A3. I think interpretability of the models, and the ability for a subject matter expert to help correct bias without rewriting the entire input dataset for the models is a huge struggle. It’s a great place where in my direct experience, #SAS can help. #saschat #government

A4. In one study with the FDA on premarket drug adverse events, we addressed interpretability by enabling subject matter experts to assess and directly influence the models. You can learn more here: communities.sas.com/t5/SAS-Global-… #saschat #trustworthyAI #government

A4: Yes agree, diversity in #datascience teams overseeing algorithms is a good way to ensure trust in the algorithms. #saschat Lots of good thoughts coming out of this tweet chat I like it.

A5. Enable processes to highlight potential areas of bias within models through interpretability, as well as to work directly with the models to counteract bias and ensure full representation. Keep the ethical human in-the-loop! #saschat #trustworthyAI #government

A5: #saschat And if this is embedded in the culture of the organizations... here's some encouragement for people to vote with their choice of spending. #AI

#saschat: I think that we could review this entire chat and come up with a set of general recommendations for responsible AI and innovation. Would be a good project for both #government and #commercial. Cheers folks good chatting with you today!

Hi folks I'm Tom Sabo, a solutions architect for #SAS focusing on government. I'm happy to talk today about how #AI and #analytics can help address #fraud in #government. Nice to be back on the #saschat (s) it's been awhile!

#A1 #saschat With new methods in #AI come new ways to thwart traditional methods of #fraud prevention. Think about how much more realistic a phishing e-mail from #ChatGPT would sound vs. what you are used to!

#A2 #saschat Leveraging #textanalytics can greatly assist in manually reading or using ctrl-F through large documents looking for fraud indications. This can flag patterns of risk, for instance, in agreements between #government and its regulated entities.

#saschat A3: Govts. can most effectively deploy #AI and #analytics to combat fraud largely by working with a team who can help automate certain aspects of their process. We want to leverage fraud experts for their knowledge, not for mindless rote tasks!

SAS Forums A5. AI software - such as SAS Viya - has been developed in such a way that employees do not have to be super specialized because most operations are drag-and-drop. #saschat, #government, #fraud, #AI

SAS Forums start with running pilot programs that demonstrate the efficiencies and effectiveness of these solutions and then scale it. And we are good in this. #saschat #fraud

#saschat Good times taking #sas and #ai - catch you all next time!

Hi folks glad for another round of analytics discussion hope to touch on some #NLP and #AI topics today. #saschat

A1: Methods of interpretability for LLMs... how do we tune these without just throwing more data into the mix? Should we be filtering or doing work pre-filters? #saschat

A3: Participate in a #saschat I'm getting smarter here by the minute :)

A4: I've seen partnerships where data scientists work with business stakeholders who have traditionally done processes like comment analysis through largely manual means. Enable #NLP automation at key points guided by data scientists and SMEs. #saschat

A5: Clearly, adaptability. #saschat

Good times today, stay adaptable and don't lose sight of the broad array of capabilities in #NLP! #saschat