IEEE Consumer Electronics Magazine - July 2018 - 27

BACKGROUND
Users find it beneficial to personalize their services when
using electronic equipment in their day-to-day lives (e.g., content on a smart TV, music on an audio device, or games in a
video game console) [1]. Either from entertainment or from
home-office devices, users are more than ever searching for
products to consume among a large volume of content. However, the wide range of products and their specific and varied
characteristics make it difficult for a user to find the proper
device. Recommender systems [2] assist users with this
search, providing suggestions for items to purchase by taking
into consideration their unique requirements. Recommender
systems centered on collaborative filtering techniques [3] are
used to suggest highly rated items for a target user based on
the products that similar users have experienced in the past.
In high-risk product domains, where the products are very
expensive and in which users are likely to search for and buy
products for the first time, the recommender cannot establish a
meaningful profile for many of its recommendation seekers [4].
For the average customer navigating through this domain, the
task of locating a desired choice among a large set of options is
intimidating [5]. To overcome such cold-start problems, conversational recommender systems [6] are widely recognized as an
effective preference-based search and recommender technology.
Conversational recommender systems use a product's features
to help users navigate through a product space, offer alternative
product suggestions, and elicit user feedback [7].
Most of these recommender systems, however, lack the
online collaborative capacity to enable users to be aware of
and interact with other users who are simultaneously searching for a product. A CCR that integrates a conversational recommendation process in a 3-D collaborative environment is
effective in addressing this shortcoming [8]. In this framework, which consists of two main layers that provide the user
interface and the recommender systems, and a third layer that
is responsible for communicating both [9]-[11], users have
the ability to join and search for a desired product.
In this article, a model that defines users' states and their
transitions during user-recommender interaction is proposed.
This model considers individual and collaborative states that
allow a more detailed analysis of the efficiency, as gauged by the
number of recommendation cycles used to reach a desired product, and efficacy, as measured by the decision accuracy, using
the same methodology described in [12], of the CCR algorithm
with real users in a 3-D VW implementation. Accordingly, an
individual incremental critiquing (IC) [13] algorithm, which has
no transitions since the user does not collaborate with anyone, is
compared with a collaborative approach (where users' states
change over time) that enables interaction among users. This
evaluation produced adequate results concerning usability as
well as the significant improvement of efficiency and efficacy of
a collaborative framework to that of a noncollaborative one.

RELATED WORK
Analyzed in this section are the main approaches that integrate recommenders within virtual environments according to

In the cultural domain, a
recommender has been used to
help users navigate in 3-D spaces
(i.e., museums and galleries).
the following different points of view: application domains,
recommender visualization and interaction platforms, recommendation methods, and recommender collaboration capabilities. With regard to the application domain, previous studies
have focused on implementing shopping assistants, e.g., a
virtual shopping mall on the Internet [14], or recommending
virtual objects inside a virtual-reality interface [15], whereas
others have focused on recommending locations inside the
VW [16]. In the cultural domain, a recommender has been
used to help users navigate in 3-D spaces [17] (i.e., museums
and galleries). Although the application example of the CCR
framework in this article focuses on an e-commerce domain,
it is applicable to other domains.
Relative to recommender visualization and interaction platforms, very few studies have focused on exploiting 3-D/virtual-reality interfaces [18]. Bonis et al. [17] proposed a 3-D
desktop platform-widely used by other researchers [16],
[19]-based on the Second Life protocol, a massive, online
3-D VW that permits users to construct, inhabit, and interact
in their own 3-D world. OpenSimulator [28], an open-source
3-D VW platform that follows Second Life protocols, has also
been demonstrated [20]. Authors using more specific libraries
have developed 3-D virtual environments, such as Java 3D,
with virtual-reality markup language [14], [15]. The CCR
framework in this article uses the OpenSimulator platform,
which is an out-of-the-box solution.
Most prior studies of the recommendation method have
used a collaborative-filtering (CF) method for generating user
recommendations in a 3-D virtual environment [14], [16],
[20]. The CF method [3] is based on historical data and does
not necessarily imply a direct online interaction among users.
A hybrid recommendation method based on CF [3] and content-based [21] filtering (recommendations based on a user's
past preferences) has been demonstrated [15]; however, this
approach does not allow online user collaboration either. The
CCR framework in this article employs a conversational recommender based on eliciting feedback with critiquing, a
preference-based mechanism in which users provide feedback by constraining a feature's value at the feature level,
while allowing online collaboration among users. (In this
context, online collaboration refers to users interacting while
searching for content.)
To this point, most of the studies conducted regarding collaboration among users have been limited to the implementation of a CF method based on offline historical data, i.e., the
use of aggregated values of other users' profiles. Differing
from these approaches, the research discussed in this article
July 2018

IEEE ConsumEr ElECtronICs magazInE

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - July 2018