309: Conversational AI: Trust, Values and the Illusion of Understanding with Dr Malak Sadek

And what's the impact you're finding?

With chatbots and LLMs now, everything is kind of on steroids.

Everything essentially has an impact, from something as small as the gender of voice assistant, whether it's male or female, to what it's saying to you, to how it's speaking, to what it refuses to answer, for example.

All of that has an impact, especially as we use these things for, like, months and years.

The impact just compounds and compounds, and you're left with a lot of psychological effects.

So for example, like I came across a study that was mentioning that because a lot of voice assistants have been female-voiced by default, it can lead to negative impressions of people viewing females as assistants, for example, or secretaries or things like that.

So there is a lot of psychological impact that we are discovering, and I'm sure there's plenty more we're going to discover, and there's constantly new stuff coming out.

Yeah, it's very interesting, and we're going back to our own biases in society, in a sense.

Just had a conversation with a colleague earlier today about also skin color or provenance of, for instance, of the coach bots or avatars.

Many are tend to have, for instance, an African American woman there.

And again, it's like, "And I'm there to serve you." [laughs] So like we're discussing where are we going with this.

So it's the same with the female.

I totally hear you.

So that's very interesting to look at the impact it all brings.

What do you find is different between conversational AI compared with other technology?

technology?I think because conversation is something that's so natural to us as humans, and it's something that you already know how to do versus any app or any computer, anything, you have to learn at some point how to use it, and there is a kind of learning curve there.

But anyone can just have a conversation, and I think the ease of that just blurs some lines sometimes, especially now because they're much more powerful, they're much more natural sounding.

You'd be surprised how often even I feel it sometimes when you're chatting away with a LLM, it's not that you necessarily will treat it like a friend or you know it's not a human being, it doesn't have feelings, things like that.

But for example, the amount of people who still will say please and thank you when they're talking to them or, you know, they wouldn't ever think about treating them rude, things like that.

It's just very interesting to see that creep up despite you being consciously aware, and obviously we don't see that with other apps or things that don't feel so human-like.

And as we start seeing them take on more social use cases like companion bots and these roles that we typically reserve for human beings, the lines just get blurred more and more.

And it's interesting as a researcher, but it's also very scary as a user and human being.