Anglais

     
Recruteur
Parution
Lieu de travailParis, Ile-de-France, France
Catégorie
Fonction

Description

CNRS - Portail emploi recrutement

  • Les annuaires de l’organisme
  • Travailler au CNRS


Nudges in human-machine interaction: linguistic analyses and automatic modeling of a conversational

This offer is available in the following languages:
Français - Anglais

Faites connaître cette offre !

General information

Reference : UPR3251-IOAVAS-001
Workplace : ORSAY
Date of publication : Thursday, December 06, 2018
Scientific Responsible name : Laurence Devillers
Type of Contract : PhD candidate contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 1 March 2019
Proportion of work : Full time
Remuneration : 1 768,55 € gross monthly

Description of the thesis topic

Title
Nudges in human-machine interaction: linguistic analyses and automatic modeling of a conversational emotional agent capable of nudging

Keywords
human-machine vocal interaction, affective computing, social and affective robotics, nudge, dialogue policies, ethical and evaluation issues

Profile and skills required
dialog system
human-machine interaction
affective computing
nudging

Project description The connected objects and specifically the conversational agents such as Google Home, already present in many homes, open new perspective on verbal interaction and are likely to influence the human interlocutors. They are neither regulated at the moment, nor evaluated and their operating mode remains very opaque. "Nudges", outcomes of the behavioral sciences, consist in inciting the individuals to change their behavior without forcing them and by using their cognitive biases. Often used advisedly for the health, nudges are effective tools of public policies. But the use of personalized vocal interfaces could enlarge the notion of "nudges" to many other contexts and amplify the phenomena of manipulation.

This doctoral project is an interdisciplinary project between the CNRS (NATIONAL CENTER FOR SCIENTIFIC RESEARCH) and the University Paris-South, between IT specialists, linguists, economists and jurists and aims to answer five bolts of research on the scientific issues of nudges in the human-machine interaction:

1-the definition of nudges in general and of bad nudges

2-the development of evaluation tools for nudges

3-the implementation of strategies of nudging with robots showing the impact of nudges

4-the development of an agent assessor and a longitudinal study on an existing vocal connected object (ex: Google Home)

5- the elaboration of ethical and legal recommendations for public policies, at the interface of several scientific domains and for the regulation of such objects in the society. A greater transparency of the systems would allow to improve the confidence and the understanding of such systems by the users, in particular with regards to the vulnerable subjects such as the children and the elderly people.

The PhD candidate in computer science will work in close collaboration with a PhD candidate in economy (Unity of Network search Innovation Territories and Globalization, Paris-South university). Among the priority work questions, the thesis in computer science will focus on an in depth state of the art of the literature in meaning behavioral analysis, in interactive agents, and assessor agents modeling. Classical works on nudging such as R. Thaler, Nobel Prize in Economics in 2017 (R. Thaler, on 2016, AER) and on economy of the personal data and the private life (In. Acquisti and al. 2016, JEL) will provide a base view and a starting point for the expected PhD work. As a second step, the PhD candidate will propose and run several laboratory and real-life experiments in order to define and narrow the influence of a conversational agent with a nudging verbal behavior on several groups of human subjects: children, adults, elderly people. Expected results will provide inputs for reflection on various issues concerning the regulation of connected objects: transparency, ethics-by-design and empowerment.

INFORMATIONS CLÉS
Établissement Université Paris-Saclay
École doctorale Informatique
Sciences et Technologies de l’Information et de la Communication
Unité de recherche Traitement automatique du Langage Parlé
Directeur de la thèse Laurence Devillers Co-Directeurs Ioana
Vasilescu Gilles Adda

FINANCEMENT
DATAIA

Mots clés
Agent émotionnel, agent évaluateur, modélisation, analyse linguistique

Profil et compétences recherchées M2 en informatique.
Très bonnes compétences en programmation.

