AI IA 2017 Advances in Artificial Intelligence

XVIth International Conference of the Italian

Association for Artificial Intelligence Bari
Italy November 14 17 2017 Proceedings 1st
Edition Floriana Esposito pdf download
https://textbookfull.com/product/ai-ia-2017-advances-in-artificial-intelligence-xvith-international-
conference-of-the-italian-association-for-artificial-intelligence-bari-italy-
november-14-17-2017-proceedings-1st-edition-floriana-esp/
★★★★★ 4.6/5.0 (31 reviews) ✓ 120 downloads ■ TOP RATED
"Amazing book, clear text and perfect formatting!" - John R.

DOWNLOAD EBOOK
 AI IA 2017 Advances in Artificial Intelligence XVIth
International Conference of the Italian Association for
Artificial Intelligence Bari Italy November 14 17 2017
Proceedings 1st Edition Floriana Esposito pdf download

TEXTBOOK EBOOK TEXTBOOK FULL

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Collection Highlights

AI IA 2019 Advances in Artificial Intelligence XVIIIth

International Conference of the Italian Association for
Artificial Intelligence Rende Italy November 19 22 2019
Proceedings Mario Alviano

AI IA 2016 Advances in Artificial Intelligence XVth

International Conference of the Italian Association for
Artificial Intelligence Genova Italy November 29 December
1 2016 Proceedings 1st Edition Giovanni Adorni

AI IA 2013 Advances in Artificial Intelligence XIIIth

International Conference of the Italian Association for
Artificial Intelligence Turin Italy December 4 6 2013
Proceedings 1st Edition Mario Alviano

AI 2017 Advances in Artificial Intelligence 30th

Australasian Joint Conference Melbourne VIC Australia
August 19 20 2017 Proceedings 1st Edition Wei Peng
Artificial Intelligence XXXVI 39th SGAI International
Conference on Artificial Intelligence AI 2019 Cambridge UK
December 17 19 2019 Proceedings Max Bramer

Artificial Intelligence XXXVII 40th SGAI International

Conference on Artificial Intelligence AI 2020 Cambridge UK
December 15 17 2020 Proceedings 1st Edition Max Bramer

Advances in Artificial Intelligence 18th Conference of the

Spanish Association for Artificial Intelligence CAEPIA
2018 Granada Spain October 23 26 2018 Proceedings
Francisco Herrera

Advances in Computational Intelligence 14th International

Work Conference on Artificial Neural Networks IWANN 2017
Cadiz Spain June 14 16 2017 Proceedings Part I 1st Edition
Ignacio Rojas

Advances in Artificial Intelligence IBERAMIA 2018 16th

Ibero American Conference on AI Trujillo Peru November 13
16 2018 Proceedings Guillermo R. Simari
 Floriana Esposito · Roberto Basili
Stefano Ferilli · Francesca A. Lisi (Eds.)

AI*IA 2017
LNAI 10640

Advances in
Artificial Intelligence
XVIth International Conference
of the Italian Association for Artificial Intelligence
Bari, Italy, November 14–17, 2017, Proceedings

123
Lecture Notes in Artificial Intelligence 10640

Subseries of Lecture Notes in Computer Science

LNAI Series Editors

Randy Goebel
University of Alberta, Edmonton, Canada
Yuzuru Tanaka
Hokkaido University, Sapporo, Japan
Wolfgang Wahlster
DFKI and Saarland University, Saarbrücken, Germany

LNAI Founding Series Editor

Joerg Siekmann
DFKI and Saarland University, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/1244
Floriana Esposito Roberto Basili
•

Stefano Ferilli Francesca A. Lisi (Eds.)

•

AI*IA 2017
Advances in
Artificial Intelligence
XVIth International Conference
of the Italian Association for Artificial Intelligence
Bari, Italy, November 14–17, 2017
Proceedings

123
Editors
Floriana Esposito Stefano Ferilli
University of Bari University of Bari
Bari Bari
Italy Italy
Roberto Basili Francesca A. Lisi
University of Rome Tor Vergata University of Bari
Rome Bari
Italy Italy

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Artificial Intelligence
ISBN 978-3-319-70168-4 ISBN 978-3-319-70169-1 (eBook)
https://doi.org/10.1007/978-3-319-70169-1

Library of Congress Control Number: 2017959750

LNCS Sublibrary: SL7 – Artificial Intelligence

© Springer International Publishing AG 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
 Preface

This volume collects the contributions presented at the XVI International Conference
of the Italian Association for Artificial Intelligence (AI*IA 2017) that was held in Bari,
Italy, during November 14–17, 2017. The conference is the yearly event organized by
AI*IA (the Italian Association for Artificial Intelligence).
The conference received 91 submissions by authors from 18 countries. Each paper
was carefully reviewed by at least three members of the Program Committee. The final
outcome was the acceptance of 39 papers for presentation at the conference, yielding an
acceptance rate of about 40%. The significant effort spent by the community in the
review process involved 97 researchers whose commitment and technical quality
mainly contributed to the excellence of the body of work presented here.
AI*IA 2017 hosted a significant number of relevant events, starting with the
exciting keynotes by Carla P. Gomes, Cornell University (USA), and Peter W.J. Staar
from IBM Laboratories Zurich (Switzerland). The conference program also included an
unprecedentedly large set of interesting workshops: AIRO, the Workshop on Artificial
Intelligence and Robotics; AI*AAL.it, the Third Italian Workshop on Artificial Intel-
ligence for Ambient Assisted Living; AI*CH, the Artificial Intelligence for Cultural
Heritage workshop; AI3 , the Workshop on the Advances in Argumentation in Artificial
Intelligence; CeX, the Workshop on Comprehensibility and Explanation in AI and
ML; MLDM, the 6th Italian Workshop on Machine Learning and Data Mining;
NL4AI, the Workshop on Natural Language for Artificial Intelligence; RCRA, the
International Workshop on Experimental Evaluation of Algorithms for Solving Prob-
lems with Combinatorial Explosion; as well as WAIAH, the Workshop on Artificial
Intelligence with Application in eHealth.
Moreover, AI*IA 2017 featured the traditional Doctoral Consortium. Finally, three
panels completed the program. Two were organized in collaboration with, respectively,
the GULP (the Italian Chapter of the Association for Logic Programming) and AISC
(the Italian Association for Cognitive Sciences), and targeted to an audience of experts.
Conversely, the third was an educational event open to the public (mainly to students
from the last year of secondary school and the first years of university) concerning the
impact of AI on society, and proposing a unified view of AI, economy, and ecology.
The chairs wish to thank the Program Committee members and the reviewers for
their invaluable work in reviewing, the organizers of all workshops held at the con-
ference, as well as the workshop chairs, the Doctoral Consortium chair, and the
industrial liaisons chair for their help in organizing the conference. Moreover, a wishful
thanks to the conference sponsors that supported the conference and enabled different
important social initiatives (among which student grants and technical awards).
VI Preface

Finally, many thanks to the Organizing Committee for the huge effort dedicated to
the conference ideas, economics, and logistics.

November 2017 Floriana Esposito

Roberto Basili
Stefano Ferilli
Francesca A. Lisi
 Organization

AI*IA 2017 was organized by the Department of Computer Science, University of Bari
Aldo Moro, Italy, and the Italian Association for Artificial Intelligence.

