Social capabilities across human development | Part 5
A person's movement can suggest what they are trying to do, but similar movements can serve different goals. Social assistance involves checking the person's aim before supplementing their actions.
Asking for a tool does not mean asking someone to take over
Two friends are making a small birdhouse. One reaches toward the back of the table. The other points to a blue screwdriver and asks, "This one?" After a nod, the screwdriver is handed over. The original builder continues the assembly.
In this fictional scene, help removes a small obstacle to accessing a tool; it does not take away the task. Rushing to finish the birdhouse for the person might produce the object while missing their wish to build it themselves. The task outcome and the person's goal need to be distinguished.
Thinking from early helping research
Warneken and Tomasello examined young children's behavior supporting others' goals across several experimental situations. An important research question concerns understanding actions in relation to another person's goals, alongside one's own actions. See Altruistic helping in human infants and young chimpanzees.
Helping in an experiment does not establish understanding of someone's long-term wishes. Read findings within the conditions investigated: what information prompted help, and when help did not occur. The birdhouse scene and AI proposals here are reflections prompted by the research, rather than the research findings themselves.
Goals become more complex
At school, someone may want an entry point to a solution rather than the answer. At work, a person may want a message checked rather than sent on their behalf. At home, doing something independently may matter more than shortening a task.
A general impression that someone is struggling cannot distinguish these goals. Questions such as "Which part can I help with?" or "Just the tool?" make the scope clearer. Afterwards, we can observe whether the person could continue their own activity.
Goals can also change. A request for a tool may become a request to discuss assembly. One choice should not be stored as a permanent preference.
A design hypothesis for physical AI
A possible design separately retains candidate goals, the requested tool, and the permitted scope. A request to hand over a blue screwdriver must not automatically become an instruction to finish the birdhouse. Ambiguous targets can be checked through features or location.
Evaluate whether the robot took away an activity the person wanted to retain, alongside speed. Use scenarios with similar objects, changed requests, and withdrawn assistance to examine updates to both targets and authority.
Being able to infer another person's goal differs from being permitted to decide that goal for them. That boundary should not be handed entirely to a learning model.
A decision in four panels
- A birdhouse builder reaches for an inaccessible tool.
- A friend indicates the blue screwdriver to check.
- The builder selects that tool.
- After receiving it, the builder continues.
This fictional comic focuses on what the other person can continue doing, rather than what the helper completes.
A question to consider
Have you expanded a request so far that you took over something the other person valued? Briefly checking the scope of help can preserve their agency.
References
About this series
This series uses published literature to examine everyday situations. Its 7 domains and 22 capabilities are an editorial framework, not a standard academic classification. Differences in age or culture are not rankings of people's social worth or capability. The proposed AI designs do not establish humanlike emotional experience or demonstrated product performance. The comics illustrate fictional examples.
LOGIHEART's research and development vision
LOGIHEART is researching how to implement social capabilities in physical AI that interacts with people.
Explore the research and development vision at the LOGIHEART website and in the English introduction video.
