Internet platforms tend to enable scams to spread and thrive, both because of the anonymity and global reach that the internet provides, and the capacity to scale up scam operations by utilizing automation that platforms offer. In concert, these forces often enable scams with even infinitesimally small rates of successful victimization to become profitable enterprises.
Scams on the internet range a gamut from highly personalized frauds designed to appeal to the online profile of an individual target, all the way to spam emails proffering goods to any email address that will accept delivery. Additionally, scams come in all shapes and sizes on platforms: they're often directly spread through messages, try to game recommendation algorithms, and abuse search tools to find potential victims. Despite this range in tactics, approach, and mechanisms, because scams are consistently financially motivated, a simple market-derived idea points true: if platforms can slightly increase the cost to using a platform for a scam, or slightly decrease the risk that a potential victim will bite, the scammer will seek out better returns elsewhere.
This, and a few other commonalities within scams, point to mechanisms by which they can be stymied in online environments:
It's also worth noting: as large-language models proliferate the capability to build lots of high quality human-sounding speech, content analysis techniques are going to become less and less valuable for detecting and acting upon scams. It's essential that platforms start thinking about how to build their systems to safeguard against scams from the lens of features, and soon.