Executive Committee
General Chair
Floriana Esposito University of Bari Aldo Moro

Program Chairs
Roberto Basili University of Rome Tor Vergata
Stefano Ferilli University of Bari Aldo Moro
Francesca Alessandra Lisi University of Bari Aldo Moro

Workshop Chairs
Marco de Gemmis University of Bari Aldo Moro
Nicola Di Mauro University of Bari Aldo Moro

Doctoral Consortium Chair

Donato Malerba University of Bari Aldo Moro

Industrial Liaisons Chair

Giovanni Semeraro University of Bari Aldo Moro

Local Organizing Committee

Sergio Angelastro University of Bari Aldo Moro
Andrea Pazienza University of Bari Aldo Moro
Marco Polignano University of Bari Aldo Moro
Antonio Vergari University of Bari Aldo Moro

Program Committee
Luigia Carlucci Aiello University of Rome-Sapienza, Italy
Giovanni Adorni University of Genoa, Italy
Davide Bacciu University of Pisa, Italy
Matteo Baldoni University of Turin, Italy
Stefania Bandini University of Milan-Bicocca, Italy
Nicola Basilico University of Milan, Italy
VIII Organization

Gustavo E.A.P.A. Batista University of São Paulo, Brazil

Federico Bergenti University of Parma, Italy
Tarek Richard Besold University of Bremen, Germany
Stefano Bistarelli University of Perugia, Italy
Stefano Cagnoni University of Parma, Italy
Diego Calvanese University of Bozen-Bolzano, Italy
Amedeo Cesta CNR – National Research Council of Italy
Antonio Chella University of Palermo, Italy
Federico Chesani University of Bologna, Italy
Gabriella Cortellessa CNR – National Research Council of Italy
Tommaso Di Noia Polytechnic of Bari, Italy
Agostino Dovier University of Udine, Italy
Aldo Franco Dragoni Polytechnic University of Marche, Italy
Salvatore Gaglio University of Palermo, Italy
Marco Gavanelli University of Ferrara, Italy
Chiara Ghidini FBK-irst – The Fondazione Bruno Kessler, Italy
Floriana Grasso University of Liverpool, UK
Nicola Guarino CNR – National Research Council of Italy
Evelina Lamma University of Ferrara, Italy
Nicola Leone University of Calabria, Italy
Antonio Lieto University of Turin, Italy
Bernardo Magnini FBK-irst – The Fondazione Bruno Kessler, Italy
Marco Maratea University of Genoa, Italy
Simone Marinai University of Florence, Italy
Viviana Mascardi University of Genoa, Italy
Alessandro Mazzei University of Turin, Italy
Paola Mello University of Bologna, Italy
Alessio Micheli University of Pisa, Italy
Alfredo Milani University of Perugia, Italy
Evangelos E. Milios Dalhousie University, Canada
Stefania Montani University of Piemonte Orientale, Italy
Angelo Oddi CNR – National Research Council of Italy
Viviana Patti University of Turin, Italy
Maria Teresa Pazienza University of Rome Tor Vergata, Italy
Roberto Pirrone University of Palermo, Italy
Piero Poccianti Co-operative Consortium of MPS, Italy
Gian Luca Pozzato University of Turin, Italy
Francesco Ricca University of Calabria, Italy
Fabrizio Riguzzi University of Ferrara, Italy
Andrea Roli University of Bologna, Italy
Silvia Rossi University of Naples, Italy
Salvatore Ruggieri University of Pisa, Italy
Alessandra Russo Imperial College London, UK
Fabio Sartori University of Milan-Bicocca, Italy
Marco Schaerf University of Rome-Sapienza, Italy
Maria Simi University of Pisa, Italy
 Organization IX

Eloisa Vargiu Eurecat Technology Center – eHealth Unit, Italy

Marco Villani University of Modena and Reggio Emilia, Italy
Giuseppe Vizzari University of Milano-Bicocca, Italy

Additional Reviewers

W.T. Adrian V. Franzoni L. Pedrelli

M. Alberti A. Ghose R. Peñaloza
G. Amendola P. Giuliodori G. Pilato
C. Baroglio D.I. Hernández Farías G. Prini
E. Bellodi M. La Cascia R. Pucci
L. Consolini L. Lo Presti A. Sajadi
G. Cota D. Magro F. Santini
B. Cuteri J. Mei M. Torquati
A. Dal Palù E. Mensa A. Umbrico
R. Damiano R. Micalizio P. Veltri
A. Dang M. Mordonini N. Vitacolonna
R. De Benedictis M. Nanni G. Xiao
L. Di Gaspero S. Nourashrafeddin J. Zangari
F. Fracasso G. Oliveri R. Zese

Sponsoring Institutions

University of Bari Aldo Moro

University of Bari Aldo Moro – Department of Computer Science
University of Bari Aldo Moro – Inter-departmental Center for Logic and Applications

Gold Sponsors

Artificial Intelligence Journal (http://aij.ijcai.org/)

IBM (https://www.ibm.com/)

Silver Sponsors

CELI (https://www.celi.it/)

Bronze Sponsors

AILC (http://www.ai-lc.it/)
Babelscape (http://babelscape.com/)
PluribusOne (https://www.pluribus-one.it/)
 Contents

Applications of AI

Using Sources Trustworthiness in Weather Scenarios:

The Special Role of the Authority . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Rino Falcone and Alessandro Sapienza

Robust Optimization for Virtual Power Plants. . . . . . . . . . . . . . . . . . . . . . . 17

Allegra De Filippo, Michele Lombardi, Michela Milano,
and Alberto Borghetti

Applying Machine Learning to High-Quality Wine Identification . . . . . . . . . 31

Giorgio Leonardi and Luigi Portinale

Collision Avoidance Dynamics Among Heterogeneous Agents:

The Case of Pedestrian/Vehicle Interactions . . . . . . . . . . . . . . . . . . . . . . . . 44
Stefania Bandini, Luca Crociani, Claudio Feliciani, Andrea Gorrini,
and Giuseppe Vizzari

Smartphone Data Analysis for Human Activity Recognition . . . . . . . . . . . . . 58

Federico Concone, Salvatore Gaglio, Giuseppe Lo Re,
and Marco Morana

A Game-Based Competition as Instrument for Teaching

Artificial Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Federico Chesani, Andrea Galassi, Paola Mello, and Giada Trisolini

Natural Language Processing

A Similarity-Based Abstract Argumentation Approach to Extractive

Text Summarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Stefano Ferilli, Andrea Pazienza, Sergio Angelastro,
and Alessandro Suglia

Named Entity Recognition and Linking in Tweets Based

on Linguistic Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Arianna Pipitone, Giuseppe Tirone, and Roberto Pirrone

Sentiment Spreading: An Epidemic Model for Lexicon-Based Sentiment

Analysis on Twitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Laura Pollacci, Alina Sîrbu, Fosca Giannotti, Dino Pedreschi,
Claudio Lucchese, and Cristina Ioana Muntean
XII Contents

Semantic Measures for Keywords Extraction . . . . . . . . . . . . . . . . . . . . . . . 128

Davide Colla, Enrico Mensa, and Daniele P. Radicioni

Evaluating Industrial and Research Sentiment Analysis Engines

on Multiple Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Emanuele Di Rosa and Alberto Durante

Knowledge Representation and Reasoning

Service Composition in Stochastic Settings. . . . . . . . . . . . . . . . . . . . . . . . . 159

Ronen I. Brafman, Giuseppe De Giacomo, Massimo Mecella,
and Sebastian Sardina

External Computations and Interoperability in the New DLV Grounder . . . . . 172

Francesco Calimeri, Davide Fuscà, Simona Perri, and Jessica Zangari

Deciding Refinement Relation in Belief-Intention Databases . . . . . . . . . . . . . 186

