In the age of AI and Machine learning, almost all activities are getting automated and much faster also.
We always read that judiciary has major challenge of heavy work load and therefore the decision/verdict takes lot of time.
I believe that AI enabled system could greatly help in lowering the impact of workload and fastening the decision process.
Opinions are invited!
It is a major challenge, let us see what it would take:
Analysis of Evidence
The system must weigh in to allow or reject the evidence as well as pass judgement of relevance.
Interpretation of Cited Cases
contextualizing them and distilling what is relevant for the case law
Reasoning
The system would need to know when to use the following classes of reasoning:
The system must also deal with missing premises on all of the above
Text Generation
It must also be able to do text generation of the decision and case law that supports it. The system must be able to explain the reasoning behind the decision.
Hi Arturo! Thanks for your wonderful input. I totally agree and appreciate your first point (heading). However, last three are doable. Thanks for sharing your thought.
Professor G R Sinha
Do not get me wrong, I think there is promise to the point that I am currently engaged in it. But there are some major challenges. Some of them are technical, others are not not. Below are some of the challenges on the last three categories.
Extra logical assumptions:
Example1: I am working with processing patents and doing programs to automate them. The challenge that I tackled (which I think worthy enough to invest a lot of money funding the research) is on the concept of non obviousness. I solved it by introducing the concept of a metric for such problem. But this is just one instance of what I think will happen on other areas.
Example 2: How do you model a law text, through Alethic modality or Doxastic logic? Do you engage case law through temporal logic or treat it as static? Do you interpret expert witness testimony through mathematical logic or epistemic logic? This impacts the selection and implementation of your system and the interpretation of the cases.
Text generation
Example 3: just yesterday there was a CNN interview on how the Muller report should be given out to the public. What was interesting was that there is no comparative example on which to generate a report to congress due to possible conflict on public interest and possible classified information. I thought that was intriguing notion and very difficult. How would a machine handle this?
The best approach that I could think of is to generate frames for text generation and then use machine learning for selection of the frames. The counter argument that I thought of is based on what happened to Amazon (https://www.popularmechanics.com/technology/robots/a23708450/amazon-resume-ai-sexism/ ).
Again, I am very hopeful and engaged in it and it may be worth for us to see if we can solve them.
Regards
Professor G R Sinha
Saw your profile and given your background there are some interesting areas where your expertise in image processing might be helpful. For example, one area where I have worked is in determining where in an image are numbers located and assigning a radius that as more elements are detected in the neighborhood, the radius of the neighbors radius is reduced. This can help localize the area of interest within the patent. Similar problems are common in exhibits such as images used in scene identification for crime scenes. These can then be correlated with the case narrative to detect inconsistencies.
These areas are ripe for the taking and good opportunities, but a lot of work is still necessary
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude