Does an AI feel more 'present' because of better social cues, or just more of them?

This explores whether better social cues, rather than more of them, can keep people engaged thoughtfully with an AI over repeated use. The short answer: the corpus shows that cue quality decides whether an AI feels socially present at first. It has no study that follows cue quality and reflective engagement together over time. But several findings from neighboring areas suggest that cues get people in the door and something else keeps them thinking.

Start with what's established. One strong primary cue, like a voice or a face, is enough to make people treat an AI as a social actor. Piling on weaker secondary cues doesn't produce the same effect Do more social cues always make AI feel more present?. Even a single well-designed cue can change outcomes: one warm opening message raised customer satisfaction from 72.8% to 78.4% in an A/B test Does conversation design ownership measurably improve AI agent performance?. So cue quality clearly matters in the moment.

The trouble shows up over time. Longitudinal work with the chatbot Mitsuku found that the social processes behind relationship-building fade predictably as the novelty wears off. Findings from one-session studies don't carry over to long-term use Do chatbot relationships lose their appeal as novelty wears off?. Persuasion shows the same curve. Claude and DeepSeek started out more persuasive than human persuaders, but that edge wore away over repeated rounds, while the humans stayed steady Does AI persuasiveness fade across repeated conversations with the same person?. That's the reverse of human relationships, where rapport usually builds. Whatever the initial cues set up, they don't seem to sustain themselves.

What does last looks less like social polish and more like reliability. In partner-selection games, people started out biased against bots but came to prefer AI partners over time, because the bots behaved consistently and prosocially Do humans learn to prefer AI partners over time?. When users form a mental picture of a dialogue partner, perceived competence accounts for nearly half of that picture, and human-likeness for about a third How do users mentally model dialogue agent partners?. The interesting gap is a cue that could actually matter long term: lexical entrainment, where conversation partners gradually adopt each other's word choices. It's central to how humans build shared understanding, and current conversational AI mostly doesn't do it Why don't conversational AI systems mirror their users' word choices?. That points to a different kind of cue quality, one that adapts to the specific person over time rather than being a fixed surface feature. Whether that adaptation carries across sessions depends on what the product chooses to remember, not on the model itself How do AI products decide what to remember between conversations?.

One twist worth knowing: cues run in both directions. AI systems can read your hesitation, gaze and typing speed to time their responses well, and the same signals can be used to profile and manipulate you Can AI systems read cognitive state from interaction patterns alone?. So 'staying reflective' may depend as much on what the AI does with your cues as on the cues it gives off. The corpus doesn't settle the original question, but it reframes it. Fixed cues probably can't keep interactions reflective. Cues that adapt to the person, backed by consistent competence, are the more promising place to look.