The key to AI trustworthiness? Mitigating bias. This webinar explores SAS techniques designed to mitigate bias in machine learning models.     Watch the Webinar   You will learn about: Common sources of bias in machine learning models. SAS procedures and metrics used to assess fairness. Practical approaches for monitoring and reducing model bias. The questions from the Q&A segment held at the end of the webinar are listed below and the slides from the webinar are attached.   1. Which fairness metrics should I use, and how do I choose among them? There really isn’t one fairness metric that works for every situation. It depends on what you’re trying to protect against. For example, demographic parity looks at whether groups receive positive outcomes at similar rates, while equal opportunity focuses on whether qualified individuals have a similar chance of getting a positive outcome. I usually recommend starting with the business use case, identifying who could be impacted, and then choosing the metric that reflects that specific risk.   2. How can fairness assessments support Responsible AI and AI Governance initiatives? Fairness assessments turn Responsible AI principles into something we can actually measure and govern. Instead of simply saying a model should be fair, we can evaluate outcomes across groups, document what we find, establish thresholds, and monitor whether that changes over time. That gives governance teams evidence they can use for reviews, approvals, and ongoing monitoring.   3. At what stage of the AI lifecycle is bias mitigation most effective? Ideally, throughout the lifecycle. It’s much easier to address bias when you’re looking for it early, starting with how the problem is defined and what data is being used. But it shouldn’t stop there. We should assess fairness during model development, before deployment, and continue monitoring after deployment because populations, data, and outcomes can change over time.   Recommended Resources SAS Trustworthy AI Examples Github Procedures for assessing and mitigating bias in SAS Monitoring Model Fairness in SAS Model Manager A Model Card for Every Model: New Model Card Available SAS Developers - Trustworthy AI Beyond balancing data: How conditional generation makes synthetic data more useful Please see additional resources in the attached slide deck.   Want more tips? Be sure to subscribe to the Ask the Expert board to receive follow-up Q&A, slides and recordings from other SAS Ask the Expert webinars. ... View more

Team Name SASquatch Track Telecom & Media Use Case Analyze where growing electricity demand from AI data centers may outpace the grid’s ability to add new power capacity.  Technology SAS Viya (various applications) Region NA Team lead Riley McGowan Team members aubrykhuff, u64589018, (2 pending) Social media handles linkedin.com/in/Aubry-K-Huff , Siddharth Birajdar - Globe Life | LinkedIn , linkedin.com/in/DheerushaTiwari , linkedin.com/in/RileyMcGowan  Is your team interested in participating in an interview? Y Optional: Expand on your technology expertise Our team plans to use SAS Viya to combine public electricity-demand and grid-interconnection data, analyze demand and supply trends across U.S. power regions, and identify where capacity constraints may be developing. Pitch Video   Jury Video Team Photo   ... View more

CONCERN: PROC SQL DESCRIBE should correctly list the formats applied to the columns in the queried table. However, if a format is applied to one or more columns in the table, any format name that exceeds 8 characters in length is truncated to 8 characters in the table definition that is written to the SAS log. RESOLUTION:  PROC SQL DESCRIBE TABLE should displays the full format name rather than truncating them to 8 characters. Change PROC SQL so that full format names are displayed.   proc format; value TellMeIfItWorks 1 = 'Yes' 2 = 'No' 3 = 'Maybe'; run; data LONGFMTS; length a b c 8.; format a roman8. b comma12.2 c TellMeIfItWorks.; a = 1; b = 2; c = 3; run; proc sql; describe table work.longfmts; quit; * work around; %let tellmeabout = LONGFMTS; data _null_; set sashelp.vcolumn end=eof; where libname = 'WORK' and upcase(memname) = upcase("&tellmeabout"); if _n_ = 1 then put "WORK.&tellmeabout." / "("; comma = ifc(eof, '', ','); if format = '' then put ' ' name type comma; else put ' ' name type format= comma; if eof then put ")"; run;   ... View more

ICU patients who need a nursing home, rehab or hospice often wait in hospital beds while the search starts late. Day One flags them on their first day in intensive care, with a draft plan a case manager can act on. ... View more

Hi everyone, I am working in SAS Visual Analytics on SAS Viya and encountering an issue with how interactive filters behave in my report: The Problem: My report includes controls (such as a date slider and drop-down list filters) linked to a table. When a user clicks or highlights a specific row in the table, SAS automatically applies that row selection as an active filter back to the other control filters. As a result, the drop-down list narrows its available options to display only the values associated with that single selected row. Desired Behavior: It is very important for me that the options displayed in the drop-down list remain strictly relevant to the data present in the table (dynamic/relevant values only). However, I want the main controls to filter the table without a single row click inside the table filtering back and reducing the drop-down list options to just that one row. Is there a recommended way or a workaround to keep the drop-down list synced to show only relevant values from the table, while preventing a table row selection from filtering the control options back? Thanks in advance for your help! @VisualAnalytics  ... View more