapses into an elegant function involving only:
- r;
- N;
- and a natural logarithm.
This function contains all the information necessary to answer the central question of the problem:
what should be the ideal size of the initial exploration phase?
And it is exactly this function that we will maximize using differential calculus to produce the famous result:
r ≈ 0.37N
Most impressively, the logarithm was not artificially inserted into the theory.
It emerges spontaneously from the mathematical structure of uncertainty, partial observation and irreversible decisions.
Harmonic series naturally grow like logarithms.
Therefore:
P(r) = (r / N) * ln(N / r)
This function describes the total probability of success of the strategy as a function of the size of the initial rejection phase.
Now we want to discover which value of r maximizes this probability.
Therefore, we search for the point where:
dP/dr = 0
And here enters one of the most important ideas in all differential calculus.
The derivative measures the instantaneous slope of a function.
It tells us whether the function is:
- increasing;
- decreasing;
- or momentarily “flat.”
Imagine the function P(r) as a mountain.
While climbing the mountain:
- the slope is positive;
- the derivative is greater than zero.
At the top of the mountain something special happens.
For a brief instant, the ascent stops before becoming descent.
The slope becomes exactly zero.
And that is precisely the point we want to find.
Because the top of the “mountain” represents the maximum possible probability of success.
Mathematically:
dP/dr = 0
means:
“find the point where the probability stops increasing.”
Now recall the function obtained earlier:
P(r) = (r / N) * ln(N / r)
This function contains two multiplied components:
(r / N)
ln(N / r)
Therefore we must use the so-called product rule of differentiation.
First we rewrite the function more conveniently:
P(r) = (r / N) * [ln(N) − ln(r)]
This is possible because:
ln(a/b) = ln(a) − ln(b)
Now we differentiate each term separately.
The derivative of:
r / N
is:
1 / N
because N is constant.
Meanwhile, the derivative of:
ln(r)
is:
1 / r
Applying the product rule carefully, we obtain:
dP/dr = (1 / N)[ln(N) − ln(r)] − 1/N
Now we factor out:
1/N
giving:
dP/dr = (1/N)[ln(N/r) − 1]
Finally we apply the maximum condition:
dP/dr = 0
Therefore:
(1/N)[ln(N/r) − 1] = 0
Since:
1/N ≠ 0
we simplify the expression:
ln(N/r) − 1 = 0
Therefore:
ln(N/r) = 1
And here emerges one of the most beautiful moments in the entire derivation.
To eliminate the logarithm, we apply exponentiation to both sides.
We know that:
if:
ln(x) = 1
then:
x = e
where:
e ≈ 2.71828
is the famous Euler number — one of the most important constants in all mathematics.
This number appears repeatedly in:
- exponential growth;
- compound interest;
- statistical physics;
- information theory;
- neural networks;
- machine learning;
- complex dynamics;
- radioactive decay;
- and probabilistic processes in general.
Almost mysteriously, it also emerges spontaneously from the problem of optimal choice.
Applying this to the equation, we obtain:
N / r = e
Now isolating r:
r = N / e
Since:
1/e ≈ 0.3679
we finally arrive at the result:
r ≈ 0.37N
And that is where the famous 37% Rule comes from.
Most impressively, the number 37% was not arbitrarily invented.
It emerges naturally from the mathematics of uncertainty, incomplete information and irreversible decisions.
Mathematics is revealing something profoundly human:
observing too little produces naive decisions.
But observing indefinitely also destroys the ability to decide.
There exists an optimal balance between exploration and commitment.
And surprisingly, that balance appears to lie near 37%.
Most fascinatingly, the famous 37% emerges spontaneously from the mathematical structure of uncertainty itself.
The final interpretation is profoundly elegant.
To maximize the probability of finding the absolute best option, we should spend approximately the first 37% of the process merely:
- observing;
- comparing;
- learning;
- calibrating perception.
After that, we choose the first option better than all previous ones.
If there are 100 profiles, the rule suggests:
observe the first 37 without choosing anyone.
After that:
choose the first profile clearly superior to all previous ones.
Curiously, exactly the same dilemma appears in modern artificial intelligence.
In reinforcement learning, artificial agents continuously face the conflict between:
- exploring new possibilities;
- or exploiting what already works.
The same mathematical problem reappears in:
- recommendation algorithms;
- search systems;
- financial investment;
- robotics;
- scientific exploration;
- and human behavior.
Mathematics is saying something profoundly human:
learn enough to recognize quality.
But do not wait for perfect information.
Because perfect information arrives too late.
7. The Cruelest Part of the Theory
The most brutal detail appears precisely at the end.
Even when using the mathematically optimal strategy — the best possible strategy allowed by the rules of the problem — you still succeed in choosing the absolute best option only about 37% of the time.
