Incorporating intelligent agents into ordinary objects: methods and measures for studying family use of digital TVs

From 2007 Scientific Report
Jump to: navigation, search

Julie Marlier, Stéphanie Buisine William Turner



Can software agents help to intelligently manage TV leisure time? The French National Network for Audiovisual and Multimedia Innovation (RIAM) is financing a project called BUIS in an attempt to answer this question. BUIS, in French, stands for Intelligent, Easy to Use Set-Top Boxes. Set-Top Boxes are devices which link televisions to external sources of signal, turning the signal into content which is displayed on TV screens.

IWEDIA is a Set-Top Box producer and the industrial partner in the BUIS project. It has been developing dedicated software solutions to help users make the most out of their digital TV environments. One example is an interface for browsing electronic TV guides, however, the browsing format is rigid and not particularly well-suited for meeting the multiple demands of family leisure time management. Incorporating intelligent software agents and appropriate visualization techniques into IWEDIA Set-Top Boxes might resolve this problem.

The ACASA team working at LIP6 (ACASA (Agents Cognitifs et Apprentissage Symbolique Automatique)) is producing the intelligent agent and visualization techniques. This team’s long experience in cognitive modelling is being used to dynamically profile user needs but, in the case of digital television, representing the underlying dynamics of viewing behaviour is a particularly difficult problem. Although TV guides are used, of equal importance are non-rational considerations such as zapping, the bandwagon effect of popular events or the simple decision to watch something for want of better. Under these conditions, empirical research is needed in order to determine the best learning algorithm for user profiling.

The social informatics team at LIMSI (Social Informatics) is responsible for looking empirically at TV viewing practices. Its goal is twofold: first, help the ACASA team to design its agent and visualization techniques and, second, help IWEDIA to evaluate the potential market value of incorporating this software into its Set-Top Box. This work is being carried out in close cooperation with the Laboratory for Product Conception and Innovation at ENSAM (The National School for Industrial Engineering in France).

Studying TV viewing practices

Social-informatics focuses upon the relationship between computing and collective activity and, for this reason, we are less concerned with the psychology and viewing behaviour of individual TV watchers than with successfully describing how families use their digital TV environments. We are working with three couples who each have two children between 5 and 10 years old, three couples with teenagers and three retired couples, a total of 18 adults ranging in age from 33 to 77 and 18 children from 5 into their teens. Our analytical framework is built along three dimensions. First, actions with the Set-Top Box are automatically logged in order to continuously up-date an archive recording how each family uses the device to select channels and programs, to record films for future viewing, to access and use the interactive services of the digital environment. These data are being collected in two phases: first without and then with the assistance of ACASA’s software agent.

Figure 1: Collecting data on the family use of digital TV environments

Each family has been given a flat screen TV, an IWEDIA Set-Top Box with an integrated storage system and a wifi bridge for connecting the Set-Top Box to a LIP6 server through an ADSL Internet outlet. Family activity with the Set-Top Box is continuously monitored; the data archive is stored on the LIP6 server and data-mined as described above. Quantitative measures are used to calculate the time taken to access programs, the number of actions involved and the number of programs recorded for future viewing. The goal is to see if these measures improve with software assistance or not.

Our second set of observations is carried out in situ. Every month we spend two hours with each family, using scenarios and interviews in order to evaluate progress in designing an intelligent Set-Top Box. When families play out the task scenarios we have developed for them, the results are video-recorded so that we can play back to designers the sequences illustrating problems encountered when using the Set-Top Box. For interviewing, we are using a closed question format which not only facilitates comparisons between families but which is also being used to experiment the use of on-line questionnaires as a better, less intrusive form of data collection than the physical presence of researchers in the home each month.

Our third and final goal is to build a platform for working simultaneously with the audio-visual tapes of our in situ scenarios, the continuously recorded log archive of user actions with the Set-Top Box and the qualitative data obtained through interviewing.

Figure 2: Integrating audio-visual, quantitative and qualitative data for describing family TV practices

First results and perspectives

We have now just completed the first phase of our project which has been dedicated to studying how families use their Set-Top Boxes without the ACASA software incorporated into them. The following table shows how we have been helping IWEDIA improve the design of this Set-Top Box:

Figure 3: First results for improving the design of Set-Top Boxes

We are pursuing our work on data integration and annotation for describing TV leisure practices but, already, this work has shown the need to look seriously at the question of privacy when continuously monitoring collective activity and when carrying out data collection in the home. We are consequently going to be paying particular attention in the coming months to improving our observation techniques in order to reduce as much as possible their intrusiveness.


Marlier, J., Buisine, S., Turner, W. (2007). Vers un outil de co-conception produits-usages, Confere’07 Colloque francophone sur les Sciences de l'Innovation, pp. 11-16

Julie Marlier, Méthodologie et outils d’analyse des usages pour la co-conception des produits et des usages dans le processus d’innovation: Application à la conception d’Agent Apprenant Intelligent et d’interface cartographique pour la télévision interactive, Master Recherche Innovation, Conception, Ingénierie, Ecole Nationale Supérieure d’Arts et Métiers - LIMSI - CNRS, Paris: Septembre 2007.

J-G Ganascia, J-J Hennin, W. A. Turner, Projet BUIS (Boîtier Utilisateur Intelligent et Simple), Paris: Evaluation ANR, Octobre 2007.