Children and older adults form different mental models of what a robot is and can do

What it is

Children and older adults form mental models of a robot that are systematically different from a typical adult user's, and a single set of assumptions about how the user will understand this robot cannot be reused across both groups. The differences show up in whether the robot is judged to be alive, whether it has feelings, what it is actually capable of, whether it remembers or might leak private information, and whether the user believes they can control or refuse it.

Why it happens

On the children's side, the mechanism is animism from developmental psychology: at particular developmental stages, children judge objects that move on their own, talk, and respond to them as alive or as having feelings. This is not adequately described as children simply not knowing what's real — it is a cognitive-developmental phenomenon with specific age ranges and specific evidence children rely on, typically whether the motion looks self-generated and whether the object responds appropriately to the child's own actions. This explains why children are more prone than designers expect to form strong emotional attachment or fear toward a robot: they are not pretending to believe it, their cognitive system has not yet fully separated the categories of living and non-living things.

On the older-adult side, the mechanism is more about technical experience and differing mental models than about development. A substantial share of older users lack the default assumption that a set of programmed rules is producing this behavior — when a robot produces a human-like response, they are more likely to attribute it to human-like intent rather than program logic — while unfamiliarity with this kind of interaction paradigm can simultaneously lead them to underestimate what the robot can actually do. Both biases, overattributing intent and underestimating capability, can appear in the same person at once, which is a form of anthropomorphism driven by unfamiliarity rather than by the robot's design. Appearance, language style, and repeated exposure shift both groups' judgments, but not necessarily in the same direction: children may gradually re-sort the robot back into the category of "toy" with repeated contact, while older adults may correct their initial intent attribution as familiarity grows — except that correction can easily stall partway, leaving part of the misconception in place indefinitely.

Studying it

On the children's side, researchers commonly use interview or behavioral tasks that probe whether a child judges the robot to be alive, to have feelings, or to deserve certain treatment, comparing results across age groups since animism reliably weakens with age. On the older-adult side, usability testing combined with interviews is common, examining specifically how technical experience and prior expectations shape attribution and trust — for example, asking participants to predict the robot's next action and comparing that prediction against what it actually does, then looking at the direction of the prediction error. Teach-back — asking participants to explain in their own words what the robot can feel, what it remembers, and what to do if something goes wrong — is also useful for checking whether comprehension actually took hold, rather than accepting a polite "I understand" that masks a real misconception; teach-back exposes which specific step was misunderstood in a way a direct yes/no question does not.

Where it stops holding

This describes a group-level statistical pattern, not something true of every child or every older adult. Some children clearly know a robot is a machine and still choose to engage in role-play with it; some older users are technically experienced and attribute the robot's behavior no differently than a younger user would — individual variation can easily exceed the average difference between groups. Animism also weakens as children age, so the same assumption should not be applied across the entire span from preschool to adolescence — a design aimed at 6-year-olds should not carry the same assumptions as one aimed at 14-year-olds. Variation within the older-adult population cannot be reduced to "older means less technical" either: differences in hearing, vision, and cognitive function often affect judgment more than age itself, and those differences need independent assessment tools rather than being inferred from age or reported secondhand by a family member in place of the user's own account.

Applying it

Social robots designed for children or older adults need their own usability testing for affective expression and degree of anthropomorphism, rather than reusing conclusions drawn from general adult testing. The concrete check is to run age-specific interviews or behavioral observation that specifically probes whether users are attributing more capability or mind to the robot than it actually has — asking a child "would it be sad if you ignored it," asking an older user "does it remember what you told it last time." Once a clear attribution bias is found, add clarifying feedback into the interaction or deliberately reduce the strength of the cue producing that bias, rather than letting the misconception stand.

Related

• Same group: X6.05.2 Relying on a social robot for companionship can displace rather than supplement real human contact · X6.05.3 Deploying social robots in care settings requires ethics review beyond standard usability testing
• Nearby: X6.03 Emotional expression · X5.06 Presence of vulnerable populations
• Search terms: anthropomorphism · animism · mental model · human-robot interaction