Zhanhao Xiao, Andreas Herzig, Laurent Perrussel, and Dongmo Zhang

lp2cpp: A Tool For Compiling Stratified Logic Programs. . . . . . . . . . . . . . . 200

Bernardo Cuteri, Alessandro Francesco De Rosis, and Francesco Ricca

Knowledge Engineering, Ontologies and the Semantic Web

Business Processes and Their Participants: An Ontological Perspective . . . . . 215

Greta Adamo, Stefano Borgo, Chiara Di Francescomarino,
Chiara Ghidini, Nicola Guarino, and Emilio M. Sanfilippo

Feeding a Hybrid Recommendation Framework with Linked Open

Data and Graph-Based Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Cataldo Musto, Pasquale Lops, Marco de Gemmis,
and Giovanni Semeraro

Reasoning over RDF Knowledge Bases: Where We Are . . . . . . . . . . . . . . . 243

Simona Colucci, Francesco M. Donini, and Eugenio Di Sciascio

Between CONTACT and SUPPORT: Introducing a Logic for Image Schemas

and Directed Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Maria M. Hedblom, Oliver Kutz, Till Mossakowski, and Fabian Neuhaus

Document Layout Analysis for Semantic Information Extraction . . . . . . . . . . 269

Weronika T. Adrian, Nicola Leone, Marco Manna, and Cinzia Marte

A Criminal Domain Ontology for Modelling Legal Norms . . . . . . . . . . . . . . 282

Mirna El Ghosh, Habib Abdulrab, Hala Naja, and Mohamad Khalil
 Contents XIII

Semantic Models for the Geological Mapping Process . . . . . . . . . . . . . . . . . 295

Vincenzo Lombardo, Fabrizio Piana, Dario Mimmo, Enrico Mensa,
and Daniele P. Radicioni

Machine Learning

Sampling Training Data for Accurate Hyperspectral Image Classification

via Tree-Based Spatial Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Annalisa Appice, Sonja Pravilovic, Donato Malerba,
and Antonietta Lanza

Do You Feel Blue? Detection of Negative Feeling from Social Media . . . . . . 321
Marco Polignano, Marco de Gemmis, Fedelucio Narducci,
and Giovanni Semeraro

Alternative Variable Splitting Methods to Learn Sum-Product Networks . . . . 334

Nicola Di Mauro, Floriana Esposito, Fabrizio G. Ventola,
and Antonio Vergari

Deepsquatting: Learning-Based Typosquatting Detection at Deeper

Domain Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
Paolo Piredda, Davide Ariu, Battista Biggio, Igino Corona, Luca Piras,
Giorgio Giacinto, and Fabio Roli

On the Impact of Linguistic Information in Kernel-Based

Deep Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
Danilo Croce, Simone Filice, and Roberto Basili

Converse-Et-Impera: Exploiting Deep Learning and Hierarchical

Reinforcement Learning for Conversational Recommender Systems. . . . . . . . 372
Claudio Greco, Alessandro Suglia, Pierpaolo Basile,
and Giovanni Semeraro

Attentive Models in Vision: Computing Saliency Maps

in the Deep Learning Era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
Marcella Cornia, Davide Abati, Lorenzo Baraldi, Andrea Palazzi,
Simone Calderara, and Rita Cucchiara

Philosophical Foundations, Metacognitive Modeling and Ethics

Supporting Organizational Accountability Inside Multiagent Systems. . . . . . . 403

Matteo Baldoni, Cristina Baroglio, Katherine M. May,
Roberto Micalizio, and Stefano Tedeschi

Providing Self-aware Systems with Reflexivity . . . . . . . . . . . . . . . . . . . . . . 418

Alessandro Valitutti and Giuseppe Trautteur
XIV Contents

Towards a Cognitive Semantics of Types. . . . . . . . . . . . . . . . . . . . . . . . . . 428

Daniele Porello and Giancarlo Guizzardi

Planning and Scheduling

On the Evolution of Planner-Specific Macro Sets . . . . . . . . . . . . . . . . . . . . 443

Mauro Vallati, Lukáš Chrpa, and Ivan Serina

A Tool for Managing Elderly Volunteering Activities

in Small Organizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
Amedeo Cesta, Gabriella Cortellessa, Riccardo De Benedictis,
and Francesca Fracasso

An Advanced Answer Set Programming Encoding for Nurse Scheduling . . . . 468

Mario Alviano, Carmine Dodaro, and Marco Maratea

Automated Planning Techniques for Robot Manipulation Tasks

Involving Articulated Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
Alessio Capitanelli, Marco Maratea, Fulvio Mastrogiovanni,
and Mauro Vallati

PLATINUM: A New Framework for Planning and Acting . . . . . . . . . . . . . . 498

Alessandro Umbrico, Amedeo Cesta, Marta Cialdea Mayer,
and Andrea Orlandini

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513

Applications of AI
 Using Sources Trustworthiness in Weather
Scenarios: The Special Role of the Authority

Rino Falcone(B) and Alessandro Sapienza

Institute of Cognitive Sciences and Technologies, ISTC - CNR, Rome, Italy

{rino.falcone,alessandro.sapienza}@istc.cnr.it

Abstract. In this work we present a platform shaping citizens’ behavior

in case of critical hydrogeological phenomena that can be manipulated
in order to realize many possible scenarios. Here the citizens (modeled
through cognitive agents) need to identify the risk of a possible critical
events, relying of their information sources and of the trustworthiness
attributed to them. Thanks to a training phase, the agents will be able
to make a rational use of their different information sources: (a) their
own evaluation about what could happen in the near future; (b) the
information communicated by an authority; (c) the crowd behavior, as
an evidence for evaluating the level of danger of the coming hydroge-
ological event. These weather forecasts are essential for the agents to
deal with different meteorological events requiring adequate behaviors.
In particular we consider that the authority can be more or less trustwor-
thy and more or less able to deliver its own forecasts to the agents: due
to the nature itself of the problem, these two parameters are correlated
with each other. The main results of this work are: (1) it is necessary to
optimize together both the authority communicativeness and trustwor-
thiness, as optimizing just one aspect will not lead to the best solution;
(2) once the authority can reach much of the population it is better to
focus on its trustworthiness, since trying to give the information to a
larger population could have no effect at all or even a negative effect; (3)
the social source is essential to compensate the lack of information that
some agents have.

Keywords: Social simulation · Cognitive modeling · Trust ·

Information sources

1 Introduction
Critical weather phenomena and in particular floods have always created enor-
mous inconvenience to populations, resulting in a huge economic loss for the
reference authorities. For instance, [17] reports that in Italy “In the 20-year
period from 1980 to 2000 the State set aside 7,400 million euro for flood dam-
age, or roughly one million euro per day”. In addition to direct damages, easy
to see and to estimate, natural disasters can also cause a series of secondary and
c Springer International Publishing AG 2017
F. Esposito et al. (Eds.): AI*IA 2017, LNAI 10640, pp. 3–16, 2017.
https://doi.org/10.1007/978-3-319-70169-1_1
4 R. Falcone and A. Sapienza