Niveau de français requis: Bon
Niveau d’anglais requis: Avancé

Présentation détaillée du projet doctoral

Les objets connectés et spécifiquement les agents conversationnels comme Google Home, qui sont déjà au sein de nombreux foyers, apportent une nouvelle dimension dans l’interaction verbale, et pourraient devenir un moyen d’influence des individus. Ils ne sont pour l’instant ni régulés, ni évalués, et leur fonctionnement est opaque. Les « nudges », issues des sciences comportementales, consistent à inciter les individus à changer de comportement sans les contraindre, en utilisant leurs biais cognitifs. Souvent utilisés à bon escient pour la santé, les « nudges » sont des outils efficaces de politique publique.
Mais l’utilisation d’interfaces vocales personnalisées pourrait amplifier les phénomènes de manipulation. Une plus grande transparence des systèmes permettrait l’amélioration de la confiance et de la compréhension des systèmes par les utilisateurs, notamment par des personnes vulnérables comme les enfants et les personnes âgées.

Ce projet a fait l’objet d’une demande de financement de deux thèses dans deux disciplines différentes (informatique et économie) auprès de l’institut de convergence DATAIA de Paris-Saclay. Ce projet doctoral est un projet interdisciplinaire entre le CNRS et l’Université Paris-Sud, entre des informaticiens, linguistes, économistes et juristes pour répondre à cinq verrous de recherche sur le « nudge » dans l’interaction homme machine:
1- la définition des « nudges » et notamment des « bad nudges »
2- le développement d’outils d’évaluation des « nudges »
3- la mise en œuvre de stratégies de « nudging » avec des robots permettant de mesurer l’impact des « nudges »
4- le développement d’un agent évaluateur et la réalisation d’une étude longitudinale sur un objet connecté vocal existant (ex: Google Home) 5- l’élaboration à l’aune des différentes disciplines de préconisations éthiques et juridiques de politique publique pour la régulation de ces objets dans la société.

Le doctorant en informatique devra travailler en étroite collaboration avec un doctorant en économie (Unité de recherche Réseaux Innovation Territoires et Mondialisation, Université Paris-Sud). La thèse en informatique devra regrouper la littérature en analyse comportementale expressive et en modélisation des agents conversationnels et des agents évaluateurs. Les travaux sur le « nudging » prennent leurs sources dans les travaux de R. Thaler, prix Nobel d’économie en 2017 (R. Thaler, 2016, AER) et en économie des données personnelles et de la vie privée (A. Acquisti et al. 2016, JEL). Le doctorant mettra en place différentes expérimentations en laboratoire et en situation réelle pour étudier l’influence d’un agent conversationnel sur différents sujets: enfants, adultes ou personnes âgées. Les résultats de la thèse seront une étude des moyens de régulation des objets connectés : la régulation par la transparence, « ethics-by-design » et l’ « empowerment ».

Méthode
Définition de situation de « nudging » (« bad and good »)
Modélisation d’un agent conversationnel capable de « nudging »
Expérimentation avec des enfants et des adultes, analyses linguistiques des dialogues
Métriques sur les « bad nudge
 »
Modélisation d’un agent évaluateur

Résultats attendus
Définition et identification de « nudges » négatifs
Régulation par la surveillance

Précision sur l’encadrement
Suivi régulier des travaux de recherche avec présentation et rédaction d’articles

ENGLISH VERSION

Title
Nudges in human-machine interaction: linguistic analyses and automatic modeling of a conversational emotional agent capable of nudging

Keywords
human-machine vocal interaction, affective computing, social and affective robotics, nudge, dialogue policies, ethical and evaluation issues

Profile and skills required
dialog system
human-machine interaction
affective computing
nudging

Project description The connected objects and specifically the conversational agents such as Google Home, already present in many homes, open new perspective on verbal interaction and are likely to influence the human interlocutors. They are neither regulated at the moment, nor evaluated and their operating mode remains very opaque. "Nudges", outcomes of the behavioral sciences, consist in inciting the individuals to change their behavior without forcing them and by using their cognitive biases. Often used advisedly for the health, nudges are effective tools of public policies. But the use of personalized vocal interfaces could enlarge the notion of "nudges" to many other contexts and amplify the phenomena of manipulation.

