The paper states that they studied the HTML form element interactions but “not the keystrokes or content.”
There's a big difference. Both are more invasive than we would like, but grabbing everything you type while in the app's browser is much worse
than measuring a true or false "did this person submit their comment or did they give up and leave it unsubmitted."
Tiktok is getting the content of the text, which could be sensitive info, and it grabs from every site you visit, not just the social platform itself.
But I think the main issue is using the data for allegedly targeting of protestors and Chinese political opponents, more than the depth of the data collection itself.