Differential Privacy
A formal privacy technique that adds carefully calibrated random noise so that the presence or absence of any single individual cannot be inferred from the output. It buys a mathematical privacy guarantee at the cost of accuracy, and that cost often falls hardest on the smallest groups, which is why a differentially private dataset still needs a per-group utility check.
Defined in 4 GAGE programs, which carry 5 distinct definitions of it. The wording above is taught in AI Governance: Applied Mastery.
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
A formal privacy technique that adds carefully calibrated random noise so that the presence or absence of any single individual cannot be inferred from the output. It buys a mathematical privacy guarantee at the cost of accuracy, and that cost often falls hardest on the smallest groups, which is why a differentially private dataset still needs a per-group utility check.
A mathematical technique for adding calibrated statistical noise to an aggregated dataset or metric so that the result carries a formal guarantee no individual data point can be reliably reconstructed from the published aggregate, used in fleet-level analytics to gain insight without centralizing individually identifiable sensor data.
A formal mathematical framework for quantifying and bounding the privacy loss introduced by releasing statistics or synthetic data derived from a real dataset, expressed as a parameter (commonly called epsilon) that trades off privacy strength against statistical utility.
A technique that adds calibrated noise so that an analysis result is essentially unchanged whether or not any one person's data is included, making re-identification impractical.
A formal anonymisation technique that adds calibrated statistical noise to data or query results so that no single record materially changes the output, giving a quantifiable, testable privacy guarantee rather than a hoped-for one.
Where it is taught
The exact lessons this term appears in. The first 7 topics of every program are free with a free account.
- Privacy and Data Rights in AI Systems · Ethical and Responsible AI and Operational Governance, AI Literacy & Professional Conduct
- Synthetic data: when it protects, when it launders, and how to tell · Feeding the Machines, AI Data Governance: The Data Chair
- The retention decision: what you must keep, what you must destroy, and proving both · Data Reality, AI Governance: Applied Mastery
- Synthetic data: when it saves you and when it launders a bias · Data Reality, AI Governance: Applied Mastery
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.