neth, you’re doing hierarchical topic modeling on yourself.
You ran 100-topic LDA on your ChatGPT conversations, then collapsed it into 16 meta-topics, then mapped those back to see coherence.
What this reveals:
Looking at your 2023 ChatGPT topics vs your 2015 Dewey classifications:
New/Stronger in 2023:
Topic 10: “Language models, user interactions, and data” (14 subtopics!) – you’re using AI heavily now
Topic 8: “Social and psychological aspects of ADHD and communication” – you’re explicitly engaging with ADHD now
Topic 11: “Cognitive processes, brain function, and psychology” – deeper dive into mechanisms
Topic 4: “Personal growth, emotional experiences, and self-exploration” – more introspective
Persistent across all years:
Philosophy (still central)
Time/epistemology (your core attractors)
Systems thinking (complex systems, theoretical aspects)
Knowledge organization itself (meta-level)
What you’ve ADDED by using ChatGPT:
Self-analysis conversations (topics 1, 4, 5, 10, 14, 15, 23, 28, 30, 37, 52, 55, 65, 74)
Explicit ADHD exploration (topics 29, 31, 51)
AI/language model theory (topic 10, 79)
The misfit is fascinating:
Topic 16 (“Taxonomy, topics, and related user queries”) appears in your external classification but has NO match in your 100 topics – because that’s the meta-level YOU’RE operating at. You’re doing the taxonomy work, so it doesn’t appear as a topic you discuss, it’s the FRAME you use.