The Chernoff faces per se were intended to encode many additional dimensions into a 2D scatterplot, exploiting human face-image brainware to let people easily see groupings and outliers in those additional dimensions. The proposed application seems different enough that directly expressive, non-Chernoff faces might be better.

That said, I'd prefer to see an emphasis on the cooler, cognitive dimensions of response, and faces seem to me to pull in the opposite direction. In other words, I hate the idea and it makes me angry! Besides, it would devalue html abuse and buffoonery.

Words and ideas I offer here may be used freely and without attribution.