New experiment: does talking about an AI system in anthropomorphic terms (for example, saying that it "thinks" or "knows" something) affect people's perception of the system? Quite possible not.
https://arxiv.org/abs/2606.29121
No Digg Deeper questions have been answered for this story yet.
Anthropomorphic discourse about AI has been criticized on the grounds that it might distort public perceptions of a system: for example, people might trust it more than they should, or overestimate its capabilities.
New experiment: does talking about an AI system in anthropomorphic terms (for example, saying that it "thinks" or "knows" something) affect people's perception of the system? Quite possible not.
https://arxiv.org/abs/2606.29121
We found evidence that anthropomorphic descriptions *didn't* affect people's responses (we used Bayes factors to quantify support for the null hypothesis).
In our experiment (with 815 people), we tested if views about AI policy and capabilities are affected by whether it's described in anthropomorphic terms. We had participants read one of two versions of the same article, one with anthropomorphic language and one without it, and answer a series of questions about AI.
It's possible that anthropomorphism does have an effect on people's perceptions of AI, but we were unable to detect it because it's subtle, or requires repeated exposure over time.
We found evidence that anthropomorphic descriptions *didn't* affect people's responses (we used Bayes factors to quantify support for the null hypothesis).
In our experiment (with 815 people), we tested if views about AI policy and capabilities are affected by whether it's described in anthropomorphic terms. We had participants read one of two versions of the same article, one with anthropomorphic language and one without it, and answer a series of questions about AI.
Anthropomorphic discourse about AI has been criticized on the grounds that it might distort public perceptions of a system: for example, people might trust it more than they should, or overestimate its capabilities.
But there's also another possibility, which is that people are used to talking about technology in anthropomorphic terms ("the car doesn't want to start"), and are not misled by the metaphor even when they're talking about AI. This a distinct possibility and our study provides some empirical evidence for it.
It's possible that anthropomorphism does have an effect on people's perceptions of AI, but we were unable to detect it because it's subtle, or requires repeated exposure over time.
