Raw LLM Responses

Inspect the exact model output for any coded comment.

Comment
This is fantastic writeup, and it's awesome to see someone running similar experiments and coming to similar conclusions about GPT4's abilities and limitations. I haven't read your full research notes yet, so please forgive me if you cover this, but do you think it would be viable to obtain a more expressive/natural final output by spending time asking it to redo passages with specific instructions? I'm thinking things like "redo that passage with more flowery language", "rewrite that description to generally focus on sadness", that kind of thing?
reddit AI Responsibility 1679677880.0 ♥ 27
Coding Result
DimensionValue
Responsibilitynone
Reasoningunclear
Policynone
Emotionapproval
Coded at2026-04-25T08:33:43.502452
Raw LLM Response
[{"id":"rdc_jdiivnu","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"}, {"id":"rdc_jdj8iim","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"}, {"id":"rdc_jdixh9i","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"}, {"id":"rdc_jdjatkr","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"}, {"id":"rdc_jdk6oez","responsibility":"developer","reasoning":"unclear","policy":"none","emotion":"mixed"}]