ACODA

ACODA is a joint research effort funded in the framework of the European Eurostars programme. The consortium consists of two SMEs, Zoovu (Austria) GmbH and Es-Geht Energiesystem GmbH, and the Interactive Systems Group at the University of Duisburg-Essen.

The goal of the ACODA project is to develop innovative methods for analyzing and classifying customers’ search and decision-making behavior when interacting with online digital assistants in e-commerce and services scenarios. User behavior classification based on machine learning techniques will provide the essential basis for adapting dialog flows in digital assistants, for adjusting the type and level of detail provided on products, or for including community-based feedback in advisor dialogs. Our work also contributes to designing elements for making digital assistants more informative, user-friendly and persuasive. The Interactive Systems Group will provide contributions to the following methodological objectives:

• We combine features from overt user behavior such as question answering with implicit signals from users’ interaction behavior to feed into machine learning models.
• Classification will be session-based without requiring the availability of a long-term user profile. This means the method should be applicable in cold-start situations where no prior knowledge about the user is available.
• The method development is based on sound psychological theory about users’ cognitive styles and decision making, drawing upon established theories such as the Elaboration Likelihood Model or the Maximizer/Satisficer paradigm.
• The approach will be validated on large, real-life datasets, contributing to the empirical basis in session-based user modeling.