A study published in the British Journal of Psychology found that people consistently expect other people to use AI-assisted communication tools more often than they would use such tools themselves.
The research, led by Zoe A. Purcell of Université Paris Cité, ran two experiments. The first surveyed 477 adults in the United Kingdom, ages 18 to 75, in August 2021. The second surveyed 765 adults in the United States, ages 18 to 93, in August 2022. Participants viewed examples of AI-assisted text, audio, and video messages and rated how acceptable the technology was and how likely they, or other people, would be to use it.
In the first study, participants rated open use of AI tools as more acceptable than secret use, giving average scores of 3.41 and 2.72 out of 5. They also said other people were more likely to use the tools than they were, with average scores of 3.73 for others compared with 2.80 for themselves.
The second study placed participants in separate groups so they never compared open and secret use side by side. Under those conditions, the acceptability gap nearly disappeared, with open use scoring 56.87 out of 100 and secret use scoring 53.62, a difference the researchers said was not statistically significant. The gap between self and others held firm, however. Participants rated the likelihood that others would use the tools at 71.41, compared with 40.44 for their own likely use.
The second study also found that participants judged their own potential use by a tighter standard than they applied to others. Acceptability judgments lined up closely with expected behavior when people rated themselves, but the link was weaker when they rated other people. Familiarity with AI tools was tied to higher acceptability and greater expected use, while the belief that AI use signals something negative about a person’s character was tied to lower acceptability.
The authors wrote that “self-other differences are a central factor for understanding people’s attitudes” toward AI-mediated communication. They noted two limits to the research: participants were never told who these “other” people were, and the data were collected in 2021 and 2022, before tools such as ChatGPT became widely used.
