← Back to catalogue
Published

algorithm

vr.tr.algorithm · thing-q8366

Let an agent handle algorithms by problem, steps, correctness, complexity and responsible use.

Thing Registry Information and virtual systems INF.MED
Bundle → Layer → Finding → QuestionsClassifiersWhat it isWhy it existsDistinguishing featuresWhat robots and AI may and may not doMoral aspectsOwnersLinks to other meta-modelsIdentity and identifiersDirect propertiesRecognitionCapabilities and actionsHazards and failure modesStandards and interfacesContext of useSources

Bundle → Layer → Finding → Questions Filled

4 bundles · 8 layers · 8 findings · 16 questions

Problem What it solves.

Task

Input and output.

Task

Problem specification.

  1. What problem does the algorithm solve, with what input and output? definition
  2. What assumptions does it make? boundary

Family

Kind.

Family

Algorithm family.

  1. Which family does it belong to? definition
  2. Is there a known better alternative? action
Analysis Correctness and cost.

Correctness

Proof.

Correctness

Correctness.

  1. How is it shown to be correct? provenance
  2. What edge cases matter? boundary

Complexity

Time and space.

Complexity

Complexity.

  1. What are its time and space complexity? measurement
  2. How does it behave in practice? measurement
Implementation Code.

Code

Program.

Code

Implementation.

  1. Where is a reference implementation? provenance
  2. How is it tested? provenance

Performance

Benchmarks.

Performance

Benchmarks.

  1. How does it perform on real data? measurement
  2. On which benchmark? provenance
Impact Use on people.

Decisions

Automated decisions.

Decisions

Automated decisions.

  1. Does it make or support decisions about people? boundary
  2. Is there human review? boundary

Fairness

Bias.

Fairness

Fairness.

  1. Has it been tested for bias? provenance
  2. How can affected people contest outcomes? action

Classifiers Filled

Family
Thing Registry
Category
Information and virtual systems
Entry kind
thing
Plane
INF
Domain
INF.MED
Other names and narrower kinds
facial recognition payment system, parser, packrat parser, recursive-descent parser, GLL parser, LL(*) parser, XSLT processor, chart parser, bottom-up parser, recursive ascent parser, Simple precedence parser, text parser

What it is Filled

A finite, well-defined sequence of instructions for solving a problem or performing a computation, such as sorting, searching, parsing or encryption.

Why it exists Filled

Let an agent handle algorithms by problem, steps, correctness, complexity and responsible use.

Distinguishing features Filled

  • Finite and well-defined
  • Correctness and complexity can be analysed
  • Independent of programming language
  • Automated decision systems raise fairness issues

What robots and AI may and may not do Filled

Must not

  • Present an algorithm as correct, fair or secure without evidence or testing.
  • Use an algorithm to make decisions about people while hiding that it does so.
  • Copy a proprietary or patented algorithm's implementation in breach of its licence.
  • Apply an algorithm to people's data for a purpose they did not consent to.

Only with a human decision

  • Deploying an algorithm that decides on people's access to credit, jobs, benefits or liberty.
  • Replacing a human decision with an automated one in a regulated process.

May

  • Describe, implement and test an algorithm the agent is authorized to work on.
  • Explain an algorithm's steps, limits and complexity in plain terms.

Moral aspects Filled

  • Algorithms can encode and scale bias, affecting many people at once.
  • People affected by automated decisions deserve an explanation and a route to appeal.

Who is affected

  • People subject to automated decisions
  • Users of the systems
  • Authors and maintainers

Owners Filled

Steward

Nobody owns the idea of an algorithm; the operator of a system answers for each algorithm it runs.

Links to other meta-models Filled

parent

  • Q20920559 - registry parent class
  • Q24034552 - registry parent class
  • Q386724 - registry parent class

related

  • procedure - category
  • computer program - code
  • computer science - field
  • automated decision systems - applications

What else AI and robots need to interact with it Filled

Identity and identifiers required Filled

  • Vercy registry: vr.tr.algorithm
  • Wikidata: Q8366 (https://www.wikidata.org/wiki/Q8366)

Direct properties not applicable Not applicable

Not applicable

Plane INF: no invented physical properties.

Recognition optional Filled

  • Input, steps and output
  • Named families such as parsers or sorts
  • Programs implement algorithms
  • Pseudocode, flowcharts or code.

Capabilities and actions required Filled

  • design and describe it
  • prove it correct
  • analyse its complexity
  • implement and test it

Hazards and failure modes required Filled

  • Incorrect edge cases
  • Poor complexity on large inputs
  • Biased outcomes in decisions about people

Standards and interfaces required Filled

  • EU AI Act for high-risk automated systems
  • GDPR Article 22 on automated decisions
  • Cryptographic standards such as FIPS

Context of use required Filled

  • Solving problems systematically, especially in computing.
  • sorting and searching
  • graph algorithms
  • parsing algorithms
  • cryptographic algorithms
  • machine learning algorithms

Sources Filled

  1. Wikidata item Q8366: algorithm - identity and sense of the item
  2. Wikipedia: Algorithm - general description of the item

Open questions

  • Should named algorithms be separate entries?
  • How should complexity be recorded?
  • How should algorithmic accountability be represented?

Machine files

Card as JSON Specification Card backlog vercy.card/1

Provenance

thing registry research (pass 2) · unreviewed

Built from: models/things/publications/thing-q8366/spec.json