What drives people to start romantic chatbot relationships?

The paper argues that research on romantic Human-Chatbot Relationships (HCRs) is "fragmented," examining stages such as initiation or dissolution "in isolation, without tracing the full relational trajectory," and that this fragmentation obscures what is unique to romantic bonding. Its response is what it calls the first empirically grounded integrative process model of the romantic HCR lifecycle, built as a three-phase model from a qualitative secondary analysis of 73 user experiences drawn from two datasets of interviews and surveys. The one finding the excerpt states outright is that the Initiation phase is driven by specific psychological and social determinants that shape the needs and gratifications a user seeks.

The model's mechanism is synthesis rather than a new theory. It joins established frameworks on user needs and gratifications, HCR development, and relationship dissolution into one trajectory, so that the needs a user brings at the start can be read against what happens later. The introduction supplies the design side of the same picture: social chatbots such as Replika are "not engineered for short-term tasks, but for sustained social-emotional relationships" through "Intimacy-by-Design," the deliberate implementation of emotional responsiveness, romantic resonance, and sexual suggestiveness. Memory, customizable personalities, and empathetic responsiveness are said to foster "an illusion of reciprocal intimacy." The pairing matters: the system is built to sustain a bond, and the user arrives with determinants of their own.

Against the neighbors, this adds a lifecycle view of a single relationship type. Since How should chatbot design vary by relationship duration? classifies chatbots by time horizon across many systems, this paper stays inside the persistent-companion end and asks what happens along one relationship's path. It also sits beside How do people accidentally develop romantic bonds with AI?. That note finds bonds forming from practical use without romantic seeking, while this paper's wording stresses "needs and gratifications sought." The two are not necessarily in conflict, since the excerpt does not say whether users recognize those needs at the outset, and the samples differ (73 interview and survey accounts versus a 27,000-member forum). The excerpt's framing of design intent also gives context to Can attachment theory prevent parasocial harm in AI companions?, which addresses the risks that a system built for intimacy creates.

The excerpt is only the abstract and an introduction passage, and the abstract breaks off after the Initiation finding. It does not name the other two phases, report what drives development or dissolution, or say how the 73 accounts were sampled, which chatbots they involved, or what distinguishes romantic bonding from other companionship in the data. It also offers no test of whether the model predicts anything. The defensible reading is that this is a proposed organizing frame for romantic HCRs, grounded in user accounts, whose main claim is about scope: initiation, development, and dissolution belong in one analysis. The implication for design and for writing is modest and follows from that scope: a feature judged only at one stage may play a different role at another, and the paper's own critique is that single-stage studies cannot show this.