Monday, April 25, 2011

Paper Reading #25: Estimating user's engagement from eye-gaze behaviors in human-agent conversations

Comments:

TBD

Reference Information:

Title: Estimating user's engagement from eye-gaze behaviors in human-agent conversations

Authors:  Yukiko I. Nakano of Seikei University, Tokyo, Japan and Ryo Ishii of NTT, Kanagawa, Japan

Presentation Venue: IUI '10 Proceedings of the 15th international conference on Intelligent user interfaces

Summary:

The goal of the research contained in this paper is to using a software to determine the level of engagement in a conversation being invested by the human participant. The motivation for this study is to create an agent that can creatively control the interactions between itself and the human. The software the researchers propose is intended to be able to learn the human individual's personal gaze profile and tailor its responses accordingly if it detects the human's attention wandering.



Discussion:

This is interesting in that it looks like the authors are trying to create an agent that would have a good chance at passing the Total Turing test. This potential advance could give the agent a more "human" quality that might induce an actual human subject to wonder if the one they are communicating with is a person or a computer. If they could incorporate this appropriately with a full engine to simulate it, there would be a good chance they could fool many people into assuming it was a person on the other end of the communication.


I'm not sure what else it would take, but this could be ultimately incorporated into androids or robots to an extent that we would have intelligent agents housed in bodies that could be instructed to do dangerous or menial work. The one danger is that should these agents achieve consciousness to the extent that they question the validity of their own existence we may find they would choose to take matters into their own hands in such a way that it would prove detrimental to our species. Care should be taken when implementing this so it is mutually beneficial for creator and created.

2 comments:

  1. I didn't really think about engagement detection as an advancement in A.I., but it so totally is! I guess that being aware of how much someone else is engaged in an activity is such a natural thing that it didn't even occur to me how big of a step forward this was for artificial intelligence.

    ReplyDelete
  2. I really liked your thoughts on how this could be adapted to A.I. I didn't think about that at all while reading this paper.

    ReplyDelete