Technology

Can You Predict What Congress Will Do?

Actually, you can -- with predictive analytics.

By  
on

One of the allures of emerging AI technology is the ability to — based on past history — predict future outcomes. To this point, there’s been significant investment by vendors in the development of predictive analytics in support of litigation, looking at past case law and inferring the probability of success for different litigation strategies. While this makes sense as a starting point (there’s a large corpus of data in support of litigation analytics), it does beg the question: what other areas in the law are ripe for predictive analytics?

If we shift our focus from litigation to that of transactional law, a logical place to start is with the regulatory framework itself. Changes to this framework impact existing, past and future matters for clients of law firms. In an ideal world, we would be able to predict which changes will make it across the finish line and then plan to address (and remediate risk in response to) those changes. However, as we all know, the political world is far from ideal, and it can be challenging to predict outcomes.

To illustrate just how difficult it can be, consider this fact: between 2001 and 2015 nearly 70,000 bills were introduced in the U.S. Congress, yet only 2,513 were enacted or less than 4 percent of all bills introduced. If that statistic looks too rosy, also consider that bills that are passed often change (sometimes materially) from the time of introduction to the time of passage. In such an environment, how does one answer a client’s questions such as: “What’s the likelihood of this thing passing?” and “How does it impact me”?

To answer those questions (absent of technology), one would have to gather and wade through a fair amount of information. You would look for contextual data points such as:

  • Who is the sponsor of the bill?
  • Is the sponsor in the majority party?
  • Is the sponsor respected? Does the sponsor support from other congressmen or congresswomen?
  • Are there competing bills in play? How do those compare in terms of sponsorship and support?

And unfortunately, context is not enough. To accurately assess the likelihood of passage, you also need to look at the text of the bill itself. Not surprisingly, certain language is highly correlated with the likelihood (or lack thereof) of passage. And of course, as referenced above, there are also the additional challenges of i) ongoing changes to the text as part of the legislative process and ii) the fact that these bills are not typically short documents (as an example, the Affordable Care Act is more than 900 pages long — not an easy read through successive versions).

Tracking these factors, reading the bills, and creating (and updating) predictions is literally a full-time job that requires very specific knowledge. Large law firms employ teams of professionals in public policy groups that inform both lobbying efforts on behalf of clients and that also keep their attorneys in the know. Even those groups, though, tend to focus specifically on key issues, as the task of comprehensively tracking and predicting all legislation can be daunting.

Sponsored

Let’s consider the application of predictive analytics. In a past article, we discussed the conditions in which analytics works well:

  1. There’s sufficient data to train the algorithm, and
  2. That data is consistent in nature.

Addressing the first requirement, there is a plethora of data (granted much of it is unstructured) available for analysis and algorithm training (reference the nearly 70 thousand bills introduced in the 14-year period from 2001 to 2015) so that requirement is nicely met.

However, what about the second requirement? In the period referenced, control of the House changed twice, and control of the Senate also changed twice. At first glance, that seems to violate the rule about consistency of data (which it certainly would if the makeup of a panel of appellate judges changed twice during a similar period); but the reality is that the Republican and Democratic parties have not changed positions or ideologies materially — even if the participants themselves change. As such, the control of each body (and the associated control of committees) simply becomes one of the determining factors in the model.

Assuming that the analytics are available, there are two more considerations for implementation:

Sponsored

  • Transparency. If a bill has an 80 percent chance of passage, the logical questions of “What does that mean?” and “Why?” remain. To trust the analytics (particularly for an attorney advising a client), it’s essential that the algorithm provide transparency into what factors have led to the prediction (and the degree to which each factor contributes to same).
  • Association: If we assume that the 2,513 bills that were enacted from 2001 to 2015 were split evenly across the 14-year period, then that leaves us with roughly 180 bills passed each year… which is still a large number to track (assuming you’re only tracking the bills that pass). Where the analytics become useful is when we can create an association between the analytics and individual companies.

This capability would in turn allow one to track and set alerts for legislative events that are relevant to the observer. Consider a world in which you were able to reach out to your client and alert them that a bill suddenly became very likely to pass and explain how the bill will impact their business. Now THAT would be value added and certainly an AHA moment for the client.

 

 

The reality is that models to predict outcomes in the U.S. Congress already exist — they are surprisingly good, and well on the way to commercial application. If you’re interested in seeing some of these analytics in action, check out companies like Skopos Labs to see how they have built and refined predictive models for legislation. Wolters Kluwer partnered with the company on a solution called the Federal Developments Knowledge Center, and thus far, the AI has demonstrated an accuracy rate of 99 percent predicting when a bill will pass the first chamber (i.e., House or Senate) and a 98 percent accuracy rate predicting when a bill will be enacted into law.

Of course, when the law or associated regulatory environment does change, that brings a whole host of new issues for attorneys dealing with (suddenly) dated matters. That will be next month’s topic, where we take a look at how legal technology can help address the challenges in managing changes to the law over a time continuum.

May Goren Photography

Dean Sonderegger is Vice President & General Manager, Legal Markets and Innovation at Wolters Kluwer Legal & Regulatory U.S., a leading provider of information, business intelligence, regulatory and legal workflow solutions. Dean has more than two decades of experience at the cutting edge of technology across industries. He can be reached at Dean.Sonderegger@wolterskluwer.com.

CRM Banner

Topics

, , , ,