That’s because most debates are not won based on reason. In fact a large part of how the world works is mostly based on how you say something not what you say. Perhaps the world of engineering and mathematics (and subjects which strongly depend on these) are the only ones where reasoning is a stronger lever in an outcome (outside of perhaps economic incentives).

Please give me an example where GPT4 reasons worse than an average adult college educated human. Any topic, STEM or otherwise.