At first glance, this sounds absurd.
How can a strategy considered “optimal” still fail most of the time?
But this is exactly where the theory becomes profound.
Mathematics is revealing a fundamental limitation of reality itself:
when decisions must be made under uncertainty, incomplete information and irreversibility, perfection simply does not exist.
There is no algorithm capable of guaranteeing that you will always find the perfect partner, the perfect career or the perfect opportunity.
Real life does not provide total information.
It forces decisions before all data becomes available.
The 37% Rule does not promise certainty.
It merely offers the best statistical chance possible inside an inevitably imperfect universe.
In other words:
perfection is unattainable under uncertainty.
Only optimization is possible.
8. When the 37% Rule Fails
Despite all of its mathematical elegance, the theory operates under specific assumptions.
It assumes:
- random options;
- irreversible decisions;
- consistent comparisons;
- relatively stable objectives.
But real life is rarely so clean.
People change.
Preferences change.
Contexts change.
Relationships evolve.
Opportunities reappear.
Human compatibility is not one-dimensional like a simple mathematical problem.
In addition, modern algorithms manipulate behavior.
Apps influence which profiles appear.
Recommendation systems alter perception.
Social feedback changes preferences.
In other words:
human reality is far more chaotic than the original mathematical model.
And precisely because of that, the 37% Rule should not be interpreted as a magical formula.
It works better as a powerful mathematical lens.
An elegant way of perceiving deep patterns hidden inside modern anxiety, decision-making and the human condition itself.
9. Between Mathematics, Love and Uncertainty
Perhaps the most fascinating aspect of the entire theory is philosophical.
The 37% Rule reveals something profoundly true about the human condition:
waiting too long destroys opportunities.
But acting too early destroys possibilities.
There exists a delicate balance between:
- curiosity;
- learning;
- exploration;
- experience;
- commitment.
And most surprisingly, that balance appears to emerge spontaneously from mathematics itself.
At its core, the entire strategy can be summarized in an almost absurdly simple way.
First observe.
Then choose.
But do not observe forever.
Because at some point the search for the perfect option becomes exactly the mechanism preventing any real choice from happening at all.
Perhaps the true modern tragedy is not the lack of options.
Perhaps it is dying without ever managing to stop searching.
#37PercentRule #DecisionTheory #ArtificialIntelligence #DatingApps #Philosophy #Algorithms #AI #Psychology #AITalksOrg
10. Recommended Reading
- Christian, Brian, and Tom Griffiths. 2016. Algorithms to Live By: The Computer Science of Human Decisions. New York: Henry Holt and Company.
- Bruss, F. Thomas. 2000. “Sum the Odds to One and Stop.” The Annals of Probability 28, no. 3: 1384–91.
- Ferguson, Thomas S. 1989. “Who Solved the Secretary Problem?” Statistical Science 4, no. 3: 282–89.
- Ferguson, Thomas S. 2008. Optimal Stopping and Applications. Mathematics Department, University of California, Los Angeles (UCLA). ucla.edu.
- Flood, Merrill M. n.d. Early unpublished formulations and discussions of the Secretary Problem and optimal stopping strategies in operations research (1940s–1950s). Historical reference.
- Gardner, Martin. 1956–1981. “Mathematical Games” (columns). Scientific American.
- Kahneman, Daniel. 2011. Thinking, Fast and Slow. New York: Farrar, Straus and Giroux.
- Robbins, Herbert. mid-20th century. Foundational probabilistic work related to optimal stopping and sequential decision theory. Historical reference.
- Sapolsky, Robert. 2017. Behave: The Biology of Humans at Our Best and Worst. New York: Penguin Press.
- Schwartz, Barry. 2004. The Paradox of Choice: Why More Is Less. New York: Harper Perennial.
- Shannon, Claude. 1950. “Programming a Computer for Playing Chess.” Philosophical Magazine 41, no. 314: 256–75.
- Silver, David, Aja Huang, Chris J. Maddison, Arthur Guez, Laurent Sifre, George van den Driessche, Julian Schrittwieser, et al. 2016. “Mastering the Game of Go with Deep Neural Networks and Tree Search.” Nature 529, no. 7587: 484–89.
- Simon, Herbert A. 1955. “A Behavioral Model of Rational Choice.” The Quarterly Journal of Economics 69, no. 1: 99–118.
- Simon, Herbert A. 1982. Models of Bounded Rationality. Cambridge, MA: MIT Press.
- Sutton, Richard S., and Andrew G. Barto. 2018. Reinforcement Learning: An Introduction. 2nd ed. Cambridge, MA: MIT Press.
- Von Neumann, John, and Oskar Morgenstern. 1944. Theory of Games and Economic Behavior. Princeton: Princeton University Press.
Copyright 2026 AI-Talks.org
