Social media companies are working to keep you on site or attempting to deliver results useful to you. You can see this personalization reflected in the results of search engines, which try to maximize relevance: searching for “bakery” in New York should return different results than searching in Paris. Features like the little predictive-text cues of autocomplete also look to maximize relevance by suggesting refinements: typing in “cats” might prompt the user with “cats movie” or “cats that don’t shed.” Autocomplete suggestions offer a look into the algorithm’s understanding of accumulated user behavior ― an immensely important part of how key social media algorithms work. The outputs are personalized, but the data sets are massive, and the platforms are mining for similarities across vast quantities of user information. If a whole lot of users near you or similar to you complete the search that you’ve begun with specific words, the autocomplete algorithm takes that into account and may be more likely to suggest them to you. Since the results are derived from data related to millions of searches, autocomplete is, in a sense, holding a mirror up to society