ECDIS, Regulation, Public Comment
Dec 2023: Colorado’s Quantitative Testing Protocol with Real World Data
Case study on Colorado’s draft protocol for testing unfair discrimination showcases challenges devising globally-applicable approach.
Elaine Gibbs, March 13, 2024
In response to Colorado Division of Insurance’s DRAFT PROPOSED Algorithm and Predictive Model Quantitative Testing Regulation, released in October 2023, Bell Analytics conducted a case study, applying the draft protocol to real-world data.
First, we inferred racial and ethnic classifications using BIFSG and compared to self-reported identifications, examining the applicability of the BIFSG approach.
We then followed the draft protocol. For certain model types and datasets, we found the test procedure:
May miss clear examples of illegal and unfair discrimination
May flag certain data points for remediation, without flagging others that are very closely related
Does not fully account for reasonable correlations to the underlying insurance practice
Omits an explicit measure of impact on protected classes or business objective specific to the data point in question
Find the full analysis here.
Latest articles
Get in touch.
Curious how we can help? Let's discuss.
Reach us at hello@bell-analytics.com.