Comparing Humor Theories

Philosophers have been arguing about why things are funny since before anyone had a podcast. The good news: they are all right. The better news: they are all right about different parts of the same joke.

Here is the breakdown.

1. Superiority

Thinkers: Plato, Hobbes

Core Question: Why do we laugh at others?

Big Idea: Laughter can arise from feeling above weakness, folly, or failure.

The Joke:

A pharmacist looks at a prescription and says, "I can't fill this. This name isn't real." The customer says, "I legally changed it." The pharmacist says, "To Laughing?" The customer says, "To Dr. Laughing."

Hobbes laughs because he feels briefly superior to both of them. Plato is concerned this is happening at all. The pharmacist does not laugh. The customer does — which is where this theory starts to fall apart and theory two picks up.

2. Incongruity

Thinkers: Kant, Schopenhauer, Suls

Core Question: Why does surprise make us laugh?

Big Idea: Humor appears when expectation collides with an unexpected result.

The Joke:

A philosopher legally changes her name. Her colleagues ask: "To what?" She says: "Laughing." They say: "That is not a philosophical position." She says: "It sure is. I upped my attitude, now up yours! 

The collision is everything here. Sophy (wisdom) plus Laughing — two things with no obvious business sharing a last name — meet at a pharmacy counter, a philosophy conference, and an international executive boardroom. Every single time, the incongruity does the work. Kant nods slowly. Schopenhauer finds this mildly less bleak than usual, which for him counts as delight.

3. Relief

Thinkers: Spencer, Freud

Core Question: Why does laughter feel like release?

Big Idea: Jokes discharge pressure created by tension, repression, or restraint.

The Joke:

A CEO walks into a board meeting, delivers a quarterly report across four continents, navigates stakeholder tensions, manages an infrastructure crisis, and then goes home and writes a joke about waffles.

The next morning: 40 million pageviews.

Spencer would say the nervous energy had to go somewhere. Freud would say the waffles are not about waffles. They are both onto something. Fifteen years of philosophical humor published alongside an executive career is not a hobby. It is a pressure release system with a search bar and a sidebar. The blog is the valve. The jokes are the steam. The waffles are — fine, Freud, have this one — probably not just waffles.

4. Social Correction

Thinker: Bergson

Core Question: Why do we laugh at rigidity?

Big Idea: Laughter corrects stiffness, repetition, and mechanical behavior.

The Joke:

A guru walks onto the internet. He has a certification, a retreat package, a crystal subscription, and a downloadable guide to manifesting abundance for $297. He has not manifested the irony.

Happy Thoughts Travel Fast writes the article. The internet laughs. The guru updates his pricing.

Bergson's whole theory is that we laugh when a living thing behaves like a machine — when a person becomes so predictable, so scripted, so rigidly formatted that the flexibility of actual humanity disappears. The Guru Crime Syndicate post did not need to explain this. It demonstrated it. That is the blog in one article: not a lecture about rigidity, but a joke that corrects it. Bergson approves. He is French. Approval is not his default setting. Note accordingly.

Takeaway

Different theories explain different parts of the laugh.

Superiority explains why the pharmacist story is funny to everyone except the pharmacist. Incongruity explains why the name works at all. Relief explains why a CEO writes jokes about waffles at midnight and why that is, actually, a completely rational decision. Social correction explains why the Guru Crime Syndicate has never gone out of style.

One laugh. Four theories. Fifteen years of evidence.

The blog has been running the experiment the whole time.

---

Happy Thoughts Travel Fast | happythoughtstravelfast.com

The Laughing Timeline: What Philosopher's Said When They Found HTTF

What the philosophers said when they found the blog.

Two thousand five hundred years of philosophers thinking very seriously about humor. One blog. Fifteen years of evidence.

We checked the comments section. Here is what they said.

1. Plato (moral risk)

"I must warn you: humor is dangerous. It bypasses reason, inflames the passions, and undermines the orderly soul. I read every post twice. Do not tell Aristotle."

2. Aristotle (comic virtue)

"Plato told me. I have reviewed the archive and I am pleased to report that the humor here occupies the precise mean between buffoonery and boorishness. Four stars. I would have preferred a syllogism in the sidebar."

 

3. Cicero & Quintilian (persuasion)

"We have studied Happy Thoughts Travel Fast extensively and wish to confirm: this is rhetoric. The jokes are doing argument. We taught this. You are welcome. Also, the About page could be longer. Much longer. We are available."