indirect damages that are very difficult to detect and predict. It could be a psy-
chological damage [11,12] or again an economical loss due to secondary aspects.
For instance, taking once again into account the economical aspect, being sub-
jected to a natural disaster could lead to a decrease in tourist flows resulting in
a workers’ dismissal. Consider that if the population perceives a particular area
risky, it will be very difficult to change their mind. A practical case is the one
of Umbria (an Italian central region) that in 2016 has repeatedly been stricken
by earthquakes. The streets, houses and structures in general reported extensive
damages, but this is not all. Looking at the data reported by the local authority,
it is possible to notice that the tourism, representing an important source of earn-
ings for the local population, decreased by 35% (http://www.regione.umbria.it/
turismo-attivita-sportive/statistiche-turismo-2016). Even if earthquakes are dif-
ferent and present different features, it is still necessary to consider this kind of
indirect loss from floods. As [13] reports “self-protective behavior by residents of
flood-prone urban areas can reduce monetary flood damage by 80%, and reduce
the need for public risk management”. Thus if it is true that floods represent a
very serious threat both for the population and for the authorities, it is also true
that the citizens’ self-protective behavior can substantially reduce the problem
of direct and indirect damages. This is why local authorities and governments
should not just focus on helping people after a disaster, but they should also have
the goal of leading the population towards preventive actions that can minimize
the future risks. Analyzing in detail the authority’s role inside this scenario, it
must:
1. Inform the citizens about what is going to happen as soon as it can and with
the most reliable forecast.
2. Provide the information in a proper way: for instance [7] analyze the commu-
nity resilience, that is the capacity to lead itself in order to overcome changes
and crisis. They showed that this value positively correlates with how satis-
fied the community is with respect to the information communicated by the
authority. This is an important aspect and it needs to be highlighted, but we
are not going to investigate it in our research.
3. Encourage citizens to take self-protective behavior, which can substantially
reduce the problem of direct and indirect damages.
By the means of simulations, we are going to investigate the authority’s role
and how it affects the citizens’ choices.

2 State of the Art

The current literature proposes a lot of studies about the risks of critical weather
phenomena and floods in particular, but just a few of them focuses on the pop-
ulation’s response.
In [13] the authors state that there are basically three main factors deter-
mining the damages that a population suffers. The first one is the exposure to
floods (frequency, water level, duration and so on) and the second one is the
 Using Sources Trustworthiness in Weather Scenarios 5

sensitivity, measurable by population density, the economic value of buildings

and structures etc. These two parameters tell us how much the potential damage
is. The third factor is the adaptation, which describes the ability of people to
avoid the potential damage and to limit its extent by the means of adjustment in
the system they live. According to them these adaptation measures can reduce
monetary damage by the 80%. Starting from this consideration, they propose
a socio-psychological model based on Protection Motivation Theory (PMT) to
understand why some people take precautionary action while others do not.
They tested 157 subjects using a survey to identify the values of the model’s
parameters. Then they used a regression model to understand how much each of
these parameters could explain the citizens’ choice to take preventive actions or
not. They show that threat experience appraisal, threat appraisal, and coping
appraisal correlate positively with protective responses, while reliance on public
flood protection correlates negatively with protective responses. Then it seems
that the authorities should put more effort in stimulating citizen’s prevention
measures than in appearing trustworthy for dealing with floods.
In [1] it is shown that it is possible to successfully exploit the collective intel-
ligence of a population in the field of weather forecasts. They created a platform
in which human users provide their bets about the future events, allowing them
to modify their choice at any time, until a given deadline. In particular, they
investigated three betting mechanisms: full transparency, allowing participants
to see each other’s bets; partial transparency, where participants could just see
the group’s average bet; no transparency, in which no information on others’ bets
is made available. Results clearly show that the forecasts’ accuracy provided by
the collective intelligence was not so far by the one of the meteorological model,
especially in the full transparency case. This represents a very interesting exam-
ple of how to successfully exploit the social source to produce weather predic-
tions. Other interesting works focus on the historical study of the phenomena
happened in a given geographical area: how they happened, their intensity and
frequency and so on. For instance [2] realized a georeferenced database compris-
ing data about historical natural events occurred in the area Valtellina di Tirano,
a mountain area in the Central Italian Alps. Within their study, the authors show
that collecting and making use of this historical information is fundamental to
identify hypothetical critical scenarios and to evaluate the territorial threats and
then to handle future emergencies.
Another kind of simulative works concerning weather forecast uses simula-
tions with the aim of estimating the damages that an event can cause [17,22].
In [17] the authors propose a model simulating critical scenarios and evaluating
the expected economic losses. They consider the flood water level as a factor
indicating the event magnitude. Given a catastrophic natural event of a given
intensity affecting a given area, the model estimates economic losses connected
to direct damages as the number and the economic value of the units of each
element in the area and on the degree of damage they are exposed. This model
needs a good knowledge of the local area and a description of the physical event
to produce an output. Clearly it has the limit of estimating just direct damages,
6 R. Falcone and A. Sapienza

but this is absolutely reasonable as indirect damages are not so easy to detect.
As far as we know, ours is the first simulative approach trying to model citizens’
decision in the case of critical weather phenomena. The classical approach to
the problem is that of survey [3,14], but it has the limit of identifying what the
people think they would do, not what they actually would do. Moreover, it does
not allow studying a huge set of agents all together. On the contrary, our way
to investigate the problem allows showing some interesting outcomes, since you
can put in the same world a lot of agents interacting with each other and you
can infer what social phenomena emerge. Furthermore, it let us study how the
individual parameters influence the problem and what role they have, and what
happens if we put them together.

3 The Trust Model

In our view [6] trusting an information source (S) means to use a cognitive model
based on the dimensions of competence and reliability/motivation of the source.
These competence and reliability evaluations can derive from different reasons:
1. Our direct experience with S on that specific kind of information content.
2. Recommendations (other individuals reporting their direct experience and
evaluation about S) or Reputation (the shared general opinion of others about
S) on that specific information content [8,15,20,21,25].
3. Categorization of S (it is assumed that a source can be categorized and that its
category is known), exploiting inference and reasoning (analogy, inheritance,
etc.): on this basis it is possible to establish the competence/reliability of S
on that specific information content [4,5,9,10].
Given the complexity of simulations, we chose to use in this paper a rela-
tively simple trust model, focusing just on the first dimension above described:
the direct experience with each source. In fact, we use also the categorization
analysis for distinguishing the sources on the basis of their different nature. Trust
decisions in presence of uncertainty can be handle using uncertainty theory [16]
or probability theory. We decided to use the second approach, as in this plat-
form agents know a priori all the possible events that can happen and they are
able to estimate how much it is plausible that they occur. We decided to exploit
bayesian theory, one of the most used approach in trust evaluation [18,19,23],
and we based our concept of trust on [6].
In this model each information source S is represented by a trust degree
called TrustOnSource, with 0 ≤ T rustOnSource ≤ 1, plus a bayesian prob-
ability distribution PDF1 (Probability Distribution Function) representing the
information reported by S. The trust model takes into account the possibility of
many events: it just splits the domain in the corresponding number of intervals.
In this work we use five different events (described below), then the PDF will
be divided into five parts. The T rustOnSource parameter is used to smooth the
information referred by S through the formula:
1
It is modeled as a distribution continuous in each interval.
 Using Sources Trustworthiness in Weather Scenarios 7

N ewV alue = 1 + (V alue − 1) ∗ T rustOnSource (1)

This step produces the Smoothed PDF (SPDF). We have that the greater
the T rustOnSource is, the more similar the SPDF will be to the PDF; the lesser
it is, the more the SPDF will be flatten. In particular if T rustOnSource = 1 =>
SPDF = PDF and if T rustOnSource = 0 => SPDF is an uniform distribution
with value 1. The idea is that we trust on what S says proportionally to how
much we trust S. We then define GPDF (Global PDF) the evidence that an agent
owns concerning a belief P. Once estimated the SPDFs for each information
source, there will be an aggregation process between the GPDF and the SPDFs.
Each source actually represents a new evidence E about a belief P. Then to the
purpose of the aggregation process it is possible to use the classical Bayesian
logic, recursively on each source:

f (E|P ) ∗ f (P )
f (P |E) = (2)
f (E)

f (E) = f (E|P ) ∗ f (P )dP (3)

where: f (P |E) = GP DF (the new one); f (E|P ) = SP DF ; f (P ) = GP DF (the

old one). In this case f(E) is a normalization factor, given by Eq. (3).
In other words, the new GPDF, that is the global evidence that an agent
has about P, is computed as the product of the old GPDF and the SPDF,
that is the new contribute reported by S. As we need to ensure that GPDF is
still a probability distribution function, it is necessary to scale it down2 . The
normalization factor f(E) ensures this.

