In AI-Mediated Communication: Definition, Research Agenda, and Ethical Considerations, Jeffrey T. Hancock, Mor Naaman and Karen Levy argue that the introduction of artificial intelligence mediated communication (AI-MC) will transform the way we interact with others, thereby impacting our culture and raising new ethical concerns around communication technology. 

The authors define AI “to refer broadly to computational systems that involve algorithms, machine learning methods, natural language processing, and other techniques that operate on behalf of an individual to improve communication outcome.” This definition and the very term “artificial intelligence” may seem intimidating or futuristic, but as the authors point out, practical, easy-to-use AI technologies are already beginning to exist in our lives. Auto-correct, predictive text, grammar correction, smart replies, auto-completion and auto-responses are just some examples of widely used and accepted AI communication technologies. 

The authors argue that these technologies are impacting culture in ways we may not be fully aware of. For example, research shows that predictive technologies are impacting language. For example, a study of “smart reply” suggestions revealed automated responses were overly positive. “Given its scale, Gmail’s overly-positive language suggestions have the potential to shift language norms and expectations even when communicators are not using these tools, and produce long-term language change over time.” This is just one example, and the authors believe that AI-MC will have effects on interpersonal dynamics, self-presentation, impression format and trust.

There are also concerns about the effects of AI-MC on policy, culture and ethics. The authors point out that since AI is trained on human behavior, biases are likely to be replicated and amplified by AI-MC. However, AI also has the potential to mitigate bias, such as issuing warnings when an individual attempts to post negative comments. 

The authors also made an interesting distinction between AI-MC and AI-human interaction. They define AI-human interaction as “the study of human interactions with bots and other smart virtual agents who do not represent other individuals.” Under this definition, Apple’s Siri or Amazon’s Alexa would be considered AI-human interaction. I would never have made this distinction myself, and it makes me wonder, since this distinction has been made, will it make it easier to regulate these technologies separately in the future?