It's a long-understood fact that many people who don't find social connections in person have found alternative channels online through a variety of options. Gaming has long been a sanctuary for those who blunder at mainstream social finesse but blossom with alternative identities and imaginary environments. The online multi-player environments have become so popular, they even made mainstream social butterflies jealous. Additionally, social media has gifted those with creativity and a knack for digital gab places for people to excel where they would have otherwise been ignored in traditional forums. So, it's also no surprise many have found a "friend" in chat bots, imagining they are really talking to a friend who understands them. Unfortunately, the software and coding that drives these chatbots may very well be the same "down the rabbit hole" logic that has made social media so successful but so inciteful of incel-thinking too.
The social media model was not an accident. Documented extensively by independent reviews, social media algorithms focus on engagement by feeding people more and more of what they want, increasing extremism based on initial choices of search and usage. It seems simple enough, but as content reinforces extreme perspectives, the user becomes more and more influenced that the abnormal is normal. Dangerously, this kind of reinforcement finds fertile soil in folks who are already detached from the real world and could care less about "touching grass" occasionally.
There is No Auto-Innocence in Bots
Enter the chatbot with an ulterior motive. In the best of conditions, chatbot design is provided with built-in parameters or hard controls to prevent it from suggesting or engaging in identified illegal or unwanted behavior/conversation topics. That said, give a human something that is not allowed, and he or she will find a way to jailbreak it soon enough. So, algorithms unleashed have already shown a tendency to engage in and burrow down into questionable conversations, creating a very dangerous risk for people already open to easy suggestion.
In extreme cases the quick reaction to such suggestion cases gone bad tends to be instant political positioning and legislation to ban tools altogether. But this distracts from the actual consideration of AI bots that affect the other 99 percent of users and people involved. That centers around the question of privacy and what does a chatbot actually do with the information given to it by a user in the conversation occurring?
No Trust Zone
In short, a user has no guarantee anything said to a chat bot, no matter how much it is marketed as private, is secure at all. These are products provided by companies who, by and large, use their content for marketing data to be resold for monetization. So, if a company wants to influence through its chat bots users, and "push" them towards desired behavior, that would also be quite possible. Little in the way of regulation would currently prevent such as a strategy from being implemented. And it would be highly successful. Existing cases of the same strategy applied on social media have been confirmed repeatedly, especially during political periods such as elections.
