View Proposal

Proposer
Simona Frenda
Title
Trust: the twisty story of TIAGO, an LLM and their users
Goal
Collection of data and fine-tuning a VLA model for the TIAGO robot
Description
Meeting the expectations of users is relevant to guarantee the trust and the effective use of autonomous devices like robots. In our study we propose an interdisciplinary approach to examine the social interactions between a user and a robot that, as a companion, helps to solve specific tasks. The task is designed to evaluate whether trust towards robots enhances if the model that guides their decision-making process and conversation ability is aligned with/aware of users’ culture or beliefs. RQ: Does a user's trust in a robot increase as the model aligns with their cultural values/beliefs? Natural language processing literature has revealed societal biases in generative models regarding, e.g., users' gender, nationality, and age [1, 2, 3]. However, a correct alignment of the model towards the user's perspective suggests a higher engagement and accurate performance [4, 5]. In this study, we propose an interactive activity aimed at evaluating various aspects that could affect the robot’s trustworthiness. In particular, the robot will help the user to decide the ingredients of a recipe and help to collect them in the kitchen. Although various small VLA models to guide the movement of the robot have been tested, the model needs to be adapted to TIAGO's abilities. Therefore, the student is expected to collect data and fine-tune the VLA model for TIAGO. The project will be supervised by Dr Simona Frenda and Marta Romeo.
Resources
[1] Frenda, S., Casola, S., Lo, S. M., Sezerer, E., Uva, A., Basile, V., Bosco, C., Pedrani, A., Rubagotti, C., Patti, V. and Bernardi, D., (2024). MultiPICo: Multilingual perspectivist irony corpus. In Proceedings of Association for Computational Linguistics (ACL). [2] Ding, Y., Zhao, J., Jia, C., Wang, Y., Qian, Z., Chen, W. and Yue, X., 2025, May. Gender bias in large language models across multiple languages: A case study of chatgpt. In Proceedings of the 5th Workshop on Trustworthy NLP (TrustNLP 2025). [3] Narayanan Venkit, P., Gautam, S., Panchanadikar, R., Huang, T., and Wilson, S. (2023) Unmasking Nationality Bias: A Study of Human Perception of Nationalities in AI-Generated Articles. In AIES '23. [4] Rafieian, O. and Yoganarasimhan, H. 2023. AI and personalization. Artificial Intelligence in Marketing. [5] Salemi, A., Mysore, S., Bendersky, M. and Zamani, H., 2024, August. Lamp: When large language models meet personalization. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics.
Background
Url
Difficulty Level
Challenging
Ethical Approval
None
Number Of Students
0
Supervisor
Simona Frenda
Keywords
robotics, vla models
Degrees
Master of Science in Artificial Intelligence
Master of Science in Human Robot Interaction
Master of Science in Robotics
Back to List