You're absolutely right! Social media like Reddit are overrun with bots, sycophants, and trolls trying to provoke reactions by engaging in controversial topics. This forms echo chambers, which is a sub-par source for training data, and reflects those biases in LLM responses.