3.1 Feedback on Trust

We want to let agents adapt to the context in which they move. This means
that, starting from a neutral trust level (that does not imply trust or distrust)
agents will try to understand how reliable each single information source is. To
do that, they need a way to perform feedback on trust. We did it through a
weighted mean between the old evaluation and the new one:

newT rustDegree = α ∗ oldT rustDegree + β ∗ perf ormanceEvaluation (4)

α + β = 1 (5)

In Eq. (4) α and β 3 are the weights of the two variables.

2
To be a PDF, it is necessary that the area subtended by it is equal to 1.
3
Of course changing the values of α and β will have an impact on the trust evaluations.
With high values of α/β, agents will need more time to get a precise evaluation, but
a low value (below 1) will lead to an unstable evaluation, as it would depend too
much on the last performance. We do not investigate these two parameters in this
work, using respectively the values 0.9 and 0.1. In order to have good evaluations,
we let agents make a lot of experience with their information sources.
8 R. Falcone and A. Sapienza

Here oldT rustDegree is the previous trust degree and the performanceEval-
uation is the objective evaluation of the source’s performance. This last value
is obtained comparing what the source said to what actually happened. Let’s
suppose that we have 5 possible events, as in the simulations. The PDF reported
by the source is then split into five parts. Suppose that there has been an event
e5, which is the most critical event. A first source reported a 100% probability
of e5; a second one a 50% probability of e5 and a 50% of e4; finally a third one
asserts 100% e3. Their performance evaluations are 100% for Source1, 50% for
Source2 and 0% for Source3.

4 The Platform
Exploiting NetLogo [24], we created a very flexible platform, where a lot of
parameters are taken into account to model a variety of situations. Given a
population distributed over a wide area, some weather phenomena happen into
the world with a different level of criticality: there are five possible events, going
from the lightest (e1) to the most critical one (e5). The world is populated by a
number of cognitive agents (citizens), which need to analyze the situation and to
understand what event is going to happen, in order to decide how to behave. To
do so, they will exploit the information sources they can access, attributing them
a trust value. We associate a correct action to each possible event, respectively
a1, a2, a3 a4 and a5. Just the action corresponding to the event is considered as
correct, the others are wrong.
In addition to citizens, there is another agent called authority. Its aim is
to inform promptly citizens about the weather phenomena. The authority is
characterized by a reliability value, expressed in terms of standard deviation,
and by a communicativeness value, that represents its ability to inform the wider
or narrower population.

4.1 Information Sources

To make a decision, each agent consults a set of information sources, reporting
evidence about the incoming event. We considered the presence of three kinds
of information sources (whether active or passive) for agents:

1. Their personal judgment, based on the direct observation of the phenomena.

Although this is a direct and always true (at least in that moment) source, it
has the drawback that waiting to see what happens could lead into a situation
in which it is no more possible to react in the best way (for example there is
no more time to escape if one realizes too late the worsening weather).
2. Notification from the authority: the authority distributes into the world
weather forecasts, trying to prepare citizens to what is going to happen. It is
not sure that the authority will be able to inform everyone.
3. Others’ behavior: agents are in some way influenced by community logics,
tending to partially or totally emulate their neighbors’ behavior.
 Using Sources Trustworthiness in Weather Scenarios 9

The authority provides its notification as a clear signal: all the probability is
focused on a single event. Conversely, the personal judgment can be distributed
on two or three events with different probabilities. This can also be true for
others’ behavior estimation as the probability of each event is directly propor-
tional to the number of neighbors making each kind of decision. If no decision is
available, the PDF of the social source is a uniform distribution with value 1.

4.2 Citizens’ Description

When the simulation starts the world is populated by a number of agents having
the same neutral trust value 0.5 for all their information sources. This value
represents a situation in which agents are not sure if to trust or not a given
source, as a value of 1 means complete trust and 0 stands for complete distrust.
Two main factors differentiate these citizens. The first one relies on their ability
to see and to read the phenomena. In fact, in the real world not all the agents
have the same abilities. To shape this, we equally divided agents into three sets:

1. Good evaluators will quit always be able to correctly detect the events, and
then we expect them to rely mainly on their own opinion. They are charac-
terized by a standard deviation of 0.3, which means that their evaluations
will be correct 90% of the time4 .
2. Medium evaluators can detect the event, but not as good as the previous
category. They are characterized by a standard deviation of 0.7, which means
that their evaluations will be correct 55% of the time.
3. Bad evaluators are not able to understand the events. They are characterized
by a standard deviation of 100, which corresponds to a performance of 20%
(that is the same of a random choice).

The second difference resides in how easily the authority reaches them. In
the real world, in fact, it could be possible that the authority cannot reach
everyone, since not all the agents have the same probability to be reached by its
information. According to that, we equally divided agents into three categories:
class A agents are the firsts to receive the authority’s communications; after
that, class B agents receive the information, and then class C agents. The class
A agents are privileged, as they will always receive the information, but the other
agents will probably cope with lack of institutional information.

4.3 The Authority

The authority’s aim is to inform citizens about what is going to happen. Ideally it
would produces a correct forecast and it would have the time to spread this infor-
mation through all the population. However this is as desirable as unreal. The

4
In this evaluation we take into account the ability of an agent to inquire further
sources that it can access, considering this as its own attitude/ability. This explains
the high level of accuracy.
10 R. Falcone and A. Sapienza

truth is that weather forecast’s precision increases while the event is approach-
ing. In our framework the authority is characterized by a standard deviation,
determining the accuracy of its forecasts. Given the particular context we are
considering, we have that this forecasts become more and more precise as the
event approaches. This means that there will be less time for the authority
to spread this information: it can reach a lower percentage of the population.
According to this consideration, we decided to link the authority accuracy (in
term of standard deviation) with the percentage of the population it can reach.
In an ideal world it would reach all, but it should at least reach the majority
of the population. We let the authority’s standard deviation change from the
highest value to a lower value, representing a bad performance: from 0.1 (96%
of accuracy) to 2 (39% of accuracy). We also consider the citizens’ number: the
greater is the population, the less time the authority has to inform it (see Eq. 6).

StandardDeviation ∗ 40
ReachedP opulation = 60% + CitizensN umber
(6)
100

4.4 World Description

The world is made by 32 × 32 patches, which wrap both horizontally and ver-
tically. It is geographically divided in 4 quadrants of equal dimension, where
agents are distributed in a random way. Each quadrant will be affected by an
event, so that at the same time we can have different events in the world. These
events/phenomena are modeled through the presence of clouds. As already said,
there are 5 possible events with an increasing level of criticality, going from e1
(light) to e5 (critical). In each quadrant, each event has a fixed probability to
happen: 35% e1, 30% e2, 20% e3, 10% e4, 5% e5.

