With Artificial Intelligence (AI) promising to bring analytical power and cost savings to many customer facing functions, it is pertinent to ask how the deployment of AI impacts on the image of people / brands. For instance, if consumers believe that your online profile was developed by AI, does that impact on how they see you?
This was the question investigated by Maurice Jakesch, Megan French, Xiao Ma, Jeffrey T. Hancock and Mor Naaman via a series of experiments. The results are reported in the paper “AI-Mediated Communication: How the Perception that Profile Text was Written by AI Affects Trustworthiness”, available here.
Jakesch and his team created a series of profiles for AirBnB users, and then asked research participants to judge how trustworthy the person being profiled was. All profiles were developed by the researchers, but the research participants were led to believe that some profiles might have been written by AI. The research participants weren’t told which ones might have been written by a person vs AI.
The researchers found that, when participants thought that a profile had been written by AI, they trusted the person being profiled less than if they thought that the exact same profile had been written by the person. The profile was the same; the only thing that varied was the perception of who had written it.
To be clear, the issue here is NOT about trusting the veracity of the profile. Indeed, the research participants had no concerns about how accurate the profile was. Rather, they stopped trusting the person being profiled. That is, the suspected use of AI makes people trust other people less.
In a different study reported in the same paper, Jakesch and his team conducted a similar test, but in this scenario, they clearly labelled the reviews as either being written by a person or being written by AI. Again, all profiles had been written by the researchers themselves. In this case, the perceptions of the person being profiled did not vary significantly between the two groups. That is, for any given profile, there are no significant differences in perception about the person being profiled, when the research participants believe that the profile was written by a person vs when they believe that the exact same profile was written by AI.
The conclusion is that, in a mixed human-AI environment where there is uncertainty about the use of AI, the image of the person / brand thought to be using AI suffers. Clarity about the use (or non-use) of AI, makes the person / brand seem more trustworthy.
The authors of this paper went on to suggest that people might resent the (undisclosed) use of AI, thinking that the person was too lazy to write their own profile. What other factors could explain that the suspected use of AI makes people trust other people less?