How does AI affect the way we express emotion?

“

By default, AI behaves like a rather risk-averse and socially desirable writer. While this might not be surprising, given how much of today's writing involves AI, I think this default tendency will fundamentally change how language is understood and emotion is interpreted. For example, when we read on LinkedIn, we know (well, kind of) how to spot words AI loves to throw in as "linguistic additives": prose that looks eloquent if not grandiose but often fails to convey anything substantial. But it is far less clear what types of language have been formulaically downsized, swapped, or dropped. The low cost of AI's reformulation of human language reinforces "talk is cheap"--both in a linguistic and an economic sense. When every message is warm, warm language stops warming anyone, because nothing is cold anymore, at least not expressed so. In fact, the worry of homogenization has become a major criticism of AI-assisted content. But the discount of AI content also rests on one assumption: that writing on platforms such as LinkedIn is *believed* to be heavily AI-assisted unless proven otherwise. Strip the assumption away, and the discount may largely disappear. To test this, we recruited 170 human judges who had no idea which texts were AI vs human, and they ended up rating the AI versions as the better writing: higher on analytic style and reasoning depth, in addition to the emotional enhancement we found earlier. Now, think of the last heartfelt message you received, are you sure a human felt it?

”