One of the seeming puzzles about identity is how tags placed upon people don’t work as one might hope. For example if I tag somebody as “short,” the act can be many things: accurate, insulting, invasive, diagnostic, etc. You can’t treat the act of tagging independent from the context. You need to know a lot about the actor who generated the tag and your likely to need to know the conversational context the tagging took place in.

This problem is particularly subtle when the tag assigns a person to a group, i.e. when it ascribes group membership. It’s more fraught in that scenario because the boundaries of groups are always disputed territory. In some situations the boundaries are only marginally disputed, such as in the case of short people; but even then there are people who care deeply about the boundary and what it’s functional purposes are.

The problem becomes more acute when there is polarization being engineered on the boundary. When two groups are in dispute, when the rights of the group members are in play, when states are at war, etc. These cases are not rare, both because you can get two groups into opposition at any scale (marketing v.s. sales, offense v.s. defensive squad, freshman v.s. sophmores, etc. etc.) but also because there are numerous benefits that agents in these games can harvest from playing with these boundaries (hightened common cause, humor, separation of concerns, etc.).

So I’m amazed that Google has decided that one, just one, such case deserves a bit of special handling. This is the fundamental problem of a system like Google’s. All words are tags, all search terms have contextual meanings. I like that they have tried to do something. I hope they find a way to make it scale.