AI Triage System for EI Claim Recalculations

Employment and Social Development Canada (ESDC) uses a machine learning model to help manage a backlog of older Employment Insurance (EI) claim recalculations. The system classifies each claim into one of three categories: an increase in benefit rate, a decrease in benefit rate, or no change in benefit rate. This allows program staff to close claims that are unlikely to result in any change and to prioritize those where a claimant may be owed a different benefit amount.

This system affects Canadians who have previously filed EI claims that were flagged for recalculation. The model identifies with approximately 90% accuracy those claims where no change would result, helping direct human attention to cases most likely to impact claimants. In the worst-case scenario, a claim entitled to a higher benefit could be misclassified as 'no change,' with estimated losses of $300 (median) or $900 (average) per affected individual. All impacted claims are older, no longer active, and do not affect current or future benefit entitlements.

The system is classified as Impact Level 2 under the Government of Canada's Directive on Automated Decision-Making. Decisions can be made without direct human involvement, but a human override mechanism is in place. A recourse process exists for claimants who wish to challenge a decision. Internal consultations were conducted with Data Governance, Legal Services, Program Policy, Communications, and the Access to Information and Privacy Office, and a Privacy Impact Assessment has been completed.