Comments on the 3rd Review of the Directive on Automated Decision-Making
17 Pages Posted: 10 Oct 2022
Date Written: September 22, 2022
The amendments that the Treasury Board of Canada Secretariat (TBS) has proposed to the Directive on Automated Decision-Making (DADM) and Algorithmic Impact Assessment (AIA) tool as part of their 3rd Review are highly commendable. The amendments will do much to improve the clarity, transparency, and overall effectiveness of the DADM and AIA. These improvements are vital to TBS's pursuits of responsible AI and bolstering public trust in the DADM. However, there are many additional opportunities beyond those identified in the 3rd Review for TBS to further improve the clarity, transparency, precision, and overall effectiveness of the instrument's application, and thereby bolster public trust in the DADM, in its application, and in its outcomes.
Below are the comments I have prepared in response to the 11 issues and proposed changes described within the 3rd Review of the DADM. Additionally, I have prepared some comments pertaining to an additional 2 issues (“Implementation of Quality Assurance” and “Public Access & Engagement”) that I believe the 3rd Review does not account for in sufficient detail.
My comments on those issues touch upon 20 key topics: (1) Scope of services, (2) Scope of agencies, (3) Scope of National Security Systems, (4) Scope of system lifecycle phases, (5) Review timeframe, (6) Human-centred language, (7) Human & institutional biases in application, (8) Intersectionality, (9) Disproportionality of environmental harms, (10) Inclusive design, (11) Explainability by design, (12) Public ADS registry, (13) Transparency of QA activities, (14) QA of expected vs. actual use, (15) QA scheduling & reporting requirements, (16) Resources for QA, (17) Standards of public access and disclosure, (18) Fairness, accountability, and transparency in public engagement, (19) Review of public consultation and engagement opportunities, (20) Resources for public consultation and engagement.
Keywords: Artificial intelligence, AI policy, AI governance, tech policy, public administration
Suggested Citation: Suggested Citation