AI-based writing assistants for emotional tone: Investigating users’ acceptance and recipients’ perceptions in online negotiations

With the use of AI, modern writing-assistants can offer users recommendations on more subjective, socially relevant communication style aspects like emotional tone. While these recommendations commonly aim to affect interpersonal interaction outcomes, user acceptance determines the recommendations’ actual impact on the resulting messages and potential downstream effects on message recipients. The present research aimed to investigate this interplay of an AI-based writing-assistant’s recommendations for emotional tone, users’ acceptance, and recipients’ perceptions of the messages and the sender. Two online experiments used the context of online negotiations (Ntotal = 1,532, native English speakers residing in the UK) to investigate if users’ acceptance of a writing assistant for emotional tone was affected by whether its recommendations a) utilized AI-based functionalities and b) promoted a positive or negative emotional tone. Users tended to show greater acceptance of AI-based than non-AI recommendations, independent of whether a positive or negative emotional tone was recommended. A third study (N = 408) then investigated recipients’ perceptions of the produced messages. Recipients’ ratings of revised and original messages differed significantly regarding the messages’ emotional tone, their interpersonal impression of the messages’ sender, and their willingness to make concessions. Results thus suggest that the examined recommendations can elicit revisions to the emotional tone of interpersonal messages in the recommended direction and that these changes consequently affect the messages’ social effects. Hence, the present research illustrates the relevance of AI-based functionalities in writing-assistants for users’ acceptance and potential effects on interpersonal interaction outcomes once messages are sent.