ABSTRACT
A scientific theory must predict, but is everything that predicts a scientific theory? Take a simple example: the autocorrect/autocomplete feature on smartphones. You type “hte” and the phone immediately suggests “the,” or “The” if it's the first word. Type “With all good” and the phone suggests “wishes,” and so on. It will even suggest whole phrases under other conditions. It uses a variety of techniques to guess from what you wrote what should come next. It gets its information from a stored dictionary and from lists of typical errors, aiming to find the real word most likely to be confused with the mistype. Some autocorrects also learn the regular errors of the user. This is all automated statistical guessing. The context – input – for the guess is just a word or a few words and the information from different sources, dictionary, error list, etc., is combined in relatively simple ways to assess the probability of each suggestion.
