Interesting work! I admit I just skimmed the post quickly, so sorry if it's been mentioned already, but: is there a way for users to report/flag bad suggestions? On YouTube, for example, there's a similar list of "bubbles" for topics and I keep getting suggestions of topics I'm not interested in because I care for adjacent topics (e.g. for one soccer team because I follow another one). I'd love to be able to long-press the bubble and then get a report/flag option in a pop-up menu, for instance.

Hi, what a great post! I am working on exploring the idea of how people make agreements about concepts in a multimodal setting(language grounded in vision, just like in the app), and I was wondering if you would be interested in sharing a subset of the dataset for academic research? The mode advanced conversational level of the users, the better.

This seems very interesting! I'm curious if you'd be open to share your dataset with a phd student to help with research