4. Hobbes (superiority)

"Laughter is the sudden glory arising from the perception of some eminence in ourselves compared to others. I laughed at the pharmacist story. I am not elaborating further. Life is solitary, poor, nasty, brutish, and short, but the blog is a reasonable use of the time."

5. Kant & Schopenhauer (incongruity)

Kant: "The humor here arises from the sudden transformation of a strained expectation into nothing. I find this philosophically precise and will now spend forty pages explaining why."

Schopenhauer: "I was going to leave a comment but Kant is still typing."

 

6. Spencer & Freud (relief)

Spencer: "The nervous energy released by these jokes is entirely consistent with my hydraulic model of laughter. Very efficient. Very healthy."

Freud: "It is not about the nervous energy. It is about the name. The name is doing something. I have a theory. It involves your father."

7. Bergson (social correction)

"Laughter is society's corrective — a gentle punishment for mechanical rigidity. The Guru Crime Syndicate post alone corrected approximately forty-seven social rigidities. I counted. I am French. I had time."

8. Shaftesbury (truth-testing)

"Only that which can survive ridicule is truly true. I have subjected every article on this site to the test of ridicule and found the philosophy intact. The jokes, it turns out, are load-bearing. Remarkable. I am leaving five stars and a small portrait."

 

9. Wittgenstein & Austin (language-games)

Wittgenstein: "The meaning of a joke is its use. I said this first. I said this in 1953. I left. Someone showed up and proved it with fifteen years of public record and a legally changed name. I have no further notes. This is the note."

Austin: "How to do things with words. She did things with a name. I genuinely did not see that coming and I wrote the book."

10. Ethics, Teaching, and Human Flourishing (boundaries and learning)

The tree does not leave a comment.

The tree has been here the whole time — in the Stick Figures with their big red hearts, in Raising Funny Kids, in the Humor Challenge, in the classroom, in the executive boardroom, in the pharmacy where someone demanded three forms of ID before accepting that wisdom and laughter could legally share a last name.

The tree is the blog.

The blog was always the comment.

(Plato liked this post. Aristotle gave it four stars. Hobbes did not elaborate. Freud is still typing.)

---

Takeaway: Humor travels from moral caution to human flourishing.

Took about 2,500 years. Worth it.

---

Happy Thoughts Travel Fast | happythoughtstravelfast.com

 

Support Vector Machine: The Algorithm That Needs Personal Space

I separated my laundry 🧺 into whites and colors. Then I found one red sock 🧦 in the white pile, and the algorithm called it a hostile data point. 📉 

That is a Support Vector Machine.

A Support Vector Machine is a machine learning algorithm used for classification. It tries to separate data into groups by drawing the best possible boundary between them.

Imagine a pile of laundry. 🧺 

On one side: white shirts. 

On the other side: colorful clothes. 👚 

The goal is to draw a clear line between the two groups.

Easy enough.

Then the red sock appears. 🧦 

The red sock is not where it should be. It is sitting dangerously close to the white shirts, looking innocent while preparing to turn everything pink.

This is where SVM gets serious.

SVM does not just draw any line between groups. It looks for the boundary with the widest possible margin.

The margin is the empty space between the boundary and the nearest examples from each group.

In plain terms, SVM asks:

How can I separate these groups while leaving the most room on both sides?

That room helps.

A narrow boundary is risky. One weird example can ruin the classification.

A wide boundary is stronger. It gives the model breathing room.

The closest data points to the boundary are called support vectors. They are the examples that we focus on most because they define where the boundary goes.

In the laundry example, the red sock is absolutely a support vector. 🧦 

A dramatic one.

Possibly a needy garment. 

SVM works well when the goal is to separate things clearly:

Spam or not spam.

Cat or dog.

Safe or risky.

White laundry or laundry about to become a group project.

Sometimes the data can be separated with a straight line.

Sometimes the data is messier, so SVM uses a trick called the kernel trick. That lets it transform the data into a higher-dimensional space where separation becomes easier.

That sounds complicated, but the idea 💡 is simple.

If two groups are tangled on a flat table, lift the problem into another dimension and the separation may become obvious.

It is basically saying:

“This problem is annoying in 2D. Let’s give it a balcony.”

