The First Step in Social Ability: Not Responding, but Noticing the Invitation

Social interaction is often reduced to "who speaks first," but what truly determines the quality of a relationship is often the step before anyone speaks: someone notices another person's invitation and creates space for a response. This is the first piece in LOGIHEART's series on human social development, and also an entry point for understanding how physical AI should "interact with tact."

A True Response Begins with Coordinating Both Parties' Circumstances

A scene from one evening: a mother is working at her computer, and her daughter says, "I want to talk with you for a bit." The mother looks away from the screen and says, "I'll listen once I finish this paragraph." At a natural pause point, she closes the laptop; meanwhile, the daughter organizes her thoughts and waits.

This is, of course, a fictional example, but the key question is not whether the mother immediately stops working—she acknowledged the invitation, explained the wait, and made the next responsive moment predictable. The daughter also took the "unfinished task" into consideration. Response speed itself cannot measure the quality of this coordination.

Treating "always being responsive" as the entirety of social ability overlooks a fact: responses happen between two people who each have their own circumstances.

Back-and-Forth Is the Foundation of Early Development

The "Serve and Return" concept proposed by the Harvard Center on the Developing Child points out that when adults respond to children's sounds and actions and form back-and-forth exchanges, this is an important foundation for early development. Its mechanism can be broken down as: receiving a signal and connecting it to the next interaction. See the Harvard Center's related concept page.

Infant-caregiver interaction patterns cannot be directly transplanted into rules for adult conversation. Adults have responsibilities, fatigue, and varying relational boundaries. But this perspective still has value: noticing signals that "invite a response" can help us re-examine everyday scenarios. Rather than assuming that "this naturally gets easier with age," it is better to look at how time and environment make responding difficult.

Serve and Return is a core metaphor proposed by the Harvard Center on the Developing Child in its "Key Concepts in Child Development" series, used to describe a back-and-forth pattern of interaction between infants and caregivers: an infant "serves" through babbling, facial expressions, gestures, or crying, and if an adult can promptly notice and "return" it with eye contact, language, touch, or sound, then "hits the ball back," a complete interaction round is formed. Neuroscience research shows that this sustained bidirectional exchange strengthens neural connections in the brain responsible for language, cognition, and emotional regulation; by contrast, a long-term state of "serving with no one returning" may have stressful effects on development. It is worth noting that this concept describes the qualitative features of developmental interaction, not a set of operational procedures that can be directly copied. When translating it into rules for adult conversation or guidelines for AI behavior, one must consider adults' existing autonomy, responsibilities, and boundaries.

Being Present Does Not Equal Being Willing to Talk

Sharing the same room and wanting to converse are two different things. Eye contact, hearing one's name, or someone stopping nearby does not constitute permission that "a conversation can begin." Conversely, a person who finds eye contact difficult should not be treated as cold.

Observable social actions here include: responding to signals, making waiting periods understandable, and honoring the opportunity that has already been promised.

If the other person does not respond, do not immediately draw a conclusion—did they not hear, were they unable to respond, or did they not want to participate? Keeping these possibilities open prevents "being responsive" from sliding into surveillance or disruption. This distinction is equally important for designing machines that proactively approach people.

A Design Hypothesis for Physical AI

A robot that starts speaking as soon as it detects a person may only be reacting to "the person's presence" while ignoring "the person's circumstances." One possible design approach is to separate four stages: human detection, determination of a candidate invitation, whether conditions for responding exist, and execution of the response.

A concrete approach could be: give a brief acknowledgment of clear invitations; stop if the other person indicates they do not need to talk; if the interaction is postponed, update the other person's expectations when you can truly be helpful, rather than promising a timetable that has not yet been confirmed. These conditions should influence not only generated wording, but also physical actions.

Evaluation dimensions could include: unwelcome approaches, forgotten postponed interactions, whether corrections are accepted, and the proportion of appropriate responses. It should be noted that this is a proposed evaluation approach, not performance data already measured by LOGIHEART.

Physical AI refers to intelligent systems with a physical form that can perceive, move, and interact with people and the environment in real physical space, such as service robots, collaborative robotic arms, or mobile devices with autonomous navigation capabilities. Its key difference from purely software-based AI (such as chat assistants) is that its decisions are expressed not only as text or speech output, but also as embodied actions such as movement, approach, lingering, and turning, and these actions themselves carry social meaning. A robot approaching within one meter while a person is working and starting to speak feels completely different psychologically from a notification popping up on a screen. Therefore, the social design of physical AI must incorporate "whether to act," "when to act," and "with what physical posture to act" into consideration, rather than focusing only on what wording to generate. This is precisely why the article advocates separating detection, invitation determination, response conditions, and response execution.

A Question Worth Pondering

When you are waiting for a reply, what signal makes you feel at ease? Conversely, even if the other person responds quickly, what kind of repeated approach makes you uncomfortable? Spelling out the difference between these two is itself an entry point for studying social ability.

About This Series

This series uses public literature to discuss everyday scenarios. Its 7 domains and 22 capabilities are an editorial framework, not an academic classification standard. Differences in age or culture do not represent a ranking of social value or ability. The AI design ideas in the text do not mean that human-like emotional experience already exists or that product performance has been verified. This article was written with AI assistance, and the illustrations are fictional examples generated by AI.

LOGIHEART is researching how to implement social ability in physical AI that interacts with people. The research vision can be understood through the LOGIHEART official website and its English introduction video.

LOGIHEART is the research project to which the author of this article belongs, focusing on how to implement social ability in physical AI that interacts with people. Its proposed "7 domains and 22 capabilities" is an editorial organizing framework used to connect scattered everyday scenarios with public literature, rather than a peer-reviewed academic classification in psychology or robotics. When citing this framework, readers should treat it as a discussion tool, not a capability rating standard. The AI design ideas mentioned in the article are all proposed hypotheses, have not been validated through productization, and do not mean that the system already has human-like emotional experience; related research vision can be found on its official website and English introduction video.