I see bots being extraordinarily effective against this type of automated scoring as the sophistication of the scoring bot seems like it will always lag potenti

I see bots being extraordinarily effective against this type of automated scoring as the sophistication of the scoring bot seems like it will always lag potential creator bots. Given that scoring bots in mass use will necessarily decrease the dimensionality of the data into fewer, more tractable features, and thus creating bots need only to target the dimensionality-reduction step, rather than reproduce "human behavior." Which seems rather simple given the variance in natural language and normal human social interaction.