IEEE Consumer Electronics Magazine - July 2018 - 28

The 3-D collaborative space
client is an immersive environment
that enables interaction and
fosters collaboration between
multiple users.
takes into consideration human nature-in which cooperation
and collaboration among users have their roots-and allows
for online collaboration among users in a 3-D virtual space
[22]. Past social psychological studies concerning product
marketing, and more recent studies pertaining to e-commerce, have concluded that consumers need some kind of
decision aid, such as word of mouth or some other form of
opinion to assist them with a buying decision [23]. The main
advantage of the research discussed in this article is that users
are able to interact and collaborate with each other during the
recommendation process.

OVERVIEW OF THE CCR FRAMEWORK
The CCR framework integrates a conversational recommender
system with a collaborative 3-D virtual environment. The conversational recommender is based on a concept known as critiquing.
Given a recommended product, the user may, for example, perform critiques such as "a cheaper camera" or "a different manufacturer." The system then provides the user with a new product
that conforms to these critiques. This process is repeated until the
user either finds a product or abandons the search. Figure 1
shows the three-layer architecture of this framework.

3-D Collaborative Space Client
Room 1

Room m
RO 1

RO 1
...

...

RO r

Product
Critique
Elements

Features

Shopping
Elements

Collaborative
Elements

User Events

User Events (Display)

3-D Collaborative Space Server
Recommendation Actions

USER STATES IN THE RECOMMENDATION PROCESS

CCR

FIGURE 1. The conceptual architecture of the CCR.
28 IEEE Consumer Electronics Magazine

The 3-D collaborative space client is an immersive environment that enables interaction and fosters collaboration
between multiple users. This 3-D virtual space contains several scenes called rooms, which represent a store for a particular
product. Inside this virtual space, each user is represented by
an avatar. When users log into the space, they can navigate,
look for, and choose a free recommendation object (RO) to
begin a recommendation session. An RO is represented by a
3-D panel that displays the product and its features and facilitates users' interactions with the following:
▼ critique elements, which allow users to perform critiques
▼ collaborative elements, which are devoted to facilitating
collaboration (e.g., start or finish collaborative actions)
▼ shopping elements, which allow users to either start or finish
(i.e., activate, buy, or abandon) a recommendation process.
Due to the nature of the multiuser 3-D interface, users are
aware of the presence of other users and may observe the
products being recommended to their fellow "shoppers." If
users are interested in others' products, they are able to collaborate; in doing so, users transition from an individual to a
collaborative state. When a user interacts with an element, as
shown in Figure 1, an event is activated in the space client,
e.g., through a critique element displayed on a 3-D RO, and
the server layer receives the critique and sends it to the recommender layer. The CCR algorithm in the recommender
layer updates the user preferences with the critique and
selects the next recommendation from the full set of products
available in this particular layer. The recommender then
sends to the server layer the new recommendation to be displayed on the RO of the user (the space client layer) who performed the last critique. Additionally, the server layer stores
all of the users' information connected to the virtual environment. Finally, this layer uses a standard virtual module to
manage all functionalities of the 3-D VW.
The 3-D collaborative space server is implemented using
C#, Open Scripting Language, and Linden Scripting Language. The OpenSimulator VW platform, which stores an
inventory of objects for the correct information display of
both product and features in the recommendation panel, is
used as its engine. OpenSimulator controls communication
between the 3-D collaborative space client and the CCR using
a communication protocol based on messages and script modules. The recommender layer is implemented using Java and
includes the CCR algorithm, which extends the IC algorithm
to incorporate collaborative capacities allowing two user feedbacks: the traditional critiquing feedback and the collaboration among users. The IC algorithm is one of the most
well-known conversational recommenders that use critiquing
[13]. Details concerning modules and functionalities of the
server and the recommender layers of the CCR can be found
in [8] and [11].

July 2018

Users frequently switch to different states during the collaborative conversational recommendation process. These states
and their interactions are defined in a finite-state machine

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