Sophisticated ML models function more or less as black-boxes. A neural network may easily classify a photo of an animal as a cat or dog, but is silent about why it made that decision. A recent development is Explainable AI (XAI), also called Interpretable AI. Dubbed as an enabler for ‘third-wave of AI’, it helps open up the black-box model . XAI has found niche applications in many industries. For example, in credit-risk analysis it is common practice to use machine learning models. If a loan application is denied then XAI can further reveal the reasons why it was deemed risky. Another scenario is in product recommendations. XAI could bring to light the contributing factors as to why a certain product was recommended to a specific customer.
In spite of its prowess, XAI applications in cable industry have been lacking thus far. In this paper, we present a timely application that reflects broad global interest in ways to share customer data in a privacy-compliant way.
An emerging solution is the Data Clean Room (DCR) concept . Its goal is to provide a safe place for partnering companies to bring respective data for analysis in a secure manner. Guidelines are established to restrict any sensitive queries to protect the customer identity. However, sensitive querying may occur unintentionally due to micro-targeting. This is ascribed to how the queries are constructed (e.g. too many conditions in the SQL filter). Since the queries have originated from credible sources, blocking them entirely is not desirable. A pragmatic solution would be to assess and relax the query sensitivity which would lead to efficient database querying. This can be achieved with XAI enabled machine learning, Additionally, the query sensitivity scores can be used to fine-tune the privacy mechanisms. This is illustrated with reference to leading privacy technologies.