The strength of SVM is that it focuses on the boundary. It is especially useful when the difference between categories depends on a few important examples near the edge.

Its weakness is that it can be harder to interpret than simpler models, and it may struggle when there is too much noise or too many overlapping categories.

Because sometimes the sock is not clearly red.

Sometimes it is burgundy.

Sometimes it has white stripes.

Sometimes it has been through enough laundry cycles that nobody knows what it believes anymore.

Support Vector Machines are useful because many decisions depend on finding a clean dividing line.

They do not ask, “What is the average thing here?”

They ask:

Where is the safest boundary between these groups?

That is why SVM feels so precise.

It is not just separating laundry.🧺 

It is protecting the white shirts from chaos.

And honestly, any algorithm willing to stand between a white blouse and one suspicious red sock deserves respect. ✊🏻 

Wittgenstein's Humor Challenge

In 1953, Ludwig Wittgenstein supposedly wrote: "A serious and good philosophical work could be written consisting entirely of jokes."

Then he died and left the whole thing for someone else to figure out. Very philosophical of him.

Enter: The Experiment

In 2011, a philosopher changed her name to Sophy Laughing, started a blog called Happy Thoughts Travel Fast, and spent fifteen years writing, drawing, teaching, joking, and philosophizing in public — across four continents, in a serious executive career, in front of an audience of millions who arrived mostly through Google image search and stayed for the jokes.

Wittgenstein: issued the challenge. Sophy Laughing: accepted, executed, filed the paperwork, and submitted it to a philosophy journal.

Imaginary duel. One round. One winner. She did not even break a sweat.

What Wittgenstein Actually Meant

(Because he does deserve a brief moment of credit before we move on)

His big idea — the one that makes this whole thing work — is that meaning is not in definitions. It is in use. A word does not point at a thing. A word does something, in a context, with a speaker, in a situation.

Same goes for jokes.

A joke is not a setup plus a punchline. A joke is what happens between two people when the punchline hits home — or does not. The meaning is in the doing.

Which means you cannot explain humor in the abstract. You have to show it in use.

Challenge accepted.

Humor in Practice: Four Things You Should Know

1. Context Meaning changes with situation, speaker, and setting. The exact same sentence is either a joke or a resignation letter depending on who says it, to whom, and on which Zoom call. Context is not background. Context is the whole thing.

2. Timing A joke works through sequence, pause, and release. The pause is load-bearing. Remove it and you have a statement. Keep it and you have comedy. This is why reading jokes out loud to someone who did not ask is a relationship risk.

3. Audience Humor depends on shared recognition and response. The joke requires someone who gets it. Without that shared recognition, the funniest thing ever written is just a weird sentence. The audience does not passively receive the joke. The audience completes it by laughing.

4. Meaning What a joke means appears in what it does. Did it relieve tension? Did it expose an assumption? Did it make someone feel seen? Did it make a room of strangers feel briefly like a team? That — whatever just happened — is the meaning. Not the words. The effect.

The Takeaway

Wittgenstein said a serious philosophical work could be written in jokes.

The experiment at Happy Thoughts Travel Fast went one further: it was lived in jokes. Fifteen years of showing humor in use — in writing, in drawing, in pedagogy, in professional life, in a new name that required three forms of ID at a pharmacy.

The challenge was answered. The answer was a blog. The blog became a philosophy journal submission.

Ludwig would have found that very funny (if he had a sense of humor). 

---

Happy Thoughts Travel Fast | happythoughtstravelfast.com

Spam Text Messages: The Algorithm Behind Why Your phone Won’t Stop Buzzing

Your phone receives ten texts claiming you have an unpaid toll, a mysterious package, or a gift card waiting. By the eleventh message, even your spam filter is tired.

For many people, spam texts have become a daily event. You might be checking a message from a friend when, moments later, someone is urgently informing you that your account has been suspended, your package cannot be delivered, or you have won a prize you definitely did not enter to win.

Spam texts usually try to do one of three things: get your attention, get your information, or get your money.

Some messages pretend to come from banks, delivery companies, or government agencies. Others skip the formalities and go straight to offering miracle investments, suspicious job opportunities, or rewards that somehow require a payment first.

The volume of these messages has become impressive. Many people receive multiple spam texts every week, while others receive several every day. Although the details vary, the messages often follow familiar patterns. They create a sense of urgency, ask you to click a link, and warn that something terrible will happen if you do not act immediately.

