Amazon Patents “Toxic Player Grouping” System

Amazon Technologies has apparently filed for, and received, a patent from the US Patent and Trademark Office for “behavior-aware player selection for multiplayer electronic games.”  The filing was made back in December 2017 but was only granted last month.

In the filing, Amazon describes a system for “generating a multiplayer game session including a set of players with compatible preferences and behaviors.”  It points out that conventional matchmaking systems are set up based off player rankings and skill levels.  Such an approach, however, ignores elements of player behavior beyond actual gameplay, such as abusive language, persistent quitting, and other activities which one would consider to be poor sportsmanship.

Amazon points out that a one-size-fits-all definition of “toxic” or “non-toxic” behavior is difficult to arrive at, using an example of one player finding profanity “toxic” while not caring so much about persistent quitting, and a second player finding the quitting to be a “toxic” behavior but seeing no problem with profanity.  By using an algorithm to compare players’ listed preferences as well as records of their demonstrated behaviors in a game, Amazon’s proposed system would group players together in informal and discrete pools separate from the larger overall pool of existing players for any given game.

Food For Thought

The patent, as currently described, isn’t overly broad.  It does, however, seem to rely on certain assumptions.  The sort of node graphing which Amazon talks about, particularly with regards to demonstrated behaviors, carries some serious concerns about player privacy and data security.  And it seems to assume that people who act obnoxiously in games are accepting of other players showing that same behavior.  While it would be nice for a mechanism to help the sheep and the goats separate themselves, the practical methods of achieving that mechanism might be less acceptable than the very behaviors it is ostensibly attempting to compensate for.

Notify of

Inline Feedbacks
View all comments