Written by Pepi Barnard  

I presented a talk  ‘Charting the Ecosystem of Trust in Cat Royale: Or what it takes to trust a robot to play with cats’, for the Special Session “Robots on Stage: Performance Art, Education and Social Robotics for Children” at the International Conference on Social Robotics + AI 2025 at Parthenope University of Naples, Italy on 11th Sept.

The opportunity to disseminate this research with others working in similar contexts in the social robotics community, learn more about their work, see the latest robots, and learn about developments affecting social robotics was funded by the Turing AI Fellowship ‘Somabotics: Creatively Embodying AI’. The conference attracted around 2000 attendees, involved several interesting presentations and provided the opportunity to meet some robots, like Furhat and Tiago! 

Our presentation was sent to the organisers in advance for testing, and in case of emergency. We were called early to our Special Session to test the presentation from our USBs and everything seemed fine, however, when presentations began, technical gremlins came out to play!  It made for some chaotic moments, but our session’s comradery made up for it. 

Ways round the tangles were found, allowing each of us to present our work. I briefly described Cat Royale, our novel case study exploring trustworthiness of social robots in practice where a robot arm played more than 500 games with a family of three cats over the course of 12 days.  

For Cat Royale we adopted an ecosystems approach to understand trustworthiness where intelligent systems are considered as ̏socio-technical assemblages of heterogenous components including individual humans, technological artefacts, and social structures” Bernd Stahl, 2023. 

Our study used a Performance-led Research in the Wild method. This artist- and practice-led approach falls under the umbrella of Research Through Design, in which research knowledge emerges from open-ended and exploratory design practice. Adopting artistic methods provided alternative insights and highlighted the importance of considering nonhuman stakeholders. Our approach provoked new perspectives and offered opportunities to recognise distributed responsibility in AI ecosystems.  

Delivering Cat Royale involved tackling diverse aspects of trustworthiness, including extensive ethical review processes, the immediate interaction between cats and robot arm in the enclosure, creating a home away from home (adapting Artists’ Studio), Control Room and Enclosure design, animal welfare expertise, and carefully planned public engagement. 

Inspired by the ‘onion model’ from Wilson and Sharples (2015), we mapped out the social ecology of trust for Cat Royale to identify and locate the many agents involved in establishing the trustworthiness of the robot to play with the cats along with the various roles they played, and issues that had to be considered. We illustrated how trust in robots extends beyond technical design, involving orchestration, ethics, public engagement, and non-human stakeholders (e.g. cats, plants). Artists and researchers involved in Cat Royale experienced diverse factors involved in ensuring social robots are trustworthy. 

From our Cat Royale onion model, we were able to abstract a ‘generalised social ecology of trust for robots’ which was our main contribution to the social robotics community. Application of an ecosystems approach to Cat Royale reveals concerns about direct interaction with robots (such as safety, reliability, and social behaviours) from just one part of a much larger ecosystem of trust. Also important are layers concerned with the orchestration of experiences, embedding them into the workplace, connecting them to organisational culture, and ultimately carefully presenting them for consumption.  

 We offer a holistic perspective to understanding and designing trustworthy systems, not a simple ‘cook book’ of guidelines that can be readily translated to other settings. Rather, for those wanting to apply an ecosystem perspective, Cat Royale, as a concrete case study, illustrates different factors that should be considered as a starting point to map a social ecology of trust for a specific robot application.  

Trustworthiness in social robots is complex and multi-layered and non-humans need to be considered. Animals encounter robots, by design or accident, and we must ensure this is also trustworthy. 

When not conferencing, I took in many ancient beautiful sights and of course forgot to get my camera out most of the time (too busy taking it all in). It was lovely to be by the sea!