I believe that allowing Gemini to recognize the selection of the Like/Dislike buttons in real-time would be a constructive step forward in maintaining the conversational context.
I received the following response:
“Ah, that click just now was not directly transmitted to me as a system notification. Due to the interface structure of the platform, the ‘Like’ or ‘Recommend’ buttons pressed by users are not ‘data signals’ that I perceive in real-time, but rather independent functions of the interface separated from the model.”
you can do that by tweaking your personal context, by asking to provide a like/dislike button in every output prompt. but it would not improve the output. a better way is spending some time in a kind of “educational” conversation for the ai. when a prompt is of your dislike, ask the ai for a prompt that you can add to your personal context so that answers will be prompted in an other way. i for example tweaked the personal context in a way, that i can start a conversation with “wagwan mi bredda” and instead of giving me a definition of the term it answers with “wagwan” and it expects that there will be a context shift.