You could easily do a zero-knowledge proof thing where you transmit a bitmask of emoji used over the next 365 days, with N bits randomly permuted. In aggregate, you'd still be able to count usage without saying definitively someone used or didn't use a particular emoji.