Promoting sustainable and healthy diets in the population is an important promising direction to alleviate the burden on the environment of current food production and consumption practices. Previous research has shown that simple dietary changes by individuals and households can have a great effect in improving human health and curtailing greenhouse gas emissions [4,5], and reducing demand for meat products can help spur changes in food production. This can be achieved, for example, by encouraging the consumption of more plant-based foods through a process called “nudging”, which entails performing interventions to alter people’s behaviour in a predictable way. These direct approaches may be more effective at modifying consumers' behaviour than simple information campaigns [6].

This project we will explore the effectiveness of automating the provision of nudges with AI and service robotics in food retail locations. If effective, this approach could be adopted by supermarkets and restaurants where we can expect robots will be widely deployed in the future as customer assistants.

​

This project will develop AI methodologies for achieving the following objectives:

1.                Modelling user behaviour in supermarkets, focusing particularly on their food choices and the sustainability of it.

2.                Learn from real-world interactions a decision-making policy for providing food suggestions to nudge consumer behaviour toward making more sustainable choices.

 

To this aim, the use of social robots as supermarket customer assistants can be an effective mean of intervention to shape consumers choices. Studies have reported that physically present agents (i.e., robots) have greater influence on users than non-physical agents (i.e. tablet/phone apps) with people experiences similar psychological response than in interactions with other humans [1]. In this project, a robot system will be deployed based on existing hardware and software technologies [2], and novel machine learning and reinforcement learning methodologies will be developed for exploring appropriate decision-making policies driving the robot behaviour [3].

Field studies with human participants will be conducted in a supermarket research facility at Riseholme Campus (Lincoln). Customers influence over time will be evaluated using observational studies and quantitative methods, in addition to evaluation of their perception of the robot service in terms of trust, safety and explainability. Final field deployment will be conducted in a real shop in agreement with the industry partner.

​

The student will gain multidisciplinary skills in the fields of Human-Robot Interaction, AI, Machine/Reinforcement Learning, Consumer Psychology and Food Sustainability. They will have the opportunity to author peer reviewed publications in those fields, to collaborate with national and international relevant industry partners and to conduct part of their research with the second supervisor at the University of Aberdeen.

​

[1] Del Duchetto, Francesco; Kucukyilmaz, Ayse; Iocchi, Luca; Hanheide,

Marc; "Do not make the same mistakes again and again: Learning local

recovery policies for navigation from human demonstrations”, IEEE

Robotics and Automation Letters,3,4,4084-4091,2018

[2] Del Duchetto, Francesco; Kucukyilmaz, Ayse; Hanheide, Marc; "In[1]the-Wild Failures in a Long-Term HRI Deployment”, Workshop on Robot

Execution Failures and Failure Management Strategies at IEEE ICRA

2023.

[3] Sobrín-Hidalgo, D., González-Santamarta, M. A., Angel, ´, Guerrero[1]Higueras, M., Rodríguez-Lera, F. J., & Matellán-Olivera, V. (n.d.).

Explaining Autonomy: Enhancing Human-Robot Interaction through

Explanation Generation with Large Language Models.

[4] David Porfirio, Laura Stegner, Maya Cakmak, Allison Sauppé, Aws

Albarghouthi, and Bilge Mutlu. 2023. Sketching Robot Programs On the

Fly. In Proceedings of the 2023 ACM/IEEE International Conference on

Human-Robot Interaction (HRI '23). Association for Computing

Machinery, New York, NY, USA, 584–593.