=Paper=
{{Paper
|id=Vol-2418/AIC18_paper7
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-2418/paper7.pdf
|volume=Vol-2418
|dblpUrl=https://dblp.org/rec/conf/aic/VinanziPCC18
}}
==None==
https://ceur-ws.org/Vol-2418/paper7.pdf
Would a Robot Trust You? Developmental Robotics
Model of Trust and Theory of Mind
Samuele Vinanzi∗ , Massimiliano Patacchiola∗ , Antonio Chella† and Angelo Cangelosi∗
∗ Centre for Robotics and Neural Systems, Plymouth University, Plymouth, United Kingdom
† RoboticsLab, Università degli Studi di Palermo, Palermo, Italy
Correspondence: Samuele Vinanzi samuele.vinanzi@plymouth.ac.uk
Abstract—Trust is a critical issue in human-robot interaction:
as robotic systems gain complexity, it becomes crucial for them to
be able to blend in our society by maximizing their acceptability
and reliability. Various studies have examined how trust is
attributed by people to robots, but less have investigated the
opposite scenario, where a robot is the trustor and a human is the
trustee. The ability for an agent to evaluate the trustworthiness
of its sources of information is particularly useful in joint task
situations where people and robots must collaborate to reach
shared goals. We propose an artificial cognitive architecture based
Fig. 1. Experimental setup. A Pepper robot (1) and an informant (2) face
on the developmental robotics paradigm that can estimate the each other in front of a table where a sticker can be moved between two
reliability of its human interactors for the purpose of decision positions (3).
making. This is accomplished using Theory of Mind (ToM), the
psychological ability to assign to others beliefs and intentions that
can differ from one’s owns. Our work is focused on an humanoid
of its informants. In particular, inference is computed on the
robot cognitive architecture that integrates a probabilistic ToM
and trust model supported by an episodic memory system. probability distribution of a Bayesian network’s nodes. We
We tested our architecture on an established developmental tested this architecture replicating Vanderbilt’s experiment [3],
psychological experiment, achieving the same results obtained which consists in a sticker finding game where the child, or in
by children, thus demonstrating a new method to enhance the our case the robot, must face and learn to distinguish helpers
quality of human and robot collaborations.
and trickers. Our system is able to generate a belief network
Keywords—trust, theory of mind, episodic memory, cognitive for each user and to perform decision making and belief
robotics, developmental robotics, human-robot interaction
estimation. In addition, an episodic memory module makes
Trust is a central component of social interactions between the robot able to build a personal character that depends on
both humans and robots. It can be defined as the willingness how it has been treated in the past, thus making it more or less
of a party (the trustor) to rely on the actions of another keen to trust someone it never met. The results we obtained are
party (the trustee) with the former having no control on in line with the original experiments, thus confirming that our
the latter [1]. The fundamental role of trust evaluation is architecture correctly modeled trust and ToM mechanisms in a
to ensure successful relationships, especially during shared humanoid robot. In the future, we plan to use this model in a
goal interactions where all the parties must cooperate in a wider scenario where trust estimation and intention reading
joint task to reach a common objective. The development will generate and modulate collaborative behavior between
of trust during childhood is still under debate, but one of humans and robots.
the most interesting theories is the “trust vs mistrust” stage
ACKNOWLEDGMENT
by Erikson [2], which states that the propensity to trust is
This material is based upon work supported by the Air Force Office of
proportionate to the quality of cares received during infancy. Scientific Research, Air Force Materiel Command, USAF under Award No.
A psychological trait that relates to the mastery of one’s FA9550-15-1-0025.
self trustfulness is Theory of Mind (ToM), the ability to R EFERENCES
attribute mental states to others, as for example beliefs and [1] Roger C Mayer, James H Davis, and F David Schoorman. An inte-
intentions, that can differ from one’s owns. Vanderbilt et al. grative model of organizational trust. Academy of management review,
[3] have demonstrated that children are not good at identifying 20(3):709–734, 1995.
[2] Erik H Erikson. Childhood and Society. W. W. Norton & Company,
misleading sources of information until their fifth year of age, 1993.
when their ToM fully develops. Following these psychological [3] Kimberly E Vanderbilt, David Liu, and Gail D Heyman. The development
results, we designed an artificial cognitive architecture for of distrust. Child development, 82(5):1372–1380, 2011.
[4] Massimiliano Patacchiola and Angelo Cangelosi. A developmental
a Softbank Pepper humanoid robot that uses a probabilistic bayesian model of trust in artificial cognitive systems. In Development
approach first theoretically proposed by Patacchiola et al. [4] and Learning and Epigenetic Robotics (ICDL-EpiRob), 2016 Joint IEEE
to model trust and ToM in order to estimate the reliability International Conference on, pages 117–123. IEEE, 2016.