In a recent piece in Integration Developer News, Lori Witzel, TIBCO’s director of Research for Analytics and Data Management, stressed the importance of trust and transparency in AI innovation. As she put it, when these two components are in place, everyone benefits—businesses can reap the technology’s rewards, and the public can rest assured knowing that their privacy and ethical concerns are addressed. 

Increasingly, the regulatory environment is pushing for more ethical AI with various state and federal bills aimed at making organizations more accountable. As this happens, businesses need to prioritize establishing trust at the forefront of their AI project planning. 

According to Witzel, the two most critical components of trustworthy AI systems are transparency and auditability. Transparency ensures traceability, explainability, and clear communication of AI models, algorithms, and machine learning processes. An auditable AI system, for its part, requires that models and documentation are clear enough that auditors could produce the same results using the same AI methods as the original data science team. 

Transparency and trust also necessitate the understanding of a model’s lineage, which is the set of associations between a model and all of the components involved in its creation. This understanding can be difficult—if not impossible—to obtain without robust, scalable model operations. As such, Witzel also believes scalable model operations are an important part of preparing for AI regulations and offer companies additional benefits, including: 

  • Supporting transparency. The right model management ensures that machine learning models are traceable and explainable.
  • Enabling auditability. Organizations can make audit paths clearer by bringing machine learning models into model operations. 
  • Making AI at scale possible. With leading model management tools, organizations can significantly scale model deployment. According to Witzel, it’s not uncommon for companies to increase deployments from tens to thousands of models, giving these enterprises a significant competitive edge in the race to derive value from AI. 
  • Reducing risk. Companies can also decrease the risk of adverse AI impacts and the possibility of negative regulatory attention through transparent, auditable model management. In addition, if unintentional harm from AI does occur, the issue can be quickly traced to the source and addressed. 

For more on the above, along with tips from Lori Witzel on getting started with developing more trustworthy, transparent AI, head over to Integration Developer News.