AI for EverybodyLearn the intelligence era with better judgment.

Your progress: 0 complete

Glossary My Map Account

Glossary My Map AccountQuick answers Choose my route History of AI Sources About this platform

You are here

Current learning moveOrient

Follow a question, situation, concept, person, or historical thread.

Do nowRead, decide, then check

Your guide

Bias cannot be removed by one technical switch

Your learning move

OrientEnter through a question or situation.

Why this page exists

Explore is the connected atlas. It shows several ways to approach the same subject without pretending that every learner needs the same sequence.

What you will leave with

AI behavior reflects data selection, labels, objectives, model design, thresholds, deployment, institutions, and feedback. Some measurable disparities can be reduced.

Before you begin

No prerequisite is required for topic and Quick Answer entrances.

Do this now

Choose the entrance that matches what you already have: a question, little time, a situation, or a term.

Do now on this pageRead, decide, then checkThe plan states the outcome, time, prerequisite, immediate action, and prepared continuation.Another valid routeGet a Quick AnswerUse a short orientation before deciding whether to go deeper.

Account

Quick answer · 5 min

Bias cannot be removed by one technical switch

Can AI be unbiased?

Plan for this page

Get a clear answer, then check that you can use it.

You will leave with: Treat fairness as technical, institutional, and human work with evidence and recourse.
Time: 5 min
Before you begin: No prerequisite. Start with your question.
Do this now: Read the direct answer before the example and active check.

Prepared next stepStudy bias and fairness

AI behavior reflects data selection, labels, objectives, model design, thresholds, deployment, institutions, and feedback. Some measurable disparities can be reduced.

Fairness has competing definitions and depends on rights, history, purpose, and affected people. Governance and recourse matter alongside metrics.

Example

Hiring history

Training on past selections may reproduce unequal opportunity even if protected fields are removed because other signals act as proxies.

Important limit

No single fairness metric settles justice

Metrics can conflict; equal averages can hide subgroup or individual harm.

Interactive model

Separate before deciding

Selected layerStatistical bias

Systematic error or disparity in measurement.

Read the complete text alternative

Statistical biasSystematic error or disparity in measurement.
Social fairnessA normative judgment about legitimate treatment.

Useful distinctions

Separate the questions before judging

Statistical bias

Systematic error or disparity in measurement.

Social fairness

A normative judgment about legitimate treatment.

Active check

What is the strongest fairness review?

Study bias and fairness

Sources and scope

What supports this object

Contested

AI RMFNIST

Risk vocabulary; fairness definitions and remedies are context-dependent.

Open source

Loading page content...

…