When the inclusive emoji were first introduced in 2015 there were fears that the icons could be used inappropriately or abusively, and could even stoke racial tensions.
But when researchers analysed a sample of a billion tweets they found that most people who choose to personalize the colour of their emoji do it to make it more aligned with their own skin tone.
And even in tweets where the selected skin tone was different from that of the user, posts were found to be mostly positive.
The researchers, from the University of Edinburgh have called for even more options to personalise characters to a greater degree.
Alexander Robertson, of the University of Edinburgh's School of Informatics, said: "This result should encourage the addition of more emoji options for self-representation - adding to those that have been recently made available, such as red hair.
“I would expect to see options like more combinations of skin and hair colour (at the moment the number of combinations is restricted, for example you can’t have blonde hair and darker skin) and perhaps a wider range of skin tones than the current five.”
The study found that users with darker skin, according to their Twitter profiles, were most likely to modify their emoji, while users with light skin were less likely to make any changes. This underlines the importance to users of being able to express their identity online, researchers say.
Almost half of modified emoji used a light skin tone, because most users are white, the study found.
The darkest skin tones were used least throughout most regions of the world, even in Africa, perhaps reflecting the lack of internet access in developing regions.
Dr Walid Magdy, of the University of Edinburgh's School of Informatics, who led the research, said: "The introduction of skin tone choices for emoji has been a success in representing diversity and their extensive use shows that they meet a real demand from users."
The study, by scientists at the University of Edinburgh,will be presented at the 12th International AAAI Conference on Web and Social Media in Stanford, California, in June.