"One of the reasons that people perform poorly when trying to detect deception is the difficulty of integrating multiple cues into a binary judgement. A simple heuristic of only judging the level of detail in the message consistently allowed people to discriminate lies from truths."
This is one reason why AI hallucination is so scary - it makes up false facts with a very high level of detail and even makes up fake sources to cite, or cites real papers but lies about their contents (in great detail). Scary.
After working in tech, I can confidently say this mirrors tech management. It is scary. Watching whole teams of people jump on a bandwagon of complete lies in order to save their positions is super scary.
When you program AI to behave like humans (particularly techie humans) it's less surprising that we see this outcome. It's embedded in the training data.
