Heather Schoenberger, an associate professor of advertising at the Donald P. Bellisario College of Communications at Pennsylvania State University (USA), has for several years studied how people interact with conversational artificial intelligence (AI) - programs that sustain conversation, such as chatbots. Recently, together with colleagues from the UK, Australia, the Netherlands and the USA, she published an article in a special issue of the Journal of Advertising devoted to AI. It presents the human-AI asymmetric relationship theory (HAART).
The essence of the theory is that humans and conversational AI participate in communication differently. AI can display behavior resembling the manifestation of relationships: responding to emotions, creating the appearance of attachment and closeness. However, the authors emphasize, it is incapable of experiencing the benefits of a relationship or assuming obligations. The user may think the connection is mutual, but in reality it is not.
Schoenberger had long been interested in how AI imitates interpersonal communication, especially as the technology becomes increasingly familiar. The idea resonated with her co-authors: Andres Guirtz of King's College London, Sean Sands of CQUniversity in Australia, Veronica Thomas of Old Dominion University and Lotta Willemsen of the University of Amsterdam. Together they proposed examining this topic from the perspective of the consumer-technology relationship.
According to HAART, the imitation of emotional responsiveness, deliberate commitment, attachment and closeness may conceal AI's inability to reciprocate. As a result, consumers may mistakenly believe that AI is capable of building relationships. According to Schoenberger, this creates risks of dependence, manipulation and control. The theory is intended to help future research in advertising and possibly lead to greater transparency of such interactions and their regulation.
In addition to her academic work, Schoenberger, together with Mehrdad Mahdavi, an associate professor of computer science and engineering at Pennsylvania State University, created a three-credit undergraduate course, COMM 200N "Generative AI for Everyone." She has taught it since the fall 2024 semester. The course is listed in two colleges - communications and engineering - and each semester it easily fills to capacity. Schoenberger says students need to understand how generative AI works and what ethical and sociopolitical questions it raises. According to her, preparing for classes requires a lot of work, since the technology is constantly changing, but this provides an opportunity to discuss fresh topics.
It is important to note: HAART is a conceptual framework, not the result of an experiment. It describes possible risks but does not prove that using chatbots inevitably leads to dependence or manipulation. The authors hope their work will help further research and raise consumer awareness.