Investigating Mental Representations about Robots in Preschool Children

This paper refers to an observational research that investigates preschool children’s mental representations of robots. Our hypotheses were that: a) three to six years-old children think about robots as human-like entities, concerning to both the physical and the conceptual nature of human beings, and b) they do not understand the concept of the software programming behind the operation of a robot. The study is based on two different data collection systems: an individual-based system and a group-based one. In both cases, the investigation uses a complex and multimodal research structure that combines a drawing-based approach with a conversational-discursive methodology. This paper focuses on the individual based data collection (348 drawings and 118 interviews). Preliminary results show that the human-like representation of robots is not the only one, even though it is present in about 64%of the drawings, and that this percentage decreases after the interview about robotics. Moreover, in several cases children refer to some concepts related to programming (e.g. the presence of sensors or human beings that make robots do things).

💡 Research Summary

The paper reports an observational study that examined how preschool children (aged three to six) mentally represent robots. The authors formulated two hypotheses: (a) children will think of robots as human‑like entities, both physically and conceptually, and (b) they will lack any understanding of the software programming that drives robot behavior. To test these ideas, the researchers employed a multimodal, mixed‑methods design that combined a drawing‑based task with a conversational‑discursive interview. While the overall project included both individual‑based and group‑based data collection, this article focuses exclusively on the individual‑based component, which yielded 348 robot drawings and 118 semi‑structured interviews.

Methodology
Participants were recruited from a single Korean preschool. Each child was first asked to draw a robot on a blank sheet of paper; no instructions about style or content were given, allowing spontaneous mental imagery to emerge. After the drawing, the child took part in a one‑to‑two‑minute interview in which the researcher asked open‑ended questions such as “Can you tell me what your robot is doing?” and “Who makes the robot work?” The interview was audio‑recorded, transcribed, and later coded alongside the drawings.

The coding scheme was developed iteratively. Drawings were categorized by physical form (human‑like, animal‑like, mechanical, hybrid) and by the presence of specific features (eyes, arms, wheels, sensors, etc.). Interview transcripts were coded for conceptual attributes, especially any reference to control mechanisms (e.g., “a person tells it what to do,” “it has a button,” “it sees with a sensor”). A special “programming awareness” tag captured any mention of software‑related ideas, even if expressed in child‑like language.

Results

• Human‑like representation: 64 % of the drawings depicted robots with clearly human characteristics (faces, limbs, torso). This supports hypothesis (a) that children default to anthropomorphic images.
• Effect of interview: After the interview, the proportion of human‑like drawings dropped to roughly 55 %, indicating that the conversational prompt prompted children to reconsider or diversify their mental models.
• Programming concepts: Contrary to hypothesis (b), 27 % of the interviewees spontaneously referenced elements that can be interpreted as rudimentary programming knowledge. Examples include “the robot has a sensor that sees light,” “my dad makes the robot move,” and “there is a button that tells it to start.” These statements reveal an emerging awareness of external inputs and control agents, even if children do not use technical terminology.

Discussion
The findings suggest that while anthropomorphism remains the dominant schema for preschoolers, children are not entirely naïve about the functional underpinnings of robots. The interview appears to act as a cognitive “scaffold,” encouraging children to articulate aspects of robot operation that are not captured in their drawings alone. This aligns with constructivist perspectives that language can reveal latent knowledge not evident in static visual artifacts.

The authors acknowledge several limitations: the sample is culturally homogeneous (Korean preschoolers), which may affect the generalizability of the anthropomorphic bias; the coding relies heavily on researcher interpretation, raising concerns about subjectivity; and the short‑term interview effect cannot be definitively linked to a lasting change in mental representation. Future work is proposed to include longitudinal tracking, cross‑cultural comparisons, and experimental interventions that explicitly teach programming concepts to assess how children’s mental models evolve.

Implications
From an educational standpoint, the study argues for early robotics curricula that go beyond presenting robots as friendly, human‑shaped companions. Introducing functional elements—sensors, commands, human operators—could nurture a more nuanced technological literacy from a young age. Moreover, the multimodal methodology (drawing + interview) proves valuable for uncovering both visual and verbal dimensions of children’s cognition, offering a template for future research on emerging technologies in early childhood.

In sum, the paper provides empirical evidence that preschool children predominantly view robots through an anthropomorphic lens, yet they also possess nascent notions of control and programming that emerge when prompted verbally. These insights refine our understanding of early mental representations of technology and point toward pedagogical strategies that can scaffold more sophisticated conceptions of robotics in the formative years.