The problem is, you seem to be making a blanket statement about the theoretical possibility of prediction and suggestion rather than arguing against the limitations of current predictive systems.
It's simply that current recommendation/prediction systems have incomplete information. In this case an explicit goal or preference is not stated by the users. In the absence of explicit preference statements, a recommendation system can only act based on the behaviors it measures.
The purpose of my life's work is to align revealed preferences with stated preferences, to determine optimal action. The majority of systems are confounded because 1. They don't ask for preferences and 2. People state preferences (signaling) that are not aligned with their true beliefs.
I am indeed making a blanket statement about the theoretical possibilities of prediction and suggestion as a preferred or comparable replacement for allowing the user to choose for themselves, and that statement is that it's neither comparable nor preferred.
Your work, at least as you present it, does not seem to collide with that statement because it asks the user for intent and then examines behaviors to see how they align with that intent... there is no inferring trap. Determining optimal action, while its own complex kettle of fish, is only a concern if it over-rules the user's decision-making WITHOUT the user opting in to such an arrangement.
It's simply that current recommendation/prediction systems have incomplete information. In this case an explicit goal or preference is not stated by the users. In the absence of explicit preference statements, a recommendation system can only act based on the behaviors it measures.
The purpose of my life's work is to align revealed preferences with stated preferences, to determine optimal action. The majority of systems are confounded because 1. They don't ask for preferences and 2. People state preferences (signaling) that are not aligned with their true beliefs.