You’re probably reading this because some piece of code decided you should. That’s the reality. Whether you’re scrolling through a feed, checking your bank balance, or just trying to figure out why your GPS took you through a weird alleyway, you are interacting with the invisible hand of modern logic. But honestly, if you ask five different people for the definition of algorithm, you’ll get five different answers ranging from "it’s basically magic" to "it’s a recipe for cookies."
Both are kinda right. And both are mostly wrong.
An algorithm isn't a mysterious sentient ghost living in your phone. It’s a tool. Specifically, it is a finite set of unambiguous instructions that, if followed, accomplish a particular task. Think of it as a bridge between a problem and a solution. If the problem is "I’m hungry" and the solution is "a sandwich," the algorithm is the specific sequence of steps you take to get from point A to point B. If you skip a step—like forgetting the bread—the algorithm fails.
Why the Definition of Algorithm is Actually Simple (And Why We Overcomplicate It)
At its heart, the definition of algorithm is just a "how-to" guide for a machine. While we associate the word with high-tech giants like Google or Meta, the concept is ancient. The word itself comes from the name of a 9th-century Persian mathematician, Muḥammad ibn Mūsā al-Khwārizmī. He wasn't thinking about TikTok trends. He was thinking about algebra and how to solve equations systematically.
Computers are incredibly fast, but they are also incredibly stupid. They don't have "vibes." They can't "figure it out" if you give them vague directions. If you tell a human to "make a PB&J," they know what you mean. If you tell a computer, you have to define what a knife is, how to grip it, what "spread" means, and what happens if the jar is empty. That granular, step-by-step logic is what defines an algorithm.
It has to be discrete. It has to have an end. A loop that never finishes isn't a helpful algorithm; it's a crash.
The Ingredients of a Functional Algorithm
Every algorithm needs a few specific traits to actually be considered one. First, it needs input. This is the data you feed it. Then it needs "definiteness." This means every step is clear and has only one possible interpretation. If a step says "add some salt," that's a bad algorithm. "Add 5 grams of salt" is an algorithm.
Then there's the output. This is the result of all that work. Finally, it must be effective. All the operations must be basic enough that they can be performed exactly in a finite amount of time. You can't have a step that requires "waiting for eternity."
The Algorithms You Encounter Every Single Day
Let’s look at something like the PageRank algorithm. When Larry Page and Sergey Brin started Google, they didn't just want a list of websites. They wanted to know which ones were actually good. Their algorithm decided that a link from one site to another was essentially a "vote" of confidence. But not all votes were equal. A vote from a reputable site like The New York Times weighed more than a vote from your cousin’s Geocities page.
✨ Don't miss: Wait, What is Log of e Anyway? The Math Behind Everything
That’s a sorting algorithm. It takes a massive, messy input (the entire internet) and produces an ordered output (your search results).
Then you have recommendation engines. These are the ones people usually mean when they say "the algorithm is out to get me." Netflix uses these to predict what you want to watch. It looks at your history, compares it to millions of other users with similar tastes, and calculates a probability. It’s not reading your mind. It’s just doing math on your past behavior.
- Sorting Algorithms: Like putting a deck of cards in order.
- Search Algorithms: Finding a specific needle in a haystack of data.
- Encryption Algorithms: Turning your password into a garbled mess so hackers can’t read it.
- Compression Algorithms: Making a huge video file small enough to stream over your crappy Wi-Fi.
How Algorithms Go Wrong: The Bias Problem
Here is the thing about the definition of algorithm that gets ignored: they aren't objective. Because humans write the instructions and humans choose the data, human bias gets baked into the code.
In her book Algorithms of Oppression, Dr. Safiya Umoja Noble highlights how search engine results can reinforce negative stereotypes. If the data used to "train" an algorithm is skewed, the output will be skewed. This isn't a glitch. It's a reflection of the input.
We see this in facial recognition tech, which historically has struggled with non-white faces because the initial datasets were heavily weighted toward white subjects. We see it in credit scoring and even in resume-filtering software. If an algorithm is told to find candidates who "look like our previous successful hires," and all those hires were men, the algorithm will naturally start discarding female applicants. It's just following orders. It doesn't know it's being sexist; it just knows it's matching a pattern.
Heuristics vs. Algorithms
Sometimes, we use "heuristics" instead of strict algorithms. A heuristic is a mental shortcut. It’s a "rule of thumb" that usually works but isn't guaranteed to be perfect. If you're lost in a forest, a heuristic might be "walk downhill until you find water." It might work. An algorithm would be a 500-page manual covering every possible tree and rock formation to guarantee an exit.
In tech, we use heuristics when a perfect algorithm would take too long to compute. Finding the absolute shortest path for a delivery driver in a city with a thousand stops is computationally "expensive." So, we use an algorithm that finds a "good enough" path very quickly.
The Future: Machine Learning and "Black Box" Logic
Traditionally, a programmer wrote every single line of an algorithm. "If X happens, do Y."
But now we have Machine Learning (ML). In ML, the algorithm is essentially "an algorithm that builds other algorithms." You give the computer a goal and a massive pile of data, and it figures out the steps itself. This is where things get spooky. Even the engineers who built the system sometimes can't explain exactly why a deep-learning model made a specific decision. This is the "black box" problem.
When an AI identifies a tumor in a medical scan better than a doctor can, it’s using a complex algorithm it refined through millions of examples. We can see the input and the output, but the middle part—the actual logic—is a web of billions of weighted connections. It still fits the definition of algorithm, but it’s no longer a simple recipe humans can easily read.
✨ Don't miss: LG 4K and OLED TV: What Most People Get Wrong About Choosing a Screen
Practical Insights: Taking Control of the Code
You aren't just a passive victim of these systems. Once you understand that an algorithm is just a feedback loop, you can start to "train" the ones you interact with.
- Clean your inputs: If your YouTube feed is full of garbage, it's because you clicked on garbage. Use "Not Interested" buttons. It resets the data point.
- Diversify your data: Algorithms love "filter bubbles." They want to show you what you already like. Manually search for opposing viewpoints to break the cycle.
- Check the source: If an algorithm gives you an answer (like through a chatbot), ask for the "why." If it can't explain its logic, be skeptical.
- Privacy is a lever: By limiting the data you give (through VPNs or ad-blockers), you reduce the "input" the algorithm has to work with, making its "output" less targeted.
Moving Forward
The definition of algorithm will continue to evolve as we move into more advanced AI. But remember: at its core, it’s just a sequence. It’s a tool for managing complexity. Understanding how those sequences are built is the first step toward not being controlled by them.
The next time you see a "suggested post" that feels a little too personal, don't think of it as a spooky coincidence. Think of it as a set of mathematical instructions that saw a pattern in your life. You have the power to change that pattern. Start by being more intentional with the data you "feed" the machines daily. Check your privacy settings on your most-used apps today; look for the "ad personalization" or "data tracking" toggles and turn off what you don't need. This small act changes the input, which fundamentally changes the algorithm's power over your digital experience.