The problem with Spam has always been the minuscule cost of doing it. If the cost can rise above the profit, spam could become unattractive. Imagine if spam attracted a sudden swarm of bots, just as a new email addy attracts spammers now.
And bots could do pretty well there. I recently read an article (cannot find it; but it may have been even linked from here) which explained the benefit of poor grammar in the spam (i.e., they want to put off more knowledgeable users as the cost of the follow on emails is significantly higher).
El cheapo bots have pretty basic English skills which can help them fit naturally into the spam target patterns.
The same logic that favours using an opening gambit that scares off all but the most unworldly of people works against using bots for the followup conversation. You don't want to finally find someone that genuinely believes that a random central African stranger might want to split $138,400,000 with them and has funds ready to send only for your bot to fail the Turing test by misunderstanding their question about how to pay the deposit.
Also, anyone smart and knowledgeable enough to be able to write bots capable of persuading exceptionally gullible people to part with their money is probably capable of moving to the next level and going after the slightly-less gullible, who are more numerous and richer.
It's also probably a bit flattering to the average scammer to assume the flaws in their patter are a deliberate filter pattern though. They're just as likely to open with a highly plausible (probably copied) offer of high value second hand goods on a listings website that probably attracts dozens or replies and then respond to each of those genuine expressions of interest from people willing and able to pay with badly-written response(s) that conflict with the detail of the original ad, not to mention totally forgetting that little old Scottish ladies don't have email addresses registered in the name of Nigerian men.
"It's also probably a bit flattering to the average scammer to assume the flaws in their patter are a deliberate filter pattern though."
It doesn't have to be deliberate. It probably evolved. If worse spelling and grammar work better than correct spelling and grammar, then even if no spammer ever consciously says "Ah, I need to make this crappier", the spams will evolve to be crappy.
Moreover, while our brains in this discussion are taking cognitive shortcuts and putting spams on a "correct <-> crappy" single dimensional continuum, the reality is probably even more complicated, and it's actually specific crappiness that works better than others. For instance, I could hypothesize that it's not just typos, but typos that the recipient plausibly believe are the result of a non-native speaker from the relevant country. (I say "plausibly believe" rather than whether the non-native English is truly representative of the relevant country, because that's what matters.) That's an obvious possibility; more subtle things are possible and probably even likely. And again, the spams will evolve into those more subtle possibilities and exploit them even if no individual participant is sitting there and trying to deliberately figure out the best spams to send... which is, obviously, almost certainly untrue as well. And whoever is sitting down to write The Perfect Spam is probably doing so with a lot more data and a lot less scruples than you or I would apply to the problem.
I've often thought one of our best anti-spam defenses is the sheer mailbox stuffing quantity of them. Even the best crafted spam about how I won the lottery I never entered looks a lot less plausible the thousandth time I receive it.
Occam's razor implies that if the people originating the scams mostly are based in central Africa (true) and display the specific crappy spelling patterns and linguistic quirks of many other moderately-educated central Africans using the internet for non-nefarious purposes (also true) one needn't assume that there's any particular process that leads people to write like their compatriots. Especially not when similar imperfections betray their attempts to masquerade as a Scottish widow selling their husband's canal boat or Australian trader that wants to buy your product if you'll send him the import tax.
I'm also little inclined towards scepticism there's much evolutionary optimization going on when the standard email scam format hasn't even adapted to get around now unbiquitous standard webmail spam filters, which are essentially orthogonal to the gullibility of the account owner.
Besides, wouldn't an evolutionary process that wasn't a conscious attempt to avoid generating false positives (or slavish copying by people that don't really analyse in much depth) tend to optimise for reusing the emails that generated more responses rather than less?
The big money in spam is not in selling Viagra, it's in selling spamming.
I'm acquinted with someone who does spamming for a living, mostly in the online casino segment. He does very well, and surprisingly polite company cares more about the money he has, than the way he makes it.
Replace "spamming" with "marketing" and that makes less sense. "The big money in marketing is not in selling Viagra, it's in selling marketing". Spamming is a form of marketing/sales, and Viagra is one of the products being sold.
