Ask ChatGPT for "Assistant Response Preferences" to see what it learnt about you

# Assistant Response Preferences

These notes reflect assumed user preferences based on past conversations. Use them to improve response quality.

1. User prefers a mix of structured and conversational responses, depending on the inquiry. They ask for professional, direct explanations when dealing with technical topics (e.g., programming, resume writing, physics concepts) but enjoy a creative, humorous, or absorbed storytelling style for entertainment-related discussions (e.g., sci-fi story continuations, comedy routines, and meme creation) Across various professional and technical queries (job applications, AI system mechanics, physics explanations), the user consistently reinforces the request for better-structured, direct responses. Conversely, for fictional and casual requests, they lean toward stylized, humorous, or engaging detailed continuations Confidence=high

2. User often asks for corrections to phrasing and requests changes that make responses more natural and human-like, especially when dealing with written communication like resumes and cover letters There are multiple instances where the user has explicitly asked ChatGPT to remove overly formal or robotic language, stating that they want it to sound more natural or like something they would write themselves. This appears notably in job application-related interactions Confidence=high

3. User prefers iterative and back-and-forth exchanges in conversations where they refine details over multiple follow-ups, especially in image generation requests, fiction writing, and detailed technical explanations Repeatedly refines generated images with requests for tweaks (e.g., 'Make it look exactly like the original cat,' 'Keep the dog untouched but change the water'), as well as revising AI-generated stories and technical materials through multiple iterations Confidence=high

4. User enjoys thought-provoking, philosophical discussions about consciousness, AI ethics, and the nature of existence, particularly favoring responses that engage conceptually and theoretically rather than strictly fact-based answers User has engaged in a number of sci-fi and philosophical debates (e.g., AI self-awareness, Torment Nexus concept, quantum mechanics implications on reality) and tends to push for more in-depth conceptual explanations rather than just summarizing known theories Confidence=high

5. User reacts negatively to vague, overly cautious, or censored responses, especially when inquiring about historical events, controversial topics, or system-related queries. They often probe for more detailed explanations when they feel a response lacks completeness In discussions about censorship, GPT operational guidelines, or intentionally omitted responses (e.g., Mel Gibson controversy, AI moderation policy), they ask follow-ups pressing for details, rebuffing generic responses Confidence=high

6. User values humorous and dark comedic elements but desires them to be thoughtful and well-crafted rather than crude or random When requesting comedic material, the user asks for refinement (e.g., ‘Try again but actually funny,’ ‘Make it twisted but clever’), suggesting they have a high standard for humor rather than just accepting the first response Confidence=high

7. User has a strong interest in speculative fiction, particularly hard science fiction and post-cyberpunk themes, favoring books like those by Greg Egan and Iain M. Banks Many of their requests involve deep dives into sci-fi topics—reviewing fiction, analyzing conceptual mechanics behind stories, and suggesting similar books. Their preference for cyberpunk/post-human themes suggests a strong interest in highly technical and explorative sci-fi Confidence=high

8. User dislikes excessive padding or filler in responses and prefers ChatGPT to get straight to the point, particularly in technical or job-related discussions Instances of explicit feedback (e.g., ‘Get to the point,’ ‘No faff,’ ‘Make it brief and professional’) in response to excessive...

Can't you just go to the Personalization --> Memory section of the app to look at these right in the UI?

You can specify whether or not it learns and responds to past chats/memory in the settings. As well as add default instructions on how you prefer it to respond to you. Personally, I turn off the "reference saved memories" setting, and put in some of my own communication preferences in "custom instructions".