Chapter 09 · Sustainability & AISustainability & AI

Supervised learning

Meaning statusEstablishedSource recordDirect document linkedWhy these are different

Definition

Machine learning in which a model learns to map inputs to outputs from labelled examples — such as images tagged with their contents or histories tagged with outcomes.

References

Springer / StanfordHastie, Tibshirani & Friedman — The Elements of Statistical Learning

This source provides part of the technical or institutional basis for the definition.

Communications of the ACM 2021Gebru et al. — Datasheets for Datasets

This source supports the explanation of how the term is applied, measured or governed in practice.

Overview

What it means

The dominant paradigm in applied AI: a training set of input–output pairs defines the task, and the model generalises the pattern to unseen cases. Performance depends on the quality, volume and representativeness of the labels.

How it is used

Sustainability applications include classifying satellite scenes by land cover, screening documents for relevant disclosures, and predicting equipment failure or energy demand from historical records. Labelling the training data is often the most expensive — and most human — part.

Why it matters

Supervised systems inherit the judgements baked into their labels. Whoever defines and applies the categories — what counts as deforestation, a risk, a violation — shapes what the model can see.