This doctoral project is an interdisciplinary project between the CNRS (NATIONAL CENTER FOR SCIENTIFIC RESEARCH) and the University Paris-South, between IT specialists, linguists, economists and jurists and aims to answer five bolts of research on the scientific issues of nudges in the human-machine interaction:

1-the definition of nudges in general and of bad nudges

2-the development of evaluation tools for nudges

3-the implementation of strategies of nudging with robots showing the impact of nudges

4-the development of an agent assessor and a longitudinal study on an existing vocal connected object (ex: Google Home)

5- the elaboration of ethical and legal recommendations for public policies, at the interface of several scientific domains and for the regulation of such objects in the society. A greater transparency of the systems would allow to improve the confidence and the understanding of such systems by the users, in particular with regards to the vulnerable subjects such as the children and the elderly people.

The PhD candidate in computer science will work in close collaboration with a PhD candidate in economy (Unity of Network search Innovation Territories and Globalization, Paris-South university). Among the priority work questions, the thesis in computer science will focus on an in depth state of the art of the literature in meaning behavioral analysis, in interactive agents, and assessor agents modeling. Classical works on nudging such as R. Thaler, Nobel Prize in Economics in 2017 (R. Thaler, on 2016, AER) and on economy of the personal data and the private life (In. Acquisti and al. 2016, JEL) will provide a base view and a starting point for the expected PhD work. As a second step, the PhD candidate will propose and run several laboratory and real-life experiments in order to define and narrow the influence of a conversational agent with a nudging verbal behavior on several groups of human subjects: children, adults, elderly people. Expected results will provide inputs for reflection on various issues concerning the regulation of connected objects: transparency, ethics-by-design and empowerment.

References
Acquisti, A., et al.. (2017)., Nudges for Privacy and Security: Understanding and Assisting Users’ Choices Online. ACM Comput. Surv. 50, 3, Article 44 (August 2017), 41 pages. DOI: https://doi.org/10.1145/3054926.
Acquisti, A., Taylor, C., Wagman, L., (2016)., The economics of privacy, JEL, 54.2, June.
Cecere, G., Le Guel, F., Manant, M. Soulié, N. (2017). The Economics of Privacy. New Palgrave Dictionary of Economics.
Manant, M., Pajak, S., Soulié, N., (2017), "Can Social Media Lead to Labor Market Discrimination? Evidence from a field experiment", submitted manuscript, Minor Rev.
Thaler, R. H., (2016). Behavioral Economics: Past, Present, and Future. AER, 106.7, July.
Thaler, R. H.; Sunstein, Cass R. (2008)., Nudge: Improving Decisions about Health, Wealth, and Happiness. Yale University Press. ISBN 978-0-14 311526-7

Work Context

The candidate will work in the LIMSI CNRS laboratory, and namely in the «Speech Processing Group ». He/she will be supervised by Laurence Devillers (nudging modeling, robotics, interaction), Ioana Vasilescu (linguistics) and Gilles Adda (corpora, ethics aspects). The laboratory external work will consist in gathering data in the contexts defined by the thesis (primary school, institution for elderly people).
He/she will work in collaboration with the PhD candidate in economy.

Constraints and risks

NA

Additional Information

NA

We talk about it on Twitter!



Nudges in human-machine interaction: linguistic analyses and automatic modeling of a conversational (ORSAY) http://bit.ly/2KYXU0y #Emploi #OffreEmploi #Recrutement

— EmploiCNRS ( EmploiCNRS) Thursday, 06 December, 18

Web

Lors de votre postulation, veuillez vous référer à myScience.fr et indiquer la référence  JobID 10542.