evidoria

← Back to browse

Good practice Imported

Canada's Algorithmic Impact Assessment — a Mandatory, Published Risk Score for Every Automated Government Decision

Canada · Ottawa · See the Canada profile · See the Ottawa profile

Evidence: Observational / pre–post Top 14% 80/100 · Ask Evidence Copilot about this practice

Since April 2019, Canada's Directive on Automated Decision-Making has required every federal automated decision system to complete a public Algorithmic Impact Assessment — a risk survey whose score sets mandatory transparency and oversight requirements before deployment.

65
Risk questions in the Algorithmic Impact Assessment
41
Mitigation questions in the Algorithmic Impact Assessment
4
Impact levels produced by the assessment
Canada's Algorithmic Impact Assessment — a Mandatory, Published Risk Score for Every Automated Government Decision

Details

Maturity
Established
Promoter
Treasury Board of Canada Secretariat
Period
2019-present
Keywords
digital government, algorithmic accountability, public administration

Context

Canada's Treasury Board Secretariat introduced the Directive on Automated Decision-Making on 1 April 2019, requiring federal institutions to assess any automated system used to make or assist an administrative decision about a person.

Objectives

The assessment tool - the Algorithmic Impact Assessment (AIA) - was built openly, with its source code published, and asks 65 risk questions and 41 mitigation questions covering the system's design, the decision type, its impact and the data it uses.

Activities

The AIA produces one of four impact levels; the higher the level, the more the directive requires, from plain-language notices and human-in-the-loop review to independent peer review and public release of source code. Completed assessments are published on the Open Government Portal.

Results

Dozens of federal automated decision systems - spanning areas such as immigration case triage and benefits administration - now have a public AIA on file.

Conclusions

Independent reviewers, including the Montreal AI Ethics Institute, note the process still depends on self-scoring by the department that owns the system, and that the directive's obligations apply prospectively, so many automated tools built before 2019 are not required to have a published assessment.

Implementation
Indicative cost
Low (< €50k)
Time to results
Long (> 3 years)
Staffing & skills
Treasury Board of Canada Secretariat directive owners, departmental teams completing self-assessments

Conditions for success

  • A binding directive covering all federal automated decision systems
  • An openly published, versioned assessment tool with source code available
  • Publication of completed assessments on the Open Government Portal
  • A tiered escalation path linking risk score to mandatory oversight requirements

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful