=Paper=
{{Paper
|id=Vol-1685/paper5
|storemode=property
|title= Research Methods for Group Recommender System
|pdfUrl=https://ceur-ws.org/Vol-1685/paper5.pdf
|volume=Vol-1685
|authors=Amra Delic,Julia Neidhardt,Thuy Ngoc Nguyen,Francesco Ricci
|dblpUrl=https://dblp.org/rec/conf/recsys/DelicNNR16
}}
== Research Methods for Group Recommender System==
https://ceur-ws.org/Vol-1685/paper5.pdf
Research Methods for Group Recommender Systems
Amra Delic Julia Neidhardt Thuy Ngoc Nguyen
E-Commerce Group E-Commerce Group Free University of
TU Wien TU Wien Bozen-Bolzano
Vienna, Austria Vienna, Austria Bolzano, Italy
amra.delic@tuwien.ac.at julia.neidhardt@tuwien.ac.at ngoc.nguyen@unibz.it
Francesco Ricci
Free University of
Bozen-Bolzano
Bolzano, Italy
fricci@unibz.it
ABSTRACT Research on group recommender systems often focuses
In this article we argue that the research on group recom- on aggregation strategies, i.e., how to combine individual
mender systems must look more carefully at group dynam- preferences, sometimes conflicting preferences, into a group
ics in decision making in order to produce technologies that profile. According to Arrow’s theorem, it is clear that an
will be truly beneficial for users. Hence, we illustrate a user optimal aggregation strategy does not exist - group recom-
study method aimed at observing and measuring the evo- mender systems studies also confirmed that there is no ulti-
lution of user preferences and actions in a tourism decision mate winner. There are only a few studies that concentrate
making task: finding a destination to visit. We discuss the on decision/negotiation support in group recommender sys-
benefits and caveats of such an observational study method tems: Travel Decision Forum [12], Trip@dvice [2], Collab-
and we present the implications that the derived data and orative Advisory Travel System (CATS) [14], Choicla [21].
findings may have on the design of interactive group recom- To our best knowledge, there are no observational studies
mender systems. on group decision processes in the context of group recom-
mender systems. These types of studies are usually con-
ducted in the social disciplines: in [22] the importance of
CCS Concepts discussions, especially with respect to information that is
•Information systems → Recommender systems; •Human- shared among group members is emphasized. An extensive
centered computing → User studies; overview of studies on group dynamics and the influence of
the different aspects (e.g., group structure, group decision
Keywords process structure) on the group choices is presented in [8].
The main motivation of this paper is therefore to raise
Group Decision Making, Group recommender systems, Ob- in the group recommender systems community the aware-
servational Study ness of the importance of a new type of analysis: observing
groups in naturalistic settings. We believe that the design
1. INTRODUCTION of a novel and more effective sort of group recommender
Recommender systems for groups are becoming more and systems can be initiated if one better observes and under-
more important since many information needs originate by stands groups in actions, measures their behaviors, and tries
group and social activities, like listening to music, watching to identify concrete opportunities for computerized systems
movies, traveling, attending sport events, and many more. to become more useful to people. In this paper we will il-
The importance of group recommender systems also has in- lustrate the design, the outcome and the implications of an
creased due to the social web, where users are not isolated observational study where groups of people faced a concrete
but form interrelated groups. A high number of papers on decision task - select a destination to visit as a group - and
group recommender systems have been published [13] but the researchers monitored the groups before, during and af-
still, we believe, there is a gap between the current main fo- ter the task.
cus of the research and the information search and decision Hence, our study is motivated by a range of dimensions
making support needs of groups. and issues, that we list in the following.
• Decision making is the ultimate motivation for a group
recommender system. This is true even more than for
individual recommenders which can also be used for
expanding user knowledge or expressing self [20]. But
if group recommenders must support decision mak-
ing we must understand how this task is executed in
Copyright held by the author(s). groups and how the decision issues, the group members
RecTour 2016 - Workshop on Recommenders in Tourism held in conjunc-
tion with the 10th ACM Conference on Recommender Systems (RecSys), and the contextual situation alltogether impact on it.
September 15, 2016, Boston, MA, USA. In the past too much attention was put on how to iden-
tify “optimal” recommendations, which in the context satisfaction, how to compare and relate user prefer-
of groups is not even possible to correctly define. ences and group preferences.
• We believe that the application domain is crucial in Thus, the aim of this paper is to reflect on research meth-
a group recommender system. Recommending tourist ods for group recommender systems on the basis of an obser-
attractions or destination for a group cannot follow the vational study.To present a detailed analysis of the collected
same model used to recommending movies to watch [24]. data is not the focus of this paper; this was done in [5].
