Question $8$

$1\mathrm{.}9$ pts

Consider a recommendation system $($e$.$g$.,$ Netflix$).$ Suppose that there is a malicious user with $10,000$ fake email addresses. The user creates $10,000$ Netflix accounts, and each account has the same user characteristics $($same age, gender, likes, and dislikes$).$ However, each fake user provides contradictory movie reviews. For example, a given fake user may claim to like Romantic Comedy films, but it rates $"50$ First Dates" $($a Romantic Comedy$)1$ out of $5$ stars, whereas it rates "The Joker" $($an Action Thriller$)5$ out of $5$ stars.

What effect might this have on the recommendation system?

The system may give wrong recommendations to a user for a short time, but as usage history builds it will start to give better recommendations.

The system may give consistently wrong recommendations to a real user with no usage history.

The system may give consistently incorrect recommendations to a real user with a long, legitimate usage history.

The system may give incorrect recommendations to a user for a short time, but as more legitimate usage history builds, it will start to provide better recommendations.