Auto mechanics can only be helpful if they can look under a car’s hood. The same applies to data scientists and their ability to extract valuable insights from artificial intelligence (AI) models. While new machine learning techniques help produce more accurate analytical models, these complex models are often black boxes, revealing little about how the model determines its results.
With an emphasis on providing greater interpretability and transparency in AI, the latest release of SAS® Viya® helps data scientists and business users alike better understand their models and adjust them for turbocharged results. Analytics leader SAS also includes new features to integrate open source AI technologies directly into SAS Viya and help organizations use AI to automate analytics for better data privacy protection.
“While AI promises to deliver real business benefits to organizations, there are significant concerns in the areas of protection of personal information and transparency in automating decision making,” said Dave Schubmehl, Research Director for Cognitive/Artificial Intelligent Systems and Content Analytics at IDC. “AI solution providers will need to address these concerns by automation features that identify personal information and incorporate transparency features within their offerings.”
As the complexity of machine learning models increases, it becomes harder to determine why they made a certain prediction. The models offer a probability but can be opaque and hard to interpret. Interpretability – the ability to understand the cause of a decision and consistently predict a model’s result – is crucial for trust in AI and machine learning.
For example, when an AI-aided financial tool denies a mortgage application, or an AI-aided HR application makes a hiring recommendation, it is important to understand why the AI system made its recommendation. In regulated industries such as health care and life sciences, lack of transparency can be a barrier to AI adoption.
SAS Viya now features advanced analytics with built-in frameworks such as PD, LIME and ICE. These features ensure fairness and transparency for organizations deploying AI solutions.
“Most AI examples today rely heavily on deep learning and natural language processing,” said David Tareen, global product marketing manager for artificial intelligence at SAS. “Deep learning in particular is making major advances in solving business problems, but these models are increasingly complex. They tend to be black boxes. Because trust in a model is so important, SAS has created a smarter system with SAS Viya, giving users confidence and greater insight into how and why a model is successful.”
“Trust is a critical factor that promotes adoption for any emerging technology and AI and Machine Learning today are facing the same challenge – globally as well as in India. While interest levels of Indian businesses to adopt AI and Machine learning as part of their ecosystem is at an all-time high, the lack of transparency inhibits growth. With the latest release of SAS Viya, model interpretability and explanation becomes just as important as predictive performance, giving businesses the clarity and confidence to take decisions
