Yes, I completely agree that we are vulnerable to bots. The APIs are wide open by design. It will be interesting to see how it evolves but ActivityPub is supposedly designed with this stuff in mind. Yes it's easy to act maliciously and create accounts. There are a million ways to attack. It's a fact of life, sadly. Also probably the natural order - we've been chucking rocks at one another for millennia.
It happened on Usenet. It wasn't a paradise, it was full of spam and trolls and bots. However the fragmentation and self policing of the Usenet groups somehow kept the experience tolerable. Maybe we just expected less. Lemmy reminds me a lot of early BBS days. Not even any spam so far. It's remarkable - but probably temporary. I'm liking it though.
Mastodon is built on ActivityPub and seems to be thriving. I don't see spam or problems there so far. It seems quite civil. It's more like Twitter in format than Lemmy is, but the big instances have dealt with DOS and malicious actors and seem to be coping ok.
Your work sounds very cool. The development over the past few years in data analytics and machine learning is indeed startling when the ability for deception and manipulation is so easily scalable, but I think we will find ways to isolate or mitigate these issues gradually. It might take years and a lot of suffering, but innovation to solve problems is also our natural inclination.