AITA human vs random-AI
AITA: Human vs random-AI . PsyProxy quantitative evaluation. The excerpts reflect a diverse range of interpersonal conflicts and ethical dilemmas, primarily centered around themes of **family dynamics**, **relationship boundaries**, and **communication**. Many situations involve discussions of whether individuals are the "asshole" (YTA) or not (NTA) in various social contexts, highlighting issues such as **expectations**, **traditions**, and the importance of **empathy**. The tone fluctuates between serious and lighthearted, often encouraging self-reflection and open dialogue. Recurring topics include navigating **sibling relationships**, managing **personal autonomy**, and the complexities of **supporting loved ones** during challenging times, all while emphasizing the need for respectful interactions and understanding differing perspectives. [Summary on 50 random texts by ChatGPT 4o Mini].
Reported holdout systems from the verified card
Binary classification uses FVE as the task-primary metric. Secondary columns keep the companion metrics visible so binary, ordinal, regression, and multiclass cards are not compared through one flattened score.

