Automating Plain Language

 

From Microsoft Stock Images

Over two decades ago our NCSC Jury Studies pioneer, Tom Munsterman, ran the WordPerfect word processor language level test against some jury instructions.  He told me it reported that the reading level needed to understand the instructions were university graduate school.  So, when an article identifying “ways that lawyers could make their written documents easier for the average person to read” was posted by MIT, I thought it was time to revisit the subject?

 

---

A quick search found a Stanford Law Review article from 1981 calling for “A Model Plain Language Law” for “consumer contracts to be written so that an average consumer can understand them”.  

Thus, legal language comprehension has been a problem and concern for a long time.  Why hasn’t this improved? Going back to the MIT article, they made an interesting finding.

“After analyzing thousands of legal contracts and comparing them to other types of texts, the researchers found that lawyers have a habit of frequently inserting long definitions in the middle of sentences. Linguists have previously demonstrated that this type of structure, known as “center-embedding,” makes text much more difficult to understand.

The biggest culprit, they found, was center-embedding. In this type of construction, a writer introduces the subject of a sentence, then inserts a definition of the subject, and then continues on with the sentence. In their paper, the researchers included this sentence, with a lengthy definition in parentheses, as an example:

“In the event that any payment or benefit by the Company (all such payments and benefits, including the payments and benefits under Section 3(a) hereof, being hereinafter referred to as the ‘Total Payments’), would be subject to excise tax, then the cash severance payments shall be reduced.””

This observation is very helpful but, what are the steps that courts could take to rectify this and other language comprehension-related issues?

First, one sensible approach that the Office of Justice Programs posted was a 1982 report that called for more testing for jury instructions.  The summary of the report said: “To help correct the problem, jury instructions should be presented in written and oral form and at the beginning as well as at the end of trials. The instructions should be tested on volunteer jurors; instructions judged incomprehensible should be rewritten and retested. This rewriting/retesting cycle should continue until all points of law touched on by the instructions are easy to understand. “

But this is 2022, can’t computer intelligence help? Maybe?  Here are some systems to look at.

First, there are Contract Standards (https://www.contractstandards.com/About-Us) They are a company that are offering “plain language” contract templates and other resources.  They have also posted a “Legalese Translator” page that is interesting here. 

The second is from our old friends at “LanguageLine Solutions” that shared a service from their “Clarity team”.  We know them from international language translation services.  So, it makes sense that they can help with all areas of language presentation and meaning. 

And last, we see the first steps of having AI applied to the problem.  From the University of Toronto Mississauga we learn about a new Natural Language Processing and Machine Learning system from “Skritswap” (https://skritswap.com/).  Their system claims that it “simplifies the jargon in your contracts by replacing your original text with simpler language and clearer formatting through AI.”  The founder, Prof. Cosmin Munteanu” said that a recommended approach would be “to allow you to access the original document but also allow you to access a translation document".  I like this approach because it creates a feedback loop and can be used for teaching.

The bottom line is that it is time to get to work on our jury instructions and court documents.  After all, it has been forty years.