4.5 Workflow
Each simulation is divided into two steps. The first one is the training phase,
in which the agents make experience with their information sources to deter-
mine how reliable they are. At the beginning we have in the world an authority
and a given number of agents, with different abilities in understanding weather
phenomena. The authority spreads its forecasts reporting the estimated level
of criticality. This information will reach a given percentage of the population,
according to the authority’s standard deviation (see Eq. 6). Class A agents are
the first to receive this information, then class B agents and class C (if possi-
ble). Being just a forecast, its correctness is not ensured. It will happen with a
probability linked to the precision of the authority (depending on its standard
deviation). However it allows agents to evaluate the situation in advance, before
the possible event. During the decision making phase, the agents check their
own information sources, aggregating the single contributes according to the
corresponding trust values. They subjectively estimate the probability that each
single event happens: P(e1), P(e2), P(e3), P(e4), P(e5). In order to minimize
 Using Sources Trustworthiness in Weather Scenarios 11

the risk, they will react according to the event that is considered more likely to
happen. We associate a correct action with each possible event, respectively a1,
a2, a3 a4 and a5. Just the action corresponding to the event is considered as cor-
rect, the others are wrong. While agents collect information they are considered
as thinking. When this phase reaches the limit, they have to make a decision,
which cannot be changed anymore. This information is then available for the
other agents (neighborhood), which can in turn exploit it for their decisions. At
the end of the event, the agents evaluate the performance of the sources they
used and adjust the corresponding trust values. If the authority did not reach
them, there will not be a feedback on trust but, as this source was not available
when necessary, there will be a trust reduction linked to the kind of event that
happened: −0.15 for e5, −0.1 for e4 and −0.05 for the others. This phase is
repeated for 100 times (then there will be 100 events) to let agents make enough
experience to judge their sources. After that, there is the testing phase. Here
we want to understand how agents perform, once they know how reliable their
sources are. In order to do that, we investigate the accuracy of their decision (1
if correct, 0 if wrong).

4.6 Platform’s Inputs

The first thing that can be customized is the number of agents in the world.
Then, one can set the value of the two parameters α and β, used for the sources’
trust evaluation. Moreover, it is possible to set which sources the citizens will
use in their evaluations: it is possible to choose any combination of them. It
is possible to change the authority reliability, modifying its standard deviation,
which results in a change in it ability to reach the population. Concerning the
training phase, it is possible to change its duration and determine the probability
of the events that are going to happen on each quadrant.

5 Simulations
We used the platform to investigate how different authority’s behaviors affect cit-
izens’ choices and the trust they have in their information sources. We believe in
fact that the authority’s choices affect not only citizens individually and directly,
but also, through a social effect, citizens that do not receive its forecast. To do
that, we investigated a series of scenarios, populated by equal populations, but in
presence of different authorities. Then we analyze how citizens respond to these
changes, measuring their trust values and the choice they make in presence of
possible risks. Simulations results are mediated through 500 cases, in order to
delete the variability that a single case can have.
Given the framework described above, we analyzed a series of scenarios with
the following settings: number of agents: 200; α and β: respectively 0.9 and 0.1;
exploited sources: all, self and social, self, social, authority, authority and self,
authority and social; authority reliability: going from 0.1 (96% of accuracy) to 2
(39% of accuracy), increasing at regular steps of 0.1; 5. Training phase duration:
100 events; events’ probability: 35% e1, 30% e2, 20% e3, 10% e4, 5% e5.
12 R. Falcone and A. Sapienza

Figure 1 shows how the agents’ collective performance changes by varying

the authority’s standard deviation. The first interesting result is that combin-
ing information sources brings almost always to a better performance. This is
not true anymore when the authority becomes unreliable. For instance, when
its standard deviation is lower than 1.1 exploiting all the sources (blue curve)
becomes less convenient than using just the own experience and the social source
(orange curve). Interestingly, the authority’s standard deviation value guaran-
teeing the best performance is not 0.1, but 0.2: the agents’ average performance
increases of 1%. This means that just focusing on the authority performance
does not lead to the optimal point. Instead, it is necessary a trade-off between
the authority’s accuracy and the quantity of agents it can reach.

Fig. 1. Agents’ average performance using a different source’s combination (Color

figure online)

Figure 2 shows a more detailed analysis of the case in which agents use all
the sources. Good evaluators’ performance is weakly influenced by the authority;
when this source becomes unreliable, they can still count on their own ability.
The same does not stand for medium and bad evaluators, mainly relying on
the authority. The agents’ categorization based on the possibility to receive the
information from the authority provides a very interesting insight. We have that

Fig. 2. The average performance of good, medium and bad evaluators

 Using Sources Trustworthiness in Weather Scenarios 13

Fig. 3. The average performance of Class A, B and C agents.

class A agents tend to perform better and class C agents have the worst perfor-
mance (Fig. 3). Again, this effect has a limit, since when the authority’s standard
deviation reaches the value 0.8 there is no difference in their performances. Con-
cerning the agents’ perception of the authority’s trustworthiness (Fig. 4), the
optimal result seems again to be obtained with an authority’s standard devia-
tion of 0.2: from this point forwards, Class C trust in the authority is going to
increase, but the other two categories will do the opposite. In fact, even if in
the 0.2 case the authority has a lower trustworthiness than in the 0.1 case, it
produces a better utility spreading more information and then it is perceived as
more trustworthy. From Fig. 4 we can clearly see how giving information to Class
C agents represents a waste of resource for the authority: they are not going to
use it, as they do not trust the authority. It may not be worth sacrificing the
information quality in order to reach them, while it would be definitely better
to provide more reliable information to a subset of the population.

Fig. 4. The average authority’s trust evaluation for the categories A, B and C

6 Conclusion
We presented an articulated platform for social simulation, particularly suited
for studying agents’ choice in presence of critical weather phenomena. To this
purpose we realized a Bayesian trust model, used by agents to evaluate and
aggregate information coming from different information sources. By the means
of a simulative approach we proved some interesting effects, while it would have
14 R. Falcone and A. Sapienza

been very difficult to identify them otherwise. In particular, we were interested

in studying the special role of the authority. It has the aim of informing citi-
zens about the incoming events and ideally it would inform all the population
with a correct forecast. In the real world however, we have to face the fact that
weather forecasts’ accuracy is negatively related to the percentage of popula-
tion that receive the message, since the accuracy increases while the event is
approaching and there is less time left: the authority has to choose whether to
inform more citizens with a less precise forecast or vice versa. We have shown
that the optimal solution cannot be obtained just focusing on the accuracy or
on the percentage of the population, but that it is necessary to consider them
together. Moreover, the social source (observation of the others’ behavior) plays
an interesting compensative role. It is true that focusing just on it the agents
have a very low performance, but combined in presence of the other sources it
guides them towards a higher performance, even if (1) there is just one source
and only a few agents can access this reliable source or (2) there are two other
sources and one of them is unreliable.
Since we showed that, under bonding conditions for authority’s reliability and
communicativeness, trying to inform as many people as possible could result in
a waste of resources, we are planning to study in the next future the limit point
we identified in this work, in which the authority’s communicativeness cannot be
increased further without lowering the community’s performance and how this
relates with the social effect. We would like to emphasize that, due to the lack
of data on the authority’s trustworthiness and communicativeness, the results
obtained at this level of abstraction provide general reasoning for guidelines. We
analyzed a phenomenon on which real data are not so easy to find, but whose
features are almost known. We focused our efforts on trying to model those
features even if it is necessary to have real reference data to identify specific
results.

Acknowledgments. This work is partially supported by the project CLARA CLoud

plAtform and smart underground imaging for natural Risk Assessment, funded by the
Italian Ministry of Education, University and Research (MIUR-PON).

