Continuing on with my series on what Chat GPT can do in the Social Security context, here's another question and answer:
Question: I am an attorney representing a Social Security disability claimant. She is 56 years old, has a 10th grade education and has worked for the last 20 years as a CNA. She has two bad knees and is a diabetic with peripheral neuropathy. What arguments can I use to prove disability?
Chat GPT Answer: Given your client's circumstances, you could use the following arguments to prove disability:
1. **Medical Evidence**: Gather detailed medical records, test results, and doctor's assessments that clearly document her knee problems, diabetes, and peripheral neuropathy. Highlight any treatments, medications, surgeries, or therapies she has undergone to manage these conditions.
2. **Functional Limitations**: Describe how her medical conditions impact her daily life and ability to work. Highlight how her knee pain and neuropathy limit her mobility, standing, and ability to perform physical tasks required in her previous job as a CNA.
3. **Vocational History**: Emphasize her 20 years of work as a CNA, and how her conditions now prevent her from performing the physical demands of that job. Discuss any attempts she may have made to continue working despite her disabilities.
4. **Transferable Skills**: Evaluate whether her skills from being a CNA can be transferred to a less physically demanding job, given her current limitations. If not, this strengthens the argument that she is unable to work.
5. **Age and Education**: Her age and education level can be factors in proving disability. Explain how her limited education might restrict her ability to transition to a new job that requires less physical activity.
6. **Residual Functional Capacity (RFC)**: Obtain a detailed RFC assessment from a medical professional, outlining her ability to perform work-related activities. This can provide clear evidence of her limitations.
7. **Consistency of Treatment**: Highlight any inconsistencies in her medical records or treatment history that might suggest her conditions have not been managed consistently.
8. **Expert Opinions**: If necessary, seek expert opinions from medical professionals, vocational experts, or disability specialists who can testify to her inability to work based on her medical conditions and limitations.
9. **Impact on Daily Living**: Describe how her disabilities affect her ability to perform everyday tasks like cooking, cleaning, shopping, and personal care. This can provide additional evidence of her limitations.
10. **Social Security Blue Book**: Refer to the Social Security Administration's "Blue Book" that outlines the specific medical criteria for qualifying as disabled. Match her medical conditions to the relevant listings.
Remember, the key is to build a comprehensive case that demonstrates how her medical conditions substantially prevent her from engaging in substantial gainful activity. Consulting with legal experts experienced in Social Security disability claims can further enhance your client's chances of success.
Duh, I'm looking to prove this lady is limited to sedentary work and I may have a good shot at it given the facts I've mentioned. So, why couldn't Chat GPT give more than an extremely generic answer which doesn't speak to the facts I've given? I'd have to give this answer a failing grade -- my job seems secure -- but stay tuned. Part III shows Chat GPT doing something that's impressive in its own way.
This isnt Star Trek Charles. You dont just ask the computer to do a task and it does it. It has to be trained.
ReplyDeleteData collection and preparation are prerequisites for training AI and machine learning algorithms. Without quality data, machine & deep learning models cannot perform the required tasks and mimic human behavior. Hence, this stage of the training process is of utmost importance.
Again, the tool is only as good as the craftperson, but always gets the blame for failure.
Many legal offices are looking at or already using legal based AI. There are several AI companies out there, but I dont think anyone has released an AI for Social Security work yet. Filevine and Clio are popular but I think it is for larger practices at this time, good for contract law and document review and even some legal research. I saw something the other day about A&O (Allen & Overy) using an AI called Harvey, but I dont recall what program it is based on and who did the modeling for it.
ReplyDeleteI would think somewhere, even in the antiquated quagmire that is SSA disability work, someone is looking at AI for medical evidence review. I did a quick google on AI for medical record review and saw an ad for a company called Wisdocs, they said in the ad:
ReplyDelete"We provide the back-office brainpower with our artificial intelligence platform for insurance, legal, and independent medical evaluation firms to review medical records with ease."
I have no idea if it is good or not, I know nothing about it, but I think a tool like this or something more robust could be used by both DDS and reps to aid in disability claims. They claim it can provide clinical information in real time. Again, this is not an endorsement, but pointing out an AI tool that could aid in disability claims processing. Heck, the deduplication features and automatic organized records by date and providers would be a huge help by itself.
Odd that guidelines 201.00(f)/201.02/201.10 are not mentioned. Points 4 and 5 seem to even be going in that direction.
ReplyDeleteIts odd that a free, older, generic program when used for an extremely specialized purpose provides a less than stellar result. Kind of like asking a SSA rep to do Constitutional Law or a very detailed contract.
ReplyDeleteB/t/w - there is a sed svp 3 job - companion - that might kill the obvious grid case here, if there were little to no vocstional adjustment involved. The computer can't be expected to know the nuances from experienced litigators with decades of practice. The basic questions were accurate, but on the superficial side.
ReplyDeleteChrist Charles, this is not Deep Blue playing chase where every piece has defined parameters and limited set of possibilities every time it is prompted. This is you being someone focused on the dream instead of the process of getting there.
ReplyDeleteFor AI to work this:
--You direct and refine the prompts to the regulation, policy, ARs, SSRs and feed it key court decision.
--You then feed it favorable and unfavorable ALJ decisions, AC remands and the DDEs.
--You can even go farther and feed it the key (or all) medical evidence in those same cases.
From that you would be able to have it analyze the common flaw in the ALJ cases, the claimant's that had a common factor for the unfavorable and so on.
YOU need to be the one to commit and read how this level of AI works. You are here whining as if you put up a webpage in the 90s and wondered why the money was not flowing in.
SSA has been using AI, NLP and ML to extract and identify clinical information from medical evidence for a while now in systems like IMAGEN. These types of capabilities assist examiners in quickly identifying information that could support a disability claim.
ReplyDeleteThe agency has a number of folks looking at AI/ML tools. They are working on where and how they could be used throughout the disability determination process. Even if they still require manual oversight, they can make dramatic improvements in case processing times and operational savings.
ReplyDeleteI get frustrated whenever I see someone expecting an AI Chatbot to be a genius of all areas. AI chatbots is like a genius in the sense that it can absorb and process large amount of information and then use that information when relevant to do so. However, like any genius, it can only process information related to areas it has studied. Just like a genius, the more thorough it studies an area the more expert it will be. Asking Einstein to talk about best way to cook a turkey would likely have resulted in a less than stellar response if he never cooked or studied cooking. On the other hand, I'm sure that if he would have cooked or studied cooking he could have become the best turkey cook in the history of mankind. Point is that the AI language models are very impressive because of what they can be trained to do, and how well they can do it once trained. However, you can't judge its potential by asking it about topics that you have no idea as to the extent on which it has received training. Give AI the training described by @7:18 and it will churn out SSA legal briefs, district court and appellate briefs and a list of questions to ask the claimant and VE for any specific case, that will put all of us to shame.
ReplyDelete