Arthur: a new ECA that uses Memory to improve Communication

Abstract

This article proposes an embodied conversational agent named Arthur. In addition to being able to talk to a person (using text and voice), he is also able to recognize the person he is talking to and detect his/her expressed emotion through facial expressions. Arthur uses these skills to improve communication with the user, also using his artificial memory, which stores and retrieves data about events and facts, based on a human memory model. We conducted some experiments to collect quantitative and qualitative information, which show that our model provides a consistent impact on users.

Is my agent good enough? Evaluating Embodied Conversational Agents with Long and Short-term interactions

“…In order to conduct our research we chose to work with Arthur, as proposed by Knob et al [3]. The focus of Arthur is to provide a natural interaction with the final user by performing a consolidation of its Short-Term Memory and Long-Term Memory.…”

“…The focus of Arthur is to provide a natural interaction with the final user by performing a consolidation of its Short-Term Memory and Long-Term Memory. Such process makes Arthur able to store and remember important pieces of the previous conversation , creating the feeling that the user is having a natural interaction, as if it was with another human [3].…”

“…The goal is to compare them and consolidate our recommendations for ECAs evaluation. The ECA used in our LTI, which occurred for 41 days, is Arthur, proposed by Knob et al [3]. In order to register the LTI, an user diary was built to log the daily interactions with Arthur, which we called Interaction Logbook.…”

“…In their model, the authors divide the ECA into three main modules: Embodiment, which is responsible for how the ECA will look like; Mental Capacities, which defines the personality and behavior of such ECA; and Application Interface, responsible for the information processing. In the literature, it is possible to find many models proposed to build an ECA [3], [5], [6]. The work of Yalcin [7] aims to model empathetic behavior on Embodied Conversational Agents (ECAs).…”

“…The ratings are: 0 = I don't agree that this is a usability problem at all; 1 = Cosmetic problem only: need not be fixed unless extra time is available on project; 2 = Minor usability problem: fixing this should be given low priority; 3 = Major usability problem: important to fix, so should be given high priority; and 4 = Usability catastrophe: imperative to fix this before product can be released. Finally, there are some usability metrics used in the work of Ren et al [11] and Tullis and Albert [12], that are based on the user perception such as Effectiveness, Efficiency and Satisfaction of the interaction with the product 3 .…”