US Patent 11556839 Auditing system for machine learning decision system

Computer systems and associated methods are disclosed to implement a decision model auditing system that allows clients of a machine learning decision system to audit the decision-making process the decision system. In embodiments, the decision system is instrumented with reporting code to collect internal decision data of the decision system and send the data to a decision auditing service. In embodiments, the auditing service provides the client with an obfuscated token, which may be used to anonymize the client requests to the decision system. As client requests are handled by the decision system, the reporting code generates audit messages to the auditing service. The auditing service stores the audit information, which may later be provided to the client or used generate an audit report. In embodiments, the audit report may indicate whether the decision system contains any undesired bias.