Teaming up with social artificial agents

Type of project: European  |  Start date: 28/09/2023  |  End date: 27/09/2025

What if one of your colleagues were an intelligent artificial agent? Although hypothetical in some sectors, recent progress in AI and robotics suggests that artificial agents may become teammates even in complex task accomplishment. However, to become teammates, machines need to be more capable of today’s chatbots or vocal assistants. They need to be able to understand not only the complexities of the contest but also of the interdependencies between tasks and teammates, while taking into consideration the human psycho-emotional state during the task execution.

The aim of the project is to identify, in an experimental economic setting, which characteristics an artificial agent needs to have to help the team achieve “optimal results” when performing a task. We use the terms artificial agents to refer to both algorithms and robots, and team to refer to a group of “two or more individuals” that:

  • share common goal(s);
  • are interdependent in their tasks;
  • interact socially.

With optimal we do not only mean the best possible execution of the task but also the level of social cohesion and integration among the team members. To this end, the project brings together an interdisciplinary group of experts working in the fields of economics, computational linguistics, and bioengineering to systematically explore:

  • which types of task are better suited to particular types of artificial agents;
  • how should the agents communicate with its human peers to increase performance;
  • how should the intelligent agents perceive and process emotions when performing the task.

The project focus on intelligent agents acting as peers serving as social catalysts e.g., by promoting team engagement and reducing hostility thanks to conversational abilities. To this end, taking advantage of human-agent interactions during the lab and field experiments, a related objective is to exploit Natural Language Processing (NLP) approaches to build an ethical Italian dictionary that will be available to entire scientific communities to analyse critical aspects of human-agents collaboration. This dictionary (extendable over time) can be used by the agent to adjust its behavior in the group.

The ultimate goal is to develop a precise type of intelligent agent, i.e. an autonomous social robot, both from a cognitive and emotional perspective. One may minimally characterize a social robot as a physically embodied agent, with some (or full) autonomy, able to engage in social interactions while displaying thoughts and feelings. We can rely on two social robots at Research Center E. Piaggio – University of Pisa that use Social Emotional Artificial Intelligence:

  • ABEL, a hyper-realistic humanoid with a young aspect and a non-specific gender
  • FACE, a robot with the aesthetics of a woman.


Funding programme:
PRIN 2022 - 2022ALBSWX

Funding body:
European Union

Grant agreement:
CUP B53D23009850006


CNR-ILC role:

Project coordinator:
Caterina Giannetti (Università di Pisa)

CNR-ILC Research Unit Chair:
Giulia Venturi

Dominique Pierina Brunato
Felice Dell'Orletta
Chiara Fazzone
Noemi Terreni