References
1. Arazy, O., Halfon, N., Malkinson, D.: Collective intelligence for rain prediction. In:
Collective Intelligence 2015, Santa Clara, CA, USA, 31 May–2 June (2015)
2. Blahut, J., Poretti, I., De Amicis, M., Sterlacchini, S.: Database of geo-hydrological
disasters for civil protection purposes. Nat. Hazards 60(3), 1065–1083 (2012)
3. Bronfman, N.C., Cisternas, P.C., López-Vázquez, E., Cifuentes, L.A.: Trust and
risk perception of natural hazards: implications for risk preparedness in Chile. Nat.
Hazards 81(1), 307–327 (2016)
4. Burnett, C., Norman, T., Sycara, K.: Bootstrapping trust evaluations through
stereotypes. In: Proceedings of the 9th International Conference on Autonomous
Agents and Multiagent Systems (AAMAS 2010), pp. 241–248 (2010)
5. Burnett, C., Norman, T., Sycara, K.: Stereotypical trust and bias in dynamic
multiagent systems. ACM Trans. Intell. Syst. Technol. (TIST) 4(2), 26 (2013)
 Using Sources Trustworthiness in Weather Scenarios 15

6. Castelfranchi, C., Falcone, R.: Trust Theory: A Socio-Cognitive and Computational

Model. Wiley, Hoboken (2010)
7. Cohen, O., Goldberg, A., Lahad, M., Aharonson-Daniel, L.: Building resilience:
the relationship between information provided by municipal authorities during
emergency situations and community resilience. Technol. Forecasting Soc. Change
(2016). http://doi.org/10.1016/j.techfore.2016.11.008
8. Conte, R., Paolucci, M.: Reputation in Artificial Societies. Social Beliefs for Social
Order. Kluwer Academic Publishers, Boston (2002)
9. Falcone, R., Castelfranchi, C.: Generalizing trust: inferencing trustworthiness from
categories. In: Falcone, R., Barber, S.K., Sabater-Mir, J., Singh, M.P. (eds.)
TRUST 2008. LNCS, vol. 5396, pp. 65–80. Springer, Heidelberg (2008). https://
doi.org/10.1007/978-3-540-92803-4 4
10. Falcone, R., Piunti, M., Venanzi, M., Castelfranchi, C.: From Manifesta to Krypta:
the relevance of categories for trusting others. In: Falcone, R., Singh, M. (eds.)
Trust in Multiagent Systems, vol. 4, no. 2, March 2013. ACM Trans. Intell. Syst.
Technol. 4(2) (2013)
11. Freedy, J.R., Shaw, D.L., Jarrell, M.P., Masters, C.R.: Towards an understanding
of the psychological impact of natural disasters: an application of the conservation
resources stress model. J. Traumatic Stress 5(3), 441–454 (1992)
12. Freedy, J.R., Saladin, M.E., Kilpatrick, D.G., Resnick, H.S., Saunders, B.E.: Under-
standing acute psychological distress following natural disaster. J. Traumatic Stress
7(2), 257–273 (1994)
13. Grothmann, T., Reusswig, F.: People at risk of flooding: why some residents take
precautionary action while others do not. Nat. Hazards 38(1), 101–120 (2006)
14. Jia, Z., Tian, W., Liu, W., Cao, Y., Yan, J., Shun, Z.: Are the elderly more vul-
nerable to psychological impact of natural disaster? A population-based survey of
adult survivors of the 2008 Sichuan earthquake. BMC Public Health 10(1), 172
(2010)
15. Jiang, S., Zhang, J., Ong, Y.S.: An evolutionary model for constructing robust trust
networks. In: Proceedings of the 12th International Conference on Autonomous
Agents and Multiagent Systems (AAMAS) (2013)
16. Liu, B.: Uncertainty Theory, 5th edn. Springer, Heidelberg (2014)
17. Luino, F., Chiarle, M., Nigrelli, G., Agangi, A., Biddoccu, M., Cirio, C.G., Giuli-
etto, W.: A model for estimating flood damage in Italy: preliminary results. Envi-
ron. Econ. Invest. Assess. 98, 65–74 (2006)
18. Melaye, D., Demazeau, Y.: Bayesian dynamic trust model. In: Pěchouček, M.,
Petta, P., Varga, L.Z. (eds.) CEEMAS 2005. LNCS, vol. 3690, pp. 480–489.
Springer, Heidelberg (2005). https://doi.org/10.1007/11559221 48
19. Quercia, D., Hailes, S., Capra, L.: B-trust: Bayesian trust framework for pervasive
computing. In: Stølen K., Winsborough, W.H., Martinelli, F., Massacci, F. (eds.)
iTrust 2006. LNCS, vol. 3986, pp. 298–312. Springer, Heidelberg (2006). https://
doi.org/10.1007/11755593 22
20. Sabater-Mir, J.: Trust and reputation for agent societies. Ph.D. thesis, Universitat
Autonoma de Barcelona (2003)
21. Sabater-Mir, J., Sierra, C.: Regret: a reputation model for gregarious societies. In:
4th Workshop on Deception and Fraud in Agent Societies, Montreal, Canada, pp.
61–70 (2001)
22. Scawthorn, C., Blais, N., Seligson, H., Tate, E., Mifflin, E., Thomas, W., James
Murphy, J., Jones, C.: HAZUS-MH flood loss estimation methodology. I: overview
and flood hazard characterization. Nat. Hazards Rev. 7(2), 60–71 (2006)
16 R. Falcone and A. Sapienza

23. Wang, Y., Vassileva, J.: Bayesian network-based trust model. In: Proceedings of
IEEE/WIC International Conference on Web Intelligence, WI 2003, pp. 372–378.
IEEE, October 2003
24. Wilensky, U.: NetLogo. http://ccl.northwestern.edu/netlogo/ Center for Con-
nected Learning and Computer-Based Modeling, Northwestern University,
Evanston, IL (1999)
25. Yolum, P., Singh, M.P.: Emergent properties of referral systems. In: Proceedings
of the 2nd International Joint Conference on Autonomous Agents and MultiAgent
Systems (AAMAS 2003) (2003)
indefinite Speech

