I assume it is the same reason Facebook serves dodgy ads: it makes them money and they have had no liability for the material they choose to serve to users. There are a handful of types of content that could generate bad press (some sexual or racist content, for example), so they police those. But ordinary scams generate engagement and make them money, and since they are generally immune from liability for content under Section 230, there's no downside for them. I've always thought there might be an end-run around Section 230 where the algorithm is tuned to preferentially present fraudulent content, thereby implicating the company in the advertiser's fraud. But as far as I know, no one has tested that approach.