Reaction picker

A reaction picker exists because most replies are not worth a reply. Someone posts the invoice numbers, four people need to signal that they have seen it, and four messages saying “thanks” would bury the thread they are thanking. So the interface offers a row of five or six emoji, one tap attaches one of them to the message, and the acknowledgement lands without costing anyone a notification they have to read. The picker is the small surface that row lives in, reached from a trigger on the message: a plus on hover at a desk, a long press on a phone.

Which emoji sit in that row is the whole design, because the row is a vocabulary, not a sample. Six is about the ceiling before scanning costs more than typing would, and the six have to cover the jobs people actually need: agreement, thanks, acknowledgement, delight, and the two or three a particular community has made its own. Recent choices deserve a place, since reaction use is heavily repetitive. And there has to be a way out to the full set, which is where the picker hands off to an emoji picker rather than growing into one.

That hand-off is the line between the two words, and it runs both ways. An emoji picker inserts a character into text the reader is composing: the result is part of a message they have not sent yet, it is theirs alone, and the panel is built for retrieval out of thousands of glyphs. A reaction picker attaches a typed response to an object that already exists, and the result is not a character at all but a count that other people join. Two readers picking the same emoji produce one reaction with a two beside it, which is why the attached reaction reads like a badge on the message, and why tapping your own again removes yours rather than adding a second.

Once a reaction is counted, the interface owes the reader an answer to “who”. A reaction group that shows the emoji and the number, with the names behind a hover or tap, is the usual answer, and where the set of people matters more than the tally it becomes an avatar group instead. Two accessibility obligations come with all of this and are routinely missed: the reaction has to be announced as a control with a name and a count rather than as a bare glyph, so a screen reader says “thumbs up, two reactions, press to add yours”, and the picker has to open from the keyboard as readily as it opens from a pointer, since a surface that only exists on hover does not exist for everyone.