From a contracting notice posted by the Social Security Administration (emphasis added):
Why would the Navy have been issuing a task order on behalf of Social Security? What does "detailed performance improvement and future state analysis of various aspects of the Social Security Disability Insurance (SSDI) Program" mean? How could an applied physics lab possibly help Social Security? This may be legit but it sure sounds fishy.The Social Security Administration (SSA) intends to establish a contract with Johns Hopkins University Applied Physics Lab (JHU/APL) on a sole source basis. SSA administers the Social Security Disability Insurance (SSDI) Program under which SSA pays monthly cash benefits to millions of Americans insured under the Social Security Act. SSA annually processes and adjudicates hundreds of thousands of disability determinations, including at the administrative hearing and appeal levels. JHU/APL is currently conducting a detailed performance improvement and future state analysis of various aspects of the Social Security Disability Insurance (SSDI) Program. JHU/APL is performing this work on behalf of SSA pursuant to a task order issued by Naval Sea Systems Command (NAVSEA) under a contract between NAVSEA and JHU/APL. SSA seeks to maintain the essential engineering, research and development capability currently being provided to it by JHU/APL beyond the expiration date of the NAVSEA task order in 2020. SSA intends to use the independent, unbiased technical analysis and recommendations it receives from JHU/APL under the sole source contract to continue to identify opportunities for policy improvement and increased efficiency and accuracy in the administration of the agency’s mission. ...
13 comments:
Some of the Saul minions believe that artificial intelligence can radically remake the disability adjudication process. I suspect this is related. Skynet is coming.
I see an electronic ALJ in our future at an SSA kiosk. It's name is Bob. Bob is infallible and will always make the right decision. No humanity will be harmed as it will not be involved in this process. Our future is Bob. Hale Bob!
SSA is obviously keenly interested in artificial intelligence to reign in ALJ discretion and identify alleged errors, among other things. The Insight program which started at the Appeals Council and is now mandatory for writers at the hearing level is one example -- it provides and accumulates data on ALJ "errors" that should be corrected or even that may purport to show that the decision outcome is wrong. Making the Decision Writing Instruction template mandatory for ALJs as of early March 2020 will also provide SSA with a lot more data to use for AI. The Decision Writing Instruction template is not about efficiency as many ALJs find it slows them down -- the benefit is in data mining. Things like this and what JHU APL has been tasked to do could potentially lay the groundwork for revamping the adjudicatory system. For an overview by two leading scholars in this area, see this article discussing SSA's use of artificial intelligence tools to adjudicate disability cases and addressing the need to "retrofit" the APA if more artificial intelligence tools are used. https://administrativestate.gmu.edu/wp-content/uploads/sites/29/2019/11/Engstrom-Ho-Algorithmic-Accountability-in-the-Administrative-State.pdf
I know it sounds fishy but let's not ride the wave of negativity. I will assume it's above board until someone explains how it doesn't hold water. Right now it's all hands on deck at SSA so I have bigger fish to fry.
The idea of pulling humans out of the process by substituting computers has been a dream since the mid-80s. But the tech and the software were no match to the complex programs SSA runs. (For example, for a time, when it was first being developed, it was hoped MSSICS would allow self-help with lower graded CRs to handle SSI, thus saving money. Obviously that didn't happen but not because it is an illegitimate goal - it was just not yet practical.) So I'd agree that they've hired data nerds and are focused on an analysis of the processes. A necessary step if they wanted to use AI or substitute people with software but likely also a necessary step just because nothing as old and as complex as the SSDI program can be workable forever. Reinvention or at least reengineering is a valid and likely necessary step to keep things running. It's not getting the data or analyzing it that 's good or bad, it's how the results are used. At this point, tin foil isn't needed. Plus whatever the results, they come from a source that, unlike career agency employees, will get "respect" by the higher powers that be. And can be something SSA points to to justify future policy, budget etc.
JHAPL does statistical/mathematical modeling for all types of Navy & Coast Guard projects. For instance, the Navy wants to know how many ships to buy or the Coast Guard wants to know where to put radar stations, they hire JHAPL to do the data analysis of where it makes most sense.
Commissioner Saul spoke about the JHU/APL contract briefly at a January 29 hearing before the Senate Special Committee on Aging (starts at time index 00:56:40):
https://www.aging.senate.gov/hearings/thats-not-the-government-calling-protecting-seniors-from-the-social-security-impersonation-scam
Right now Insight is a far cry from a "robot ALJ". It basically runs a check against some types of decisions to make sure there's no obvious error on the face of the decision. It checks that all the DOT codes actually exist, and the DOT description matches the RFC, or that there's no obvious SGA within the period at issue for example. It can't really read or interpret the content of medical records or run any kind of check on the decision's internal logic. It's basically catching obvious decision writer errors based on a small set of defined rules, not adjudicating a case.
Pretty interesting that SSA has had a few AI, machine learning and natural language processing projects supporting IT Modernization. Why SSA Systems Leadership is having JHAPL perform duplicative work is par for the course. Systems executives do not trust the current work that has been going on for years or clueless to existing work going on. Probably a bit of both.
I'm not even sure why the Appeals Council needs to exist, now that Insight is used at the ALJ level. All those AC analysts and hearing examiners could have gotten to work as recon officers, in between DDS and the ALJ level.
NAVSEA runs the naval shipyards, and I suppose disability is a big issue for workers there. Non-SSDI disability insurance usually has an offset for SSDI, so the Navy might have an interest in getting their disabled shipyard workers onto SSDI as quickly as possible. They also have a research budget that they must spend. So I don't think it's suspicious that they would fund work that would be useful to SSA. It looks like the Navy funded it for a while and now SSA is picking it up. Also, as someone above noted, JHU/APL does not just do high-tech stuff like AI, and it sounds like the work here is more mundane.
Don’t get any of your hopes up too high. Anyone who has done any real work in AI, either at the graduate academic level or real world, knows that it’s just about rule processing. If you can determine and write a rule for it then the computer can process it. If it takes real intelligence to figure out the solution, as so much of SSA claims processing does then it’s just more of garbage in and garbage out.
If you listened to Sauls hearing, this is not about AI. It is about using the analysis and computing ability of the center, which had employees in Baltimore, to study efficiency and processes. He is seeing to identify areas where SSA can cut time that is wasted and perform analysis on ways to make the agency move more efficient.
Post a Comment