Emoji are less abstract than words. That is to say, the shape of a word is less connected to the thing it depicts. It barely changes the perceived meaning of the word "face" if it's depicted in Arial or Times New Roman, but it makes a big difference to the perceived meaning of a hypothetical "face" emoji whether that face is smiling or frowning in different fonts.
When you have a concrete depiction of something, you are forced to provide details. The problem is that users cannot tell what details are standard and which are only in their font. Hence more guidance is needed for creators of emoji to ensure important, necessary details are consistent.
Scott McCloud discusses iconography and the differences between how words and pictures convey ideas in "Understanding Comics". I don't think he has a perfect model, but his Big Triangle is a pretty good starting point for a discussion. http://scottmccloud.com/4-inventions/triangle/index.html
The further you go left in that triangle, the more guidance is required to ensure that all implementations are depicting the same thing.
Han/Kanji are not really more abstract either, and they aren't always words -- sometimes they are used only for their sounds, for example. They are more efficient: they pack more strokes into a double-width character than any emoji could. But again, so what.
Indeed. Kanji are quite similar in abstraction level to roman letters. That's why I don't agree with your assertion that emoji and kanji are basically the same. The two are very, very different in how they convey ideas. Emoji are concrete, directly showing the thing they represent while kanji are abstract and mostly unrecognizable (more similar to English words).
A few typical examples would be 日 (sun), 人 (person) and 目(eye). No matter the font, the glyph for 人 will not imply details like gender. For emoji, extra care needs to be taken to strip out irrelevent details or you end up with problems like the 'pedestrian' emoji[1].
When you have a concrete depiction of something, you are forced to provide details. The problem is that users cannot tell what details are standard and which are only in their font. Hence more guidance is needed for creators of emoji to ensure important, necessary details are consistent.
Scott McCloud discusses iconography and the differences between how words and pictures convey ideas in "Understanding Comics". I don't think he has a perfect model, but his Big Triangle is a pretty good starting point for a discussion. http://scottmccloud.com/4-inventions/triangle/index.html
The further you go left in that triangle, the more guidance is required to ensure that all implementations are depicting the same thing.