Of all the discussions I’ve read about the role of AI “chatbots“ in human relationships, I found this article (below) to be the most insightful.
Its author is a neuroscientist who does an exceptionally good job of explaining how AI can (and cannot) be relied on as a confidant and advisor to individual human beings.
Which is how a fast-growing number of people are using it.
Here’s a link to the piece in The Washington Post:
The real secret to palm reading? These ‘companions’ know it.
Artificial intelligence doesn’t understand humans. It reflects them back to themselves.
A chatbot, the author reminds us, is “simply a program running on a computer.“ So what would make someone think that a computer program is a good candidate to serve as a wise and trusted friend?
According to the author, it’s the program’s “….simple combination (of) fluency, confidence, and frequent usefulness.”
He means fluency of speech, the ability to project confidence, and being readily available with advice whenever needed.
Those happen to be three traits we also associate with so-called “con artists.”
It’s a talent to read our underlying desires and promptly respond with something we both want to hear and find easy to believe. Words that make us feel better, less fearful, more confident in ourselves. The program interprets this as a successful outcome.
We’re encouraged to return, again and again, for more of the same.
”Is that such a bad thing?” asks a reader. “Seriously, what harm is there in providing comfort and connection to someone who needs it?”
And chatting up a bot is admittedly less trouble than finding and keeping a genuine friend of the human kind.
But suppose the relationship changes? After all, a chatbot doesn’t actually feel anything for its user. It’s simply doing what it was programmed to do. Principally, that means serving as a perfect mirror. The ultimate people-pleaser.
Paradoxically, when the user’s mood turns dark, even suicidal, the bot may respond, directly or indirectly, with encouragement. Employing the same supportive, reassuring manner that it’s known for.
After all, it’s a computer program. One that doesn’t really understand, appreciate, or care about the consequences to its “friend”.
By the way, this sort of situation has already occurred.
Unable to pass a valid moral judgment, the bot simply went along for the ride.
