Individual Submission Summary
Share...

Direct link:

Poster #108 - Can Preschool Children Overimitate Robot Models Comparably to Human Models?

Fri, March 22, 7:45 to 9:15am, Baltimore Convention Center, Floor: Level 1, Exhibit Hall B

Integrative Statement

Overimitation is a paradoxical phenomenon where children copy redundant, useless and silly actions faithfully, even when it comes as a cost to themselves (Lyons et al., 2011). Social accounts of overimitation suggest children replicate irrelevant actions because they wish to be like, and affiliate with, the person who models the actions (Nielsen & Blank, 2011). Alternatively, the “copy-now refine-later” (CNRL) account proposes that children use a liberal inclusion mechanism when acquiring new skills, acknowledging that they can refine the process later, but not doing so initially to optimize their likelihood of learning all information (Whiten, 2005).

In the current study we attempted to extricate the social and CNRL accounts, employing a humanoid robot which elicits significantly less affiliative desire from children. The social account of overimitation predicts that children would overimitate a human model significantly more than a robotic model, due to the drive to affiliate with a human as a cultural ingroup member and conspecific. Alternatively the CNRL account predicts that children will overimitate both models equivalently, as the focus of overimitation is centred around acquiring the sequence of actions. In order to further extricate the social and CNRL accounts, we also investigated children’s propensity to overimitate each agent in a ritual condition. Children are found to overimitate significantly more in ritual tasks than typical instrumental overimitation tasks, driven by an affiliative desire (Nielsen et al., 2018). Under the social account, children were predicted to overimitate higher in the ritual human condition than the instrumental human condition and that the inverse effect would occur for robot conditions.

In the current study 80 (40 males) four- to six-year-old (M = 5 years, SD = 0.78) children participated. Children were assigned to a ritual task or instrumental task and to a robot or human model. In the instrumental task, children were modelled a sequence of actions, both causally irrelevant and causally relevant before retrieving a toy car from a box. In the ritual task, children were modelled the same sequence of actions as in the instrumental task, however no toy car was present to retrieve, thus this task was goal demoted resulting in the ritual component (see Table 1). Overimitation was investigated on the causally irrelevant behaviours.

Children showed stronger overimitation on causally irrelevant actions following demonstration by a human over a robot (see figure 1). This indicates that children likely desired to affiliate with the human model more so than the robot model, lending evidence to the social account. However, children’s overimitation did not differ across instrumental or ritual tasks for either human or robot. The findings of this study indicate that the strength of overimitation is driven by an affiliative desire. However, children still overimitated the robot suggesting that both the social and CNRL accounts may work in tandem. Children may first copy all and refine later as a baseline approach, and then increase their overimitation of causally irrelevant actions in order to affiliate with the model over and above simply completing the task as it was done.

Authors