The tourism product is more complex than other types The rest of this paper is structured as follow: in Sec-
of products (i.e., it is a bundle of products and services) tion 2 the study procedure is described in detail, Section 3
and in the same time it is less tangible. Moreover, trav- illustrates instruments used for the data collection, in Sec-
eling is an emotional experience and explicit preference tion 4 results of a first analysis are summarized, followed
characterization is problematic especially in the early by Section 5 where implications for recommender systems
phase of the travel decision-making process as different are explained. Finally, in Section 6 we discuss limitations,
users usually have different perceptions of the features challenges and possible variations of the study.
of the items. Finally, tourism products are typically
experienced in groups. For that reason, we have tried
to generate a decision task - destination selection - that
2. PROCEDURE
is believable in the context of tourism decision making In order to design a new generation of more useful and
and we made observations for users characteristics and effective group recommender systems, we do not only aim
decision outcome that have emerged as important in at gaining insights into human behavior, but also at learn-
tourism research on consumer behavior [6, 7, 23, 25]. ing how to improve and facilitate interaction of users in
a computer mediated setting. To set a basis for this, we
• Group recommendations techniques have been influ- started with an exploratory research approach that is not
enced too strongly by social choice theory [13] and not constrained by any pre-existing system functionality, i.e., we
enough by group dynamics studies [8]. It is still un- developed a study to collect observational data on human-
clear how a recommender can identify items to suggest to-human interactions in group decision making task. In
in a group decision making task, if the goal is not sim- the following we describe the procedure of this observational
ply to aggregate the votes/preferences expressed by the study in detail.
group members. But we believe that studies like the The study was initiated in a cooperation with the Interna-
presented one can help to understand the key infor- tional Federation for Information Technologies in Travel and
mation that groups need in order to make decisions, Tourism (IFITT) and 11 universities worldwide. The first
which could not simply be the suggested outcome of implementations of the study took place at the Delft Univer-
the decision. We believe that the more general concept sity of Technology (TU Delft), the University of Klagenfurt
of information recommendation, rather than product (UNI Klagenfurt) and the University of Leiden (UNI Lei-
recommendation, is important to implement [3]. den), while an extended study was carried out at the Vienna
University of Technology (TU Wien). Each implementation
• It is clear to us that the design of more effective group was conducted as a part of a regular lecture and followed
recommender systems requires a multidisciplinary ap- a three-phases structure: pre-survey questionnaire phase,
proach. In that sense the study described in this pa- groups meeting/discussion phase and post-survey question-
per brings together social science and computer science naire phase (see Figure 2).
scholars. Observational studies are not part of the Prior to the first study phase, an introduction with gen-
classical research repertoire of recommender systems eral instructions for the participants was presented. The
research methods, but, we believe that these methods first task for all participant was to form groups. At TU
are strictly required if we want to understand users in Delft, UNI Klagenfurt and UNI Leiden, students were free
naturalistic settings and be able to generate fruitful to choose their group size (between two and four group mem-
conjectures about new and useful system functions. bers). At TU Wien students were instructed to form groups
• Another important motivation of this study is the de- of six members and to select two students (referred to as
sire to collect data about group decision making that observers) whose task was to observe and record activities
can be exploited by several research groups. Hence, of their group in the next phase. All the other group mem-
in some sense, we wanted to obtain raw data that bers took part in the decision making process (referred to as
could be used to several types of analyses, from dif- decision makers).
ferent perspectives and with alternative motivations. In the first study phase, the task for the decision makers
We plan to make the data that we have collected, and was to fill in a pre-survey online questionnaire that cap-
that will also be collected in future implementations of tures their individual profiles, preferences and dislikes. De-
the study, available to everyone for further analyses. tailed data description is provided in section 3. Also, in this
phase, in Vienna, a short training for observers was orga-
• Finally, we believe that the research community on nized. The purpose was to introduce them with the follow-
group recommender systems needs to discuss and build ing study tasks and to instruct them on how to perform and
a research agenda. We must identify critical challenges record a group observation. A report template, which was
and expected results. In this study we initiate this re- constructed based on Bales’s Interaction Process Analysis
flections by raising several issues, e.g., how to measure (IPA) [1], was provided to the observers to record the ac-
the collective behavior of a group, what properties of tivities of the decision makers. The observers also received
a group are more important in recommender systems written instructions and during the rest of the study they
and how they should be measured, how to define group were in a close contact with the study organizers.
