Design and Perception of a Social Robot to Promote Hand Washing among Children in a Rural Indian School Unnikrishnan R 1 , Amol Deshmukh 2 , Shanker Ramesh 1 , Sooraj K Babu 1 , Parameswari Anitha 1 and Rao R. Bhavani 1 Abstract—We introduce “Pepe”, a social robot for encour- aging proper handwashing behaviour among children. We discuss the motivation, the robot design and a pilot study conducted at a primary school located in the Western Ghats mountain ranges of Southern India with a significant presence of indigenous tribes. The study included individual & group interviews with a randomly selected sample of 45 children to gauge their perception of the Pepe robot across various dimensions including gender, animacy & technology acceptance. We also discuss some HRI implications for running user studies with rural children. I. INTRODUCTION The Global Burden of Disease study in 2016 found that in developing countries, unsafe WASH (Water, Sanitation & Hygiene) is the third largest contributor to the global burden of disease [1]. Poor hand hygiene results in nearly 300,000 deaths annually, with the majority of deaths being among children younger than 5 years old. Despite its potential, handwashing with soap is seldom practiced in low-income countries. Freeman et al. in their systematic review of hand- washing practices across the world estimated that only 19% of people across the world washed their hands with soap after coming into contact with faeces, and presents evidence from literature showing how handwashing with soap lead to a 40% reduction in the risk of diarrhoea [2]. For India, the statistics is even more dismal at a 15% handwashing prevalence rate. Considering Indias low handwashing rates and the enormous humanitarian and economic costs of the disease burden, handwashing promotion efforts in the country are especially needed. Human-Robot Interaction technologies are advancing at a rapid pace thanks to advances in robotics and artificial intelligence and are being increasingly used in a wide variety of educational [3] and social scenarios [4]. We believe social robots have great potential to serve a supportive role in the machinery of large scale health interventions as both active agents of behaviour change as well as objective assessors of various indicators of health and hygiene. However it is important to study the design and behaviour factors of the robot that can affect childrens’ perception in the context of handwashing promotion. *This work was supported by the Amrita Vishwa Vidyapeetham Univer- sity Research Seed Grant Program. 1 AMMACHI Labs, Amrita Vishwa Vidyapeetham University, Kerala, India unnikrishnan.r@ammachilabs.org 2 University of Glasgow, School of Computing Science, Glasgow, UK amol.deshmukh@glasgow.ac.uk II. RELATED WORK Handwashing interventions are most needed in resource- poor context of rural areas. Schools in particular are a good target for interventions, as Chandrashekhar et al. [5] observed that students in Nepalese rural schools have more exposure to intestinal parasite infestations than children in urban schools due to the lack of health education and sanitation infras- tructure in schools. But if implemented alone, awareness campaigns have limited impact as seen in the UNICEF’s Great Hand Wash Yatra (Journey) [6], which used various activities like games, posters and flyers as the intervention. Though the campaign increased handwashing knowledge, it had little effect on actual hand washing behaviour. Handwashing interventions in schools are an effective way of reaching children and teaching them the habit of handwashing at a young age. An intervention study in two primary schools in rural Bangladesh showed that the proportion of handwashing after toilet use among students increased from 4% to 68% after introducing nudges. Nudges included brightly colored paths were painted from toilets to the handwashing station, and footprints and handprints were painted on the path and handwashing station [7]. Biran et al. [8] with the SuperAmma initiative takes another approach by showing how emotional drivers using triggers for disgust are an effective means for improving handwashing behaviour. However, there exists little evidence of long term maintenance of hand washing behaviour follow- ing handwashing interventions [9]. Particularly of interest in the design of robots for the persuading people to adopt better handwashing behaviour is the Hawthorne effect. This effect has been quantified by Srigley et al. [10] where hand wash events were counted and compared between the presence and lack of auditors for monitoring handwashing. The auditors were found to significantly increase hand washing behaviour. In regards to HRI studies in developing countries, Desh- mukh et al. [11] pioneered the use of studying HRI tech- niques as a means to understand robot and gender perception among rural populations. The authors observed that most of the participants viewed the social robot, in this case a utility robot for transporting water, to be useful for reducing their burden of carrying water over long distances. The partici- pants perceived the gender of the robot as female in-spite of the robot having a male voice due to cultural influence. Whatever the intervention is, it needs to be attractive and engaging to its target users to determine its success.