Preventing notification blindness

A method comprising:

• learning, using an iterative expectation maximization (EM) machine learning (ML) technique, parameter values for current intrinsic interest and affinity to be influenced nodes of a dynamic Bayesian network, the dynamic Bayesian network including the affinity to be influenced node as a child node of the intrinsic interest node, a job seeker status node, a member activeness level node, and short and long-term notifications nodes, and the intrinsic interest node a child of the job seeker status node and the member activeness level node;
• determining, based on (i) the learned parameter values, (ii) a value from the member activeness level node, and (iii) a value from the job seeker status node, an intrinsic interest of a user to visit a website;
• determining, at the affinity to be influenced node and based on probability distributions of the intrinsic interest node, the job seeker status node, the member activeness level node, and the short and long-term notifications nodes, an affinity for the user to be influenced to visit the website;
• determining, using the learned parameter values, determined intrinsic interest, and determined affinity for the user to be influenced to visit the website, a first probability indicating a likelihood that the user will, in response to viewing a badge notification on an app, turn off notifications or delete an app associated with the website and a second probability indicating a likelihood that the user will, in response to viewing the badge notification on the app, visit the website;
• and
• in response to determining the second probability is greater than the first probability, causing the app to include the badge notification when displayed on the user device or in response to determining the second probability is less than the first probability, refraining from providing the badge notification on the app on the user device.

A non-transitory machine-readable storage medium embodying instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising

• learning, using an iterative expectation maximization (EM) machine learning (ML) technique, parameter values for current intrinsic interest and affinity to be influenced nodes of a dynamic Bayesian network, the dynamic Bayesian network including the affinity to be influenced node as a child node of the intrinsic interest node, a job seeker status node, a member activeness level node, and short and long-term notifications nodes, and the intrinsic interest node a child of the job seeker status node and the member activeness level node;
• determining, based on (i) the learned parameter values, (ii) a value from the member activeness level node, and (iii) a value from the job seeker status node, an intrinsic interest of a user to visit a website;
• determining, at the affinity to be influenced node and based on probability distributions of the intrinsic interest node, the job seeker status node, the member activeness level node, and the short and long-term notifications nodes, an affinity for the user to be influenced to visit the website;
• determining, using the learned parameter values, determined intrinsic interest, and determined affinity for the user to be influenced to visit the website, a first probability indicating a likelihood that the user will, in response to viewing a badge notification on an app, turn off notifications or delete an app associated with the website and a second probability indicating a likelihood that the user will, in response to viewing the badge notification on the app, visit the website;
• and
• in response to determining the second probability is greater than the first probability, causing the app to include the badge notification when displayed on the user device or in response to determining the second probability is less than the first probability, refraining from providing the badge notification on the app on the user device.