,
9 tweets,
2 min read
Read on Twitter
(Contd) -> This is where Theory of Mind (ToM) comes into place. Theory of mind is the general understanding that one has about the other person -> in terms of what they know, they don't know, their beliefs, their wants and needs and what they might be thinking about right now.
Essentially you both try to reach a "common" understanding of the topic at hand. This all comes way too naturally to us. For example, you know when the other person hasn't even completed their sentence but you know what they are going to say? That's one example of theory of mind.
We human beings are social beings, built on interaction. Communication is one of the factors through which we have evolved cognitively as well. Babies learn very early to learn things about the world using interaction. So in some ways, it is embedded into us.
Let's go back to french press example
Me: You can make the coffee in the french press
Person: Hmm, I don't know how to use one of those
Me: It's easy. You just have to put 2 tbsps of ground coffee in it, put boiling water, let it steep for 4 minutes.
Me: You can make the coffee in the french press
Person: Hmm, I don't know how to use one of those
Me: It's easy. You just have to put 2 tbsps of ground coffee in it, put boiling water, let it steep for 4 minutes.
Me: Press it down and then it's ready.
Person: Steep means just let it stay right?
Me: Yeah, you got it.
Person: Okay cool.
Person: Steep means just let it stay right?
Me: Yeah, you got it.
Person: Okay cool.
With a robot:
Me: You can make coffee in the french press
Alexa: You can find french presses for 20$ on amazon.com
Siri: You can get coffee in these 5 coffee shops
Robot: I don't understand
Me: I think I need a glass of wine now :'(
Me: You can make coffee in the french press
Alexa: You can find french presses for 20$ on amazon.com
Siri: You can get coffee in these 5 coffee shops
Robot: I don't understand
Me: I think I need a glass of wine now :'(
Obviously, I'm exaggerating a little bit, but not quite either. This task is very complex. The problem here, is that when these agents/robots hear an instruction that they don't quite understand, they either respond to their best assumption of what you said or just fail.
Unlike, humans, they don't have a theory of mind, they can't guess what you want, nor do they quite know how to get to that understanding.
But here is where the twist occurs. We, as people talk to agents who can't communicate like us very often. And we manage pretty well. What agents am I referring to? Can you guess?
