A week or so ago, I hosted a Knowledge Café exploring the impact of anthropomorphising AI: what happens when we start to interact with an AI as if it were another person?
It was a fascinating conversation. We explored how anthropomorphism can make interaction with AI feel more natural, while also encouraging us to attribute understanding, feelings and agency to systems that do not possess them.
I have now added a summary of the Café conversation, based on the transcript, to the event page. It captures many of the different perspectives that emerged, as well as some of the questions that remained unresolved.
Since then, I came across an interesting article in Psyche, The way we talk to bots matters even if they aren’t conscious. It adds another dimension to the conversation.
The argument is not that an AI deserves our politeness because it has feelings. It is that how we talk to AI may affect us.
Research described in the article suggests that our interactions with artificial agents can spill over into subsequent interactions with humans. In one series of experiments involving more than 4,000 participants, people who first interacted with a bot subsequently behaved less cooperatively towards a human than those who had initially interacted with another person. The researchers found evidence that reduced empathy played a part.
These are early findings, but they raise an intriguing possibility. If we repeatedly practise being impatient, abrupt or dismissive with AI, perhaps we are rehearsing conversational habits that do not remain neatly confined to our interactions with machines.
Perhaps every conversation is, in some small way, practice for the next one.
We don’t have to believe that an AI has feelings to care about how we speak to it. The more interesting question may be what our conversations with AI are teaching us about how to converse.
Knowledge Letter: Issue: 316 (Subscribe)
RSS: Blog Feed
Image Credits: Midjourney
(Public Domain)
