That Time the Insurance Algorithms Creeped Me Out
Chris Cheatham
Passionate about making youth sports more fun | Make custom cards at statlegend.com
There isn’t just one time that the RiskGenius insurance algorithms creeped me out. They do it all the time. But I thought you might find the latest algorithm creepiness interesting (and maybe slightly alarming if you are a contract specialist).
When we started RiskGenius, the idea was to link construction contract clauses to insurance clauses to speed up comparison of the two. As we started building, we realized insurance was more than we could handle initially so we pushed construction aside.
But in the back of my mind, I always wanted to apply machine learning to construction contracts.
Today, I did that. I used the RiskGenius algorithms on a construction contract. I wasn’t expecting much because the algorithms I was using were trained on insurance language. Not surprisingly, most of the categories were off.
But then I started noticing some of the categories were hitting. Here’s an example of a “Construction Schedule” clause that is correctly categorized by RiskGenius insurance algorithms:
What’s strange is that "Insurance Schedule" clauses are -- to me -- totally different than construction schedule clauses. Here’s an example of an insurance schedule clause:
“Schedule of Forms Schedule of Forms To be attached to and form part of Company: Federal Insurance Company Policy No. 8246-9692 Issued to: Claim Kit Inc. ForeFront Portfolio 3 0 General Terms and Conditions Policy 14-02-17770 (12/10 ed.) 14-02-17955(4/11 ed.) ForeFront Portfolio 3 0 Directors & Officers and Entity Liability Coverage Part 14-02-17293 (1/11 ed.) 14-02-17833 (12/10 ed.) 14-02-17955 (4/11 ed.) 14-02-18527 (11/11 ed.) 14-02-19646 (2/13 ed.) 14-02-21084 (3/14 ed.) 14-02-5989 F3 (5/11 ed.) Form 14-02-0854 (Ed. 04-01)”
(Yes, that is from our company's D&O policy.)
Compare that to the construction schedule language above.
The RiskGenius machines are getting smarter, my friends. And if you spend just a couple seconds thinking about it, you will start to see how contracts will overlap and how this creates a huge opportunity for artificial intelligence.
If you are interested in how we are applying RiskGenius to insurance or construction, feel free to drop me a line. chris at riskgenius dot com
Pricing Manager| Actuary| Underwriter |Product & Data Manager| Ultrarunner & triathlete
7 年it will be interesting to know more about insurance application
Senior VP @ Ascot Group | General Liability and Professional Liability Facultative Reinsurance
7 年It will definitely be interesting to see what insights the algos will uncover particuarly with respect to gaps in coverage across multiple insurance products (often times written by multiple insurance companies).