Fat Finger panelists give their opinions, and in exchange we give them information — information about the US public with interactive poll Results and Graphics, and information about themselves with Attitudes and Research Polls.
Recently we added new information for users – personalized movie suggestions. We make the recommendations using a machine learning technique called Singular Value Decomposition, which uses your movie ratings to identify common ‘factors’ of movies you like, and to identify other movie raters that are ‘like you.’ From this, it identifies movies you are odds-on to enjoy.
We put ‘factors’ and people ‘like you’ in quotes because with machine learning we don’t know the factors, nor what makes certain people like you. This is how machine learning differs from other forms of predictive data analysis such as regression – humans cannot entirely deconstruct how it works.
60ft tall mechanical King Kong used in Dino DeLaurentis' 1976 King Kong. Getty // Mirrorpix
To get your movie recommendations, go to our Movies page, and click the ‘Training Set’ button. This will feed you 50 movies at random to rate. Rate as many of those as you are able, including movies you like and dislike. If you haven’t rated 50 movies after the first Training Set, click the Training Set button again to rate movies from a new set of 50, and so on until you’ve rated 50 movies. At that point a ‘Recommendations’ button is enabled -- click it to see what movies the model thinks you will enjoy.
Also at that point, the ‘Training Set’ button becomes a ‘Pot Luck’ button, though the functionality is the same – click it and you’ll get 50 more movies to rate. When you rate additional movies, your recommended movies will change to reflect the new information. If there is a movie you’d like to rate you can search for it using the search icon.
Hells Angels premier at Grumans Chinese Theater, Hollywood, 1930. Getty // Camerique
Be aware your Fat Finger recommendations will probably differ from what mass streaming services such as Netflix or Amazon will recommend because they base their recommendations on the popularity of movies that are similar to one you just watched, rather than on your idiosyncratic preferences. We are trying to do something different, to use your preferences to identify movies you’d like without the same emphasis on their overall popularity. In othe rwords, to help you find movies you might not otherwise have thought or known to watch.
We hope you enjoy this new feature and find the recommendations useful. Please let us know what you do and don’t like about it using the feedback button on the movie page.