SatIN - Satisficing Mixed Intention Sets in Non-Collaborative Dialogues

Funding: BMBF ICT 2020 – Research for Innovation
(embedded in Software Campus program)
Duration: August 2012 - July 2014 (24 months)
Project budget: 95.000 Euro

Do you remember your last dialogue? Perhaps, this dialogue was of collaborative nature with congruent intentions of dialogue partners, e.g., when solving a PC problem together. More likely, you were part of a non-collaborative dialogue characterized by congruent as well as conflicting intentions of dialogue partners, for instance in sales conversations. Non-collaborative dialogues represent the main part of the daily communication between people, but these kinds of dialogues were rarely considered in the context of dialogue planning.

Objectives of the project „SatIN – Satisficing Mixed Intention Sets in Non-collaborative Dialogues“ were:

  • Theoretical investigation of strategic processing of congruent and conflicting intentions in dialogues and analysis by means of empirical studies of dialogues in the domain of sales conversations
  • Modeling of results and evaluation by means of implementation of a prototypical dialogue system

One of the substantial results of the project was the specification and formal description of a model „Satisficing Mixed Intention Sets“ that uses a game-theoretical equilibrium approach for enabling dialogue systems to process congruent and conflicting intentions of dialogue partners in a strategic way. Thus, dialogues that are perceived as fair by all involved dialogue partners can be planned. Research results were implemented in a web-based Retailing Conversational Agent that conducts non-collaborative sales conversations with customers by means of distributed, semantic product information.

The perceived fairness of conversations with the dialogue system was validated by an empirical user study with 120 subjects. Results show a positive evaluation of the dialogue system regarding its ability to generate dialogues perceived as fair despite the occurrence of conflicting intentions.

The application of such dialogue services creates new scopes for design when realizing innovative, web-based sales and consulting services, e.g., in online shopping. In the context of consulting conversations in online shopping, the dialogue system takes the role of a customer consultant on the one hand. On the other hand, it represents an interface for corporate communications, i.e. it represents economic intentions of manufacturers or retailers during dialogue. An extension of existing ICT systems by the dialogue service would be possible with manageable effort due to the light-weight web-based implementation.

Software Campus

SatIN is partly funded by the Federal Ministry for Education and Research (BMBF). The duration of the project is set to two years (August 2012 – July 2014). With regard to contents, SatIN is connected with the Hightech strategy of the Federal Government as well as the funding program IKT 2020, more precisely the section software systems / knowledge technology (human-computer interaction). SatIN is part of the “Software Campus” a development program for master and PhD students of computer sciences and therefore embedded in the EIT ICT Labs. The head of the project is Sabine Janzen - a participant of the pilot phase of the “Software Campus”. The project is supervised by the business mentor Scheer Group GmbH and two scientific mentors Prof. Wolfgang Maass (Chair ISS) and Prof. Wolfgang Wahlster (CEO of German Research Center for Artificial Intelligence).

logo-scheergroup.png logo_iss.png dfki_logo_online.jpg

Contact:

Project management: Sabine Janzen


Related publications

Janzen, S. & Maass, W.

Towards Benevolent Sales Assistants in Retailing Scenarios, Proc. of 20th Int. Conf. on Applications of Natural Language to Information Systems (NLDB), 2015, pp. 180-193.
BibTeX

Krüger, A., Maass, W., Paradowski, D. & Janzen, S.

Reinartz, W. & Käuferle, M. (ed.)
Empfehlungssysteme und integrierte Informationsdienste zur Steigerung der Wertschöpfung im stationären Handel, Wertschöpfung im Handel, 2014, pp. 273-291.
BibTeX

Janzen, S.

Mixed Intention Sets in Dialogue Planning, 43. Jahrestagung der Gesellschaft für Informatik e.V. (GI) (Informatik 2013), Koblenz, Germany, Bonner-Köllen Verlag , 2013.
BibTeX