11

11 Explainability You are given a movie recommender algorithm that uses a (black box) machine learning approach to optimize its movie recommendations to users. It can give three type of explanations: 1. The algorithm presents a plot of different movies on the x-axis (ordered by how well the movie fits the label 'thriller') and shows how likely it is that they are recommended on the y-axis. 2. The algorithm explains why it recommends certain movies to Alice: if a movie is a thriller, based on real- life events, or has action in it, the movie gets recommended. If the movie is a romantic comedy, science fiction, or documentary, it does not get recommended. 3. The algorithm explains why the movie Snowden was recommended to Alice: because this movie is a thriller and based on real-life events. 6.0p Which explanation type is represented by each example? Match the correct labels to the examples. Model explanation Model inspection Outcome explanation 1. Plot of different movies 3. Specific movie 2. Features of movies