AI and Regulations

[katenamgayoung]

Goal Statement For: Incarcerated Mothers, Children and Social Workers, Prison Staff Who: female prisoners who are mothers and are separated from their children cannot be a consistent influence on their children due to incarceration and cannot be a positive influence on them Your: Hedwig, a tablet-based application Is a: provides ways for incarcerated mothers to communicate and for their children to share through pictures, text and video. That: which reconnects mothers and children and to maintain their relationship to reduce incidents based on isolation within correctional facilities as well as integrate security and regulatory compliance.

Hill Statement Who: Incarcerated Mothers and Children, Social Workers and Prison staff  What: Effectively monitors prisoner and child interactions to understand positive experiences that could affect prisoner behavior as well as negative behaviors that may harm the mother and child. Wow: through an integrated platform where content can be monitored, reducing incidents of technology misuse 24 hours a day

Sub-Hill Statement The platform automatically filters negative contents (text, image) for mothers and children through AI system which created with social workers.

Story Statement As a : Prison Staff I want: to make sure that the platform operates AI technology analysis with contents within prison regulations So that : do not have any privacy and social issues

[bhavanisp]

As a Prison Staff, I want the Hedwig platform's AI to be built and used within the prisons regulatory framework so that I can without hesitation take the responsibility of ensuring and upholding security and privacy.

[akallish]

@bhavanisp this iteration is actually very good. The question is how do we test for it?

[bhavanisp]

Maybe conducting sample tests>creating a data set of those tests>measuring those outcomes as in how many of those actions by AI is incorrect prediction (like false positives and false negatives)?. Just an example but still need to think about how those sample tests be conducted so that its not hindering the actual interaction. We need to also think about on how can we train AI to stay within the boundary and not go beyond that The boundary here is Prisons regulatory frameworks.

[bhavanisp]

@katenamgayoung I personally think this iteration is in good shape, so will be moving it to done.

[bhavanisp]

@katenamgayoung I have moved this issue to Done. Please Review it. Thank you.

[katenamgayoung]

Looks good to go! Thank you

[akallish]

@bhavanisp @katenamgayoung @crutuja The so that statement could be clearer. What about :

So that : harmful content will be blocked and flagged for review by social workers to address with the mother and/or child.

[bhavanisp]

Hello @akallish That's a good one, and I think one of the other epic addresses it. This story is more on the character limitations of AI (I am not sure if its worded right but that's the direction). Having said that, will see what we can do. Thank you for the suggestion!

[akallish]

@bhavanisp Thanks for the response. Your team knows best.

[katenamgayoung]

@bhavanisp I've reviewed it and could you move to the next step writing acceptance criteria? Thank you!

[bhavanisp]

@katenamgayoung yes will begin. Thank you for reviewing!

[bhavanisp]

Acceptance Criteria's Version 1

1. AI Filtering Verification: Given: that I am a prison staff member with access to the prison's Hedwig application When: I review the AI filtering system Then: I should be able to identify and verify that negative contents (text, videos and images) are effectively filtered out based on the established prisons' regulatory body

[akallish]

@bhavanisp This is a good iteration 0. You may need a story from the social worker perspective on the same issue and the so that statement should focus on interacting with mother and child about the questionable content.

[bhavanisp]

@akallish Thank you. Yes I am actually still trying to frame two more criterias. I will see how I can integrate this factor.

[bhavanisp]

@katenamgayoung I want your assistance in verifying if the other issue that we have finalized focuses on the perspective of social worker? Because the acceptance criterias will be according to that. Because I think issue #19 incorporates that

[bhavanisp]

Version 2 Given : that I am a prison staff member with complete access to Hedwig application When : accessing the AI settings within Hedwig Then : I must be presented with

• a detailed comparison and verification of the current AI configurations and the regulatory guidelines documented by the prison for each release
• A section providing a comprehensive report on the effectiveness of the AI filtering system. This report should include type and frequencies of filtered content and insights to AI performance for the benefit of incarcerated mothers and children.

[akallish]

@bhavanisp @katenamgayoung inam not sure you want to give prison staff complete access to Hedwig. Trust is important and a prisoner or child maybe reticent to share knowing this. The social worker should have access and the prison should get certain notifications about dangerous postings.

Kate this is your project and you will need to weigh in here.

[bhavanisp]

@akallish Right okay. @katenamgayoung @crutuja any thoughts on this? Need some perspective.