Figure 1: Overall structure of the study and differences between implementations
In the second study phase, the group meeting and dis- the behavior of the decision makers and how seriously they
cussion took place. The decision makers received written actually performed the task.
instructions with the following structure: At each university the study implementation followed the
described structure. However, still some differences existed,
1. Ten predefined destinations together with informational they are explained in section 6. After the first implementa-
Wiki pages; tion round, considering all the locations where the study was
conducted, the size of the collected data sample comprised
2. Decision task scenario: Imagine that you are work- 78 decision makers in all together 24 groups of two, three and
ing on a research paper together with the other group four group members, plus 16 observers (two for each group)
members. Interestingly, your university offers you the at TU Wien. At TU Delft, after a first implementation
opportunity to submit this paper to a conference in Eu- round (referred to as TU Delft), a second one with the same
rope. If the paper gets accepted, the university will pay configuration (without observation) took place (referred to
to each group member the trip to the conference. In as TU Delft2 ). It introduced 122 new decision makers in
addition, you will be able to spend the weekend after 31 groups. Thus, currently the data sample comprises 200
the conference at the conference destination. Ten con- decision makers in 55 groups of two, three, four and even
ferences will take place in European capitals around the five group members (see Table 1) plus 16 observers.
same summer period ;
3. Next, they were asked to discuss and decide which des- Group size 2 3 4 5
tination they would like to visit most as a group. Ad- UNI Leiden 2 2 2 /
ditionally, they also had to provide a second choice in UNI Klagenfurt 1 1 4 /
case that the first option would no longer be available. TU Delft 1 2 1 /
TU Delft2 1 8 14 8
Groups were not instructed on how to perform the deci- TU Wien 2 1 5 /
sion making task and whether they should check the infor- SUM 7 14 26 8
mational Wiki pages or not. This specific design was chosen
due to its simplicity. Usually, when a group is planning a trip
a bundle of different trip aspects have to be considered, e.g., Table 1: Groups sizes per university
timing, budget, destination, accommodation, transport, etc.
This type of task would be almost impossible to simulate in
a controlled environment. Thus, we concentrated on a sim-
ple aspect to analyze the basis of group interactions and 3. MEASUREMENTS
dynamics in this specific context. At TU Wien, observers In this section we describe the data in detail as well as
were included in the task. They audio recorded and reported the instruments were used to collect it: a pre-survey ques-
the group decision process using the previously mentioned tionnaire, a template for reporting the observations and a
report template (details in 3). post-survey questionnaire. Each of these instruments was
In the third phase, the decision makers filled in an on- designed in a way that the obtained data cover different as-
line post-survey questionnaire inquiring about the previous pects, which might impact the group decision process and
phase and the overall experience. During this phase, in- which were derived from the literature.
terviews with the observers were arranged in Vienna: for Accordingly, the first data collection instrument - a pre-
each group a meeting with the two observers of that group survey questionnaire1 captured individual profiles of the par-
took place. Firstly, we evaluated observers’ understanding ticipants in a similar way as the user profile in a recom-
of the task and the reports that they submitted, then, the
observers elaborated their reports and discussed differences 1
https://survey.aau.at/2012/index.php?sid=49577&lang=
between those. Furthermore, they were also queried about en
mender system would be represented. It is comprised of 68 were requested to identify, record and categorize each
questionnaire statements separated into four sections: “unit” of interaction (i.e., verbal and non verbal expres-
sions) according to the twelve categories of behavior;
1. Demographic data and university affiliation (i.e., age,
gender, country of origin, university and student iden- 4. Social decision scheme (i.e., delegating, averaging, vot-
tification number); ing, reaching consensus or other -explanation could be
provided);
2. 17 tourist roles and Big Five Factors:
• 30 questionnaire statements related to 17 tourist 5. Strength of group members’ preferences (i.e., for each
roles (i.e., types of touristic short term behavior) group member, the observers rated from 1 - Very un-
defined in [10]; willing to 5 - Very willing on how willing they were to
give up on their preferred options).
• 20 questionnaire statements related to the Big
Five Personality Factors (i.e., Openness to new Finally, a post-survey questionnaire2 was used to collect
experiences, Conscientiousness, Agreeableness, Ex- data about the participants’ experience with the group de-
troversion, Neuroticism) [11]. cision process and the overall study. It asked for:
3. Experience and ratings/ rankings of ten predefined 1. The first and the second group choice;
destinations:
2. Whether the provided information about the destina-
• Destinations: Amsterdam (at TU Wien and UNI tions was used during the group decision process;
Klagenfurt), Berlin, Copenhagen, Helsinki, Lis-
bon, London, Madrid, Paris, Rome, Stockholm 3. Description of the decision