In this podcast episode with the founder of conversational AI, he describes the need to make spelling mistakes (and correct them) in order to help establish that the bot is actually a human.
https://podcasts.apple.com/us/podcast/the-python-podcast-ini...
I think making sublte spelling mistakes is a much clearer sign that someone is human. The imperfection without correction makes it more believable. I still think the hackers could stand to take a creative writing workshop.
Re:Email from FBI Looks Odd
by AbrahamParangi on Saturday November 13, 2021 @10:23AM
(#29210359)
The argument is that the hacker’s operational costs are massively dominated by the manual work of social engineering, so they have a huge incentive to filter out people who are less responsive to social engineering.
If you accept that some people are more credulous than others, it becomes the best strategy to optimize for only talking to people who believe you.
I don't think there is a binary smart population and dumb population to optimise around, for every step down, some people who are otherwise convinced become hesitant and waste time, and some of that group become totally unconvinced.