Measuring Child-Robot Relationships

Short summary of the outcomes

Home page Resources

This page provides a very brief overview of the outcomes of the workshop. Participants were divided in breakout rooms with three topics: Measurements & Theory, Design, and Ethics. During the workshop we focussed on the challenges in these fields, but we also tried to think of solutions to these challenges.

Measurements & Theory

Challenges

We need to...
  • Identify and develop good theories relevant to child-robot interaction
    • Consolidate different sources of information
  • Develop standardised, validated and reliable research measures
    • Self-report measures are easier to implement, but may be less reliable, whereas behavioural measures are more difficult to develop, but can provide more objective data
    • Appropriate for long-term interactions
  • Design research paradigms which are generalisable across domains, cultures, and contexts
    • Pilot testing
  • Account for difficulties in conducting experiments with children (e.g., ceiling effects, people pleasing, or high drop-out rates / non-compliance)
    • Design for achievable sample sizes
  • Combine and report both qualitative and quantitative research results in a way that is accessible to multiple audiencesHow to identify and develop good theories relevant to child-robot interaction

Solutions

We should...
  • Acknowledge limitations of existing theory
    • Be transparent about the extent to which theory can inform specific hypotheses (confirmatory analyses) as opposed to more exploratory experimental designs
  • Select of measures which are theory grounded and relevant to the research questions being asked (i.e., minimise collection of non-relevant personal data)
  • Increase sharing of resources (e.g., measures, data, interaction designs) across labs and institutions
    • Link to open science movement (pre-registration, open sharing of materials for reproducibility)
  • Include combinations of (validated) self-report and behavioural measures in the design of research studies
  • Involve children throughout the measurement process
  • Encourage creativity in exploring different measurement ideas
  • Collaborate across disciplines to facilitate accessibility of results being reported (both quantitative and qualitative)

Design

Challenges

We need to...
  • Understand how children's needs and preferences change as they develop. How does this impact design?
  • Either design robots that adapt to a user's developmental stage, or design robots for a specific developmental stage. Or can robots be made in a way that is generally appealing in different developmental stages?
  • Consider humanlikeness in robot design. Human-like robots can be intuitive to interact with, but may also cause too high expectations.
  • Design not-humanlike robots in a way that it is clear how to interact with them.
  • Study how design is impacted by long-term deployment? How does the robot maintain engagement, avoid repetitive interactions, support memory, and support relationship building? Does long-term mean an autonomous robot is needed?
  • Study how design is impacted by the context, e.g. lab, clinic, home.
  • Design robots to be transparent, to communicate their data usage, and to manage children's expectations.
  • Assess whether existing robot platforms fit all these requirements. Do new robots need to be designed?

Solutions

We should...
  • Consider different stages of children’s development when designing
  • Implement child-centric methods, involving their caregivers, teachers, and parents. It would allow us to know for instance which robot’s features are more interesting and attractive for children
  • Increase robot’s awareness of children’s affective and cognitive states, but also physical movements
  • Incorporate machine learning and reinforcement learning to assess children’s state in real-time so that the robot can adapt to children’s needs alongside the interaction
  • Design for transparency to increase children’s awareness of robot’s capabilities
  • Start including the robot’s failures in the design, one solution is to define different levels of failures (e.g., context, task, user, environment, the robot itself)
  • Design different kinds of general robot’s behaviours to avoid repetitions and improve the naturalness of the interaction

Ethics

Challenges

We need to...
  • Create awareness with the children on how robots work and what is real and what is fantasy
  • Remain critical to what we want and what we should research
  • Gain knowledge from the children themselves; what do they think?
  • Conduct inclusive research where every child is represented
  • Gain knowledge from an interdisciplinary team
  • Educate ourselves on how we can conduct ethical research

Solutions

We should...
  • Test children’s robot literacy before the experiment, and allow them to ask questions regarding these issues during the interaction itself
  • Be transparent about any possible deception.
  • Create a framework on what specific things we want to shield children from
  • Include children in the design aspects of our research (co-design)
  • Have a clear idea about the (sub)group of children we are researching
  • Reach out to other relevant disciplines outside the HRI field
  • Design templates of ethics proposals to help ease the process