Congratulations—you have apparently missed a package for the seventh time this month.

However, if you look closely, the clues are usually there. The sender is unfamiliar, the link looks strange, the grammar is questionable, and the story often makes little sense.

Spam works because scammers know that even a tiny response rate can be profitable. If enough messages are sent, someone will eventually click. A text that creates panic can cause people to act before thinking, and a message that appears official can seem trustworthy.

The strength of spam campaigns is their scale. Sending millions of messages costs very little, which makes these campaigns persistent. Their weakness, however, is that many of the messages are easy to recognize once you understand the common patterns.

A legitimate company rarely demands immediate action through a suspicious link. Likewise, a government agency is unlikely to contact you through a random text message filled with spelling errors. And no, the mysterious reward waiting for you is probably not real.

The key is simple. Pause before clicking, verify information through official websites, ignore unexpected links, and report obvious scams when possible.

Modern spam filters help, but scammers constantly change tactics. They use new phone numbers, rewrite messages, and imitate trusted organizations. That is why awareness matters.

The practical rule is straightforward: treat unexpected texts with skepticism, verify information before responding, and never assume that a message is legitimate simply because it arrived on your phone.

That is how you avoid turning a fake delivery notification into a very real headache.

Dimensionality Reduction: Simplifying Data

I asked whether the banana was still good. The model checked color, spots, smell, firmness, emotional history, and banana bread potential. Then it said, “Let’s simplify this.”

That is dimensionality reduction.

Dimensionality reduction is a machine learning technique that takes data with many features and reduces it to fewer features while preserving the most important structure.

At its core, the question is:

Can we make something complicated easier to understand without losing what matters?

Imagine we are judging bananas.

A banana might be described by many features, including:

• Color
• Number of brown spots
• Firmness
• Smell
• Ripeness
• Bruises
• Peel texture
• Likelihood of becoming banana bread

That is a surprising amount of information for a fruit that mostly wanted a quiet life.

Dimensionality reduction takes all of those features and represents them using fewer dimensions.

Instead of tracking eight separate banana traits, the model might summarize everything along two useful axes:

• Fresh enough to eat
• Ready for banana bread

With that simplified view, the banana landscape becomes much easier to understand:

• Green bananas cluster in one region.
• Perfect yellow bananas occupy another.
• Brown, dramatic bananas gather near the “please bake me immediately” zone.

The goal is not to throw away information. The goal is to preserve the structure that helps us understand the data.

This becomes especially valuable when datasets contain too many dimensions for humans to visualize comfortably. Most people can interpret a two-dimensional chart, and three dimensions are manageable if nobody gets too ambitious. Once a dataset contains fifty, five hundred, or five thousand features, however, our brains tend to give up and start looking for snacks.

Dimensionality reduction helps by transforming high-dimensional data into something we can inspect, plot, and reason about.

One of the most common techniques is PCA, or Principal Component Analysis.

PCA identifies the directions in which the data varies the most and keeps those directions while discarding less informative detail. You can think of it as walking into a messy room and asking:

What are the main patterns here?

Not every sock needs a biography.

Dimensionality reduction works because features do not all contribute equally. In many datasets:

• Some features overlap with one another.
• Some are mostly noise.
• Some contribute only minor information.
• Some seem to be standing around with a clipboard and no clear purpose.

For bananas, color and ripeness may communicate much of the same information, while smell and banana bread potential may also be closely related. The model can compress those relationships into a simpler representation.

The real strength of dimensionality reduction is its ability to reveal hidden structure. By reducing complexity, it can make the following easier to spot:

• Clusters
• Patterns
• Outliers
• Relationships between observations

Its main weakness is that simplification inevitably removes some detail. When many features are compressed into two or three dimensions, something gets left behind.

Sometimes that trade-off is perfectly acceptable. Sometimes the missing detail matters.

A banana may look perfectly reasonable on the chart while still hiding one suspicious soft spot that quietly says, “I have made choices.”

So the practical rule is straightforward:

• Use dimensionality reduction when the data is too complex to view clearly.
• Remember that a simplified map is still a map, not the entire territory.

Dimensionality reduction does not make the banana simpler.

It makes the banana easier to understand.

And honestly, any algorithm that can look at a dramatic fruit and conclude, “This is mostly a banana bread situation,” has earned its place in machine learning.