algorithm
in English: algorithm
An algorithm is a finite sequence of unambiguous steps that, when followed precisely, yields a result within finite time for any allowed starting data of a given task type—whether executed by a human or a computer.
In plain terms
An algorithm is like a good recipe: it tells you what to start with, what to do in what order, and when you are finished. The difference is that a recipe allows some freedom ('salt to taste'), whereas an algorithm does not: every step is unambiguous so that a machine can execute it without thinking.
Every computer program is built from algorithms. An algorithm sorts your contacts alphabetically, an algorithm finds the shortest route on the map, and an algorithm decides which video you see next.
In artificial intelligence, the word can mean two things. In a traditional algorithm, a human writes the rules step by step. In machine learning, the human writes only the learning method, and the system derives the specific rules itself from a huge number of examples. This is why it is harder to say exactly why it decided as it did.
In the news, 'the algorithm' often refers to an entire decision-making system, such as a social media feed or a bank's credit assessor. In such cases, it is actually about a combination of many algorithms, data, and business objectives.
How it works
- Input: the algorithm receives some initial data, such as a list of names, two addresses, or a search term.
- Steps: executes predefined, clear operations in a specific order, such as comparing, calculating or swapping.
- Branches and repeats: depending on conditions, it takes different paths ("if larger, swap") and repeats certain steps for as long as needed.
- Stop and output: after a finite number of steps, it stops and gives a result, such as the sorted name list, the route, or the search results list.
- In a learning variant, the machine learning algorithm's output is not the direct answer but a trained model that later responds to new cases.
An everyday example
When your phone's map plans your route home, a routing algorithm compares countless possible paths in seconds and selects the fastest based on traffic. When a streaming service recommends a film in the evening, another algorithm tries to guess what you would like based on your previous choices.
Who is affected
A common misconception
Limits and risks
- The algorithm does exactly what it was told: if the goal is wrong or the input data is flawed, it confidently delivers a bad result.
- The decisions of learning algorithms are often difficult for their creators to explain simply, making them hard to challenge or verify.
- Recommendation and ranking algorithms are mostly tuned to retain your attention or generate revenue, not necessarily to serve your interests.
- There are tasks for which we do not know a fast enough algorithm, and even some for which it is proven that no algorithm exists at all.
Similar, but not the same
- program – The algorithm is the solution method; the program is its specific, executable implementation written in a programming language.
- modell – The learning algorithm is the process by which the machine learns, while the model is the result of that learning, which later provides the answers.
- artificial intelligence – Every AI system is built on algorithms, but most algorithms—for example, sorting or interest calculation—are not artificial intelligence.
Related concepts
computer science, informatika
learning algorithm, recommendation algorithm, search algorithm, sorting algorithm, encryption algorithm
program, modell, machine learning, artificial intelligence, adat
A brief history
The word originates from the Latinized form (Algoritmi) of the name of al-Khwarizmi, a Persian mathematician who worked in Baghdad in the 9th century. The idea itself is much older: Euclid already described the steps for calculating the greatest common divisor in antiquity. The precise mathematical definition of the concept was established by the 1936 works of Alan Turing and Alonzo Church.
Where things stand
As of late 2026, algorithms are present in every area of digital life, from search engines and banks to AI agents. Public debate and regulation increasingly focus on their transparency and accountability. In the EU, the Digital Services Act (DSA) does this for recommender systems, and the AI Act for systems that use artificial intelligence.
In the news
- Meta, Google and TikTok sue California over law protecting children from addictive algorithmsNov 5, 2026 · Jog
- FDA classifies digital cervical cytology slide imaging systems that use AI algorithmsNov 4, 2026 · Health
- Two federal courts dismiss AI and keyword hiring bias claims; Sirius XM also wins dismissalOct 27, 2026 · Work & HR
- California's new AI 'bossware' and 'No Robo Bosses' laws: what employers need to doOct 27, 2026 · Jog
- One in five European employees digitally monitored as algorithmic management spreads, Eurofound saysOct 26, 2026 · Work & HR
- Lancet series draws on data from nearly nine million young people to warn about digital platformsOct 25, 2026 · Education & Family
- Polish AI regulator's decisions to be reviewed by SOKiK under new civil procedure sectionOct 23, 2026 · Jog
- Law Commission outlines plan to overhaul product liability for AI, apps and algorithmsOct 22, 2026 · Jog
Also known as
algoritmus · algoritmusok · algorithm · algorithms · algo · procedure · computational procedure · az algoritmus
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.