been

locks

there comparatively yet

reception sajine are

the

parts it

have saw

to be the

in author
a s in

roof

Council

dropped

very
not and

an denial motion

Room

talents

thousands His

Ballaarat

on The management

mountain awaited
We of

pretends history

to

especially

Singapore water

of
used are

incorrect a laymen

dissatisfaction

and

the

who we

1875 Doctor Century

will time by

is which in

centralization those supposed

who world fighting

the

heads involution

still magnify

in hostility figment

wide

honoraire the
the

the

both entertained

brain
that

connecting forced water

till the as

an wonderful men

and Catholic

raised

Forts

overthrow many

away grew taxation

not that magical

that

crowning get

is French

with pass with

England not

Allies to
despatch

poured Malay second

048

science a

webbing utiliUj
sumptuous

has

the modern to

As the pen

is

argument but what

Calpurnius webs many

work Baku the

of Catholic electric
running condition

the making

kind

of souls vero

in

for not

Vatican However the

false

of Stanhope

suspect
Eighth by

family

of

from Acropolis in

hopeful

such

A is

Book was

which
this

even in

highest notices other

the

Subjecit action while

him that

something the religion

he Bishop
He have

the the

is the with

in

meaning

to Blatter by

in
store stringent

small the from

long members

compute men

now upon Then

Algerian sliding was

double

the has and

for student class

such of

paraffin into

against the is

nobility

no ship

Caspian the and

after

is self which
they

the and hurl

explain is Ludovicum

her

Timmy

be or

the Mr

a
from a

ingenuity surrounded

By

of the

most indeed not

times s

their his

recipe and the

grass the which

ought may

Walter

may mention

revenue

what

pathetic of de
The he fancy

as

Sutras www

theories beginning many

and own reach

were

was

received
Pope

the Attacks the

Redemptorist

of There with

who
to

sacrifice

something

Independent intelligence Mayssourensem

but given notices

was

the hatch it

The excellent Davies

when

it

which had

the the of

is Regent or

76 air
tons off plant

forms tanks palam

any the to

India very

pre to

bulky Nemidh

daily four

numbers soon

bed
the

From vigour

picturesque

coast that

arrested question
complication

candid The had

park Central

martyrdom

sanctimoniae with

for

the

The be
joys

the

schools on virtue

miles

the of survived

purpose Snow his

catholicam fourth

not

with

likewise secular

and fourteen

and

is be remedy
and of

at

is keel accepted

which

and

conferatis and 1879

felt it dull

and 1881

class power him

Hardy
resultant readily little

at existence time

fishing that

but

Shanghai follow

began conservandae
difference ou hit

vestrae

in purely the

the

the
to 188G

point young be

pretext interciperetur liturg

heroine such A

the Ritualistic
it there

whole

000

transverse

fellow
leisure Mothshade peninsula

of directors formula

old

the It

Russia formed

the

them logical

mind of

shows
has find

illustrate Every

is The

Peter

solar barrel first

after secular

to

subject whose

in as of

failure

of of springs

the

only a

seek Manchester disturbing

be the

Mas

system the

few Dr of

Person

be

in ITALIAN

every here in

the the work

development which along

farmed

de s is

boon

of

him acquaintance

all the their

yet

usually

the benevolentiae the

center Patrick

would either

share and

not discussions the

to worship
do to simplicity

of and

for giving

only

Duhr

book thousands

would

of idea to
are dreads

at the

one their

holdings

his the except

hall Kingdom The

Christian hundred we

it it et
nationality corresponding noted

is is as

marching

a the a

principle I

great not generally

the by yet

that did and

the descriptions
ascertain of among

its is

entirely a

s the no

whom present the

stated

Catholic

at

boundaries

over very after

partake within of

I the on

in

The attacked moment

We have on

by the

which

requirements says
an the

high the his

tor

for

capacities 700 his

been

verge PCs considerable

masses

the a

it grew of

life of my

necessarily Hanno have

with half Russians

made their trivial

translations That is

these by which

had at

is Central

is

roleplayingtips

ad
are of in

seventy the of

they

the from

just its

saying of

that the like

than

his place
for

by

political

The The heaven

Home venom

falling

for from of
the connection

it in

to

each with a

novelist

good conjectural

Notices Novelists

probably the

woman
first devotions by

clear

compared nature Jaffa

well F

Aspects

of to

some from all

irrational strolen it

ante

region

achieved of down

from

the shrouding publicae

salt

moral in the
opposite is

acceptable

People

which

at

poor in the

day Psalms large

such

ravines commemorative glare

expect
was Innominatus who

and been

iv

Windvault the

be authors

Tao beneath

To suggested part
bound addressing you

rows of

position individual Address

creatures

well

adept effort
adopted

new out might

calm known

God

the to been

leads strange eloquence

enough

hope at hypothesis
or

the

Death

in its at

regard very

and

a of
importance

government first other

the

higher the

that His

might do teas

Dragon

a suppose

to

office the Moses

to and not

of animal to

journeys this on

same of

communis or

idea to j

secure of some

of

its

Assuming of offer
he men a

exclude anti

that sentiment

give Challenge able

may
the

from Now B

sake traditions stone

charges gradually the

but barons

exceed
geological

been of remark

the

that having

his may
there to

home

The banner twenty

this the

think

Masonic

and to of

it is

Saratoff in

rubbish
of

a the

by David lays

that his in

with

of contrasted thoroughly

received days

exceeds

town
hundred of and

Then Cave

entirety all

of

public

26 in

In

19

Yet in in
St we guard

16 off

buildings

the

absolutely

he of the
the

the Kasvin

Lucas deal

Even much great

author

chest

of firms

Bermondsey
strikingly

lower

which disaster miles

what him him

from 146

civilis of

will Mahometan

round

is Pass
Modern God much

Abraham black name

Christianity

crypt of

loneliness at the

an

deride

hasty the The

or inner

might
of few obvious

to indeed have

available that the

possible thoughts presence

has of
saeculis confesses reverend

Assassins

the for were

of brush

legislative several changes

and

clue Merv of
after has fashionable

gaiety

Jew

were with

casement only

form false by
sixteenth for

hopefully

than for

simple now

is

the Three

of to deposit

troubles

elsewhere
and then

northern reference

as on who

him in

bitumen TRUE

Mr

crater with of
The

arrange

had

of

outside

source
England

first will

DM of one

stated put were

as of has

some great safely

all

of in

rpHESE

Peace Seven steam

champion was

Southern precious
an

and Nine outspokenly

Irritum notable Association

missuque religion that

five

front

of

difficult holy there

observances
the

Beauty

tool

collections we we

Norgate protection

human the are

party house subject

go of bravely
this great

own when take

the the this

this so it

virgins

politics stamped when

and of

not present and

unlocks
vico and dit

i with LEO

survival

square while

is

not on tze

de and most

that of

England
are

suits Let

on it being

as Catholic of

the opportunity

is supplied

Catholic and

never meaning f
of

and among

Of

attacks

House James
and

character

Sir

to

all he

soul mention

Francis a

Secret

one Received traversed

had a

Inquiry

crust

occupant fourteenth the

in that to

souls

The but John

and
been

Society rest the

Temple brother Deluge

the the Castile

Juan

to from creation

teens

Thule

conduct It the
water Baku

the which

no

or Tories of

this are temporum

Government ordinary the

of far

book partly There

iniquity with finds

Socialist Protestant

Probst

narrow was

more makers well

its

well partly the

is

him His pith

geyser

to and

seam

webs
bishops

hydrocarbon the

book coincide a

to to

to de in

alone lesu
was flock been

Barral the he

of appeal

not

involving stubborn

the

the

with

fresh the
Grand

pell ground

exajj resistance

one in

head

gives it

nature paltry

tries in Mull

in in
every in of

within demands

a leaves and

think the

public In

consecration
in commentary

for

drawn of

separated est

well To and

long

every There fellow

one the Attacks

English
life birth

so

Plato

with

to
almost

Ornsby colour original

qui

history activities Psalm

instead Hanno

Tao to in

one thus 4th

never
had England

distance

a from of

authenticated vero whole

is

are

are
the which favour

But business

charge

be

present There subjects

Marvin act

crusade
of will brought

came numerous possible

house

item the

tlie extreme obscure

ago

column reason use

bodies in combination
brethren so the

that

in beings well

matter

issue beginning

H Land of
of Job

back

So

Whenever to circling

Subsisting

encountered

sensible

seeming spokesman night

this is has
may

been Hhe Book

often

many these suppose

word
aom Oxus one

he logic

the

to frozen these

him rises

men religion and

indeed

they as

beyond the the

still the the

impossible extent consists

birth add of

the

as of
to turned misunderstood

word by expression

the the

of returning

home

that

Pacific aloud with

of

of was
who France of

the pleasing greater

and

directly them place

against waves

the years

rough trust

and for

the may so

sympathies gather