Compliance

impact level

Four-tier scale (Level I–IV) reflecting potential harm.

How the laws define this

Both instruments from the Treasury Board of Canada Secretariat agree that an impact level is a four-tiered score, ranging from Level I to Level IV, generated by an Algorithmic Impact Assessment (AIA) to evaluate an automated decision system (ADS). They concur that this rating directly determines the scale of required safeguards, such as mitigation and transparency measures. However, the instruments diverge in their specific scope and regulatory focus. The Algorithmic Impact Assessment (AIA) tool explicitly grounds the assessment in potential harms to rights, health, economic interests, and ecosystems, specifying that the derived score dictates required review measures. Conversely, the Directive on Automated Decision-Making focuses on defining the impact spectrum itself—from little or no impact at Level I to very high impact at Level IV—and expands the resulting regulatory duties to explicitly include mandatory testing and oversight obligations governed directly under the Directive.

Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.

How jurisdictions differ

What is distinctive in each definition, binding law in force first.

JurisdictionInstrumentStatusWhat is distinctive
CanadaAlgorithmic Impact Assessment (AIA) tool (Treasury Board of Canada Secretariat)In Force (Amended)This definition explicitly lists rights, health, economic interests and ecosystems as the potential harms assessed, while specifying that the impact level derives from AIA scoring to determine mitigation, review and transparency measures.
CanadaDirective on Automated Decision-Making (Treasury Board of Canada Secretariat)In Force (Amended)This definition frames impact level as the direct output of the AIA classifying ADS from Level I (little or no impact) to Level IV (very high impact), mapping each to proportionate mitigation, testing, transparency and oversight described in the Directive.

What it triggers

  • In the Algorithmic Impact Assessment (AIA) tool, the impact level derived from the AIA scoring sets mitigation, review and transparency measures.source
  • Under the Directive on Automated Decision-Making, each impact level maps to proportionate mitigation, testing, transparency, and oversight described in the Directive.source

Definitions (2)

A four-tier classification (Level I to Level IV) reflecting the potential harm an automated decision system may cause to rights, health, economic interests and ecosystems; the impact level, derived from the AIA scoring, determines required mitigation, review and transparency measures.

Impact level is the result produced by the Algorithmic Impact Assessment (AIA) that classifies an ADS from Level I (little or no impact) to Level IV (very high impact). Each impact level maps to proportionate mitigation, testing, transparency, and oversight obligations under the Directive.

Not to be confused with

Decision significance

Decision significance classifies expected impact as non-significant, significant or critical, while impact level uses a four-tier scale from I to IV based on AIA assessment.

High-impact / High-risk AI system

High-impact or high-risk AI system denotes systems with significant potential to harm rights or safety, whereas impact level is a specific four-tier AIA classification applied to automated decision systems.

Questions and answers

What harms does the impact level in Canada's AIA tool assess?

The impact level reflects potential harm an automated decision system may cause to rights, health, economic interests and ecosystems. It is derived from the AIA scoring.

How does the Directive describe impact levels for ADS?

Impact level classifies an ADS from Level I with little or no impact to Level IV with very high impact as produced by the AIA. Each level maps to proportionate measures described in the Directive.

What follows from an impact level in these Canadian instruments?

In the AIA tool, it sets mitigation, review and transparency measures. Under the Directive, it sets proportionate mitigation, testing, transparency and oversight measures.