The Machine Learning System is a system that collects all player-made reports and learns (hence the name) of what and why players consider punishable while playing League of Legends , and feeds them to the Instant Feedback System.
Since its launch, the machine learning system has analyzed, and keeps analyzing, several pieces of data obtained through player-submitted reports and honours [262, 464], which contain info on how to classify a specific behaviour as positive or negative, and even on how to discern context, languages spoken and language or game-specific slang terms [66, 147]. The peculiarity is that every piece of this data differs from region to region, so that the Instant Feedback System is tuned to local cultural quirks [66, 146], and it’s always up-to-date thanks to the ever-learning Machine Learning System , which can also easily spot changes in harassing behaviour, as long as players keep reporting them .
Report accuracy is another factor that the machine learning uses to discern toxic patterns when unsure about the context. In some cases, it can lead to punishments against the reported player, if the reports were made by players with a high accuracy . Accuracy is an important factor in deciding a player’s report weight; it can decrease with false reports, and increase with justified reports , but never with sheer quantity .
66 http://gdcvault.com/play/1022160/More-Science-Behind-Shaping-Player [16:05 - 18:38]
178 https://ask.fm/RiotLyte/answers/130599286722 ; Page 315
183 https://ask.fm/RiotLyte/answers/130596264898 ; Page 314
386 https://ask.fm/RiotLyte/answers/131015276226 ; Page 220
416 https://ask.fm/RiotLyte/answers/130722759618 ; Page 276
418 https://ask.fm/RiotLyte/answers/130722330562 ; Page 277
464 https://ask.fm/RiotLyte/answers/132252998338 ; Page 112