alt Hacker NewsShort of blindly outsourcing the work to an LLM and checking to see if its hallucinated problems are actually problems[0], how can more laymen (without the ability to run replication studies, and without degrees in statistics) learn to more easily identify questionable/potentially fraudulent research like this?
This is a question that has been really bothering me for a while. I have a short list of things I’ve learnt to check, but it’s all stuff I’ve picked up in a very ad hoc way (reading and listening to these kinds of critical analyses, mostly). For example, until now, I wouldn’t think about implausibly large Cohen’s d, despite this seeming like an easily generalisable rule that anyone could eyeball. But without being told, and without the reference points given in the article, I wouldn’t have a clue. And I would like to have more of a clue. And I would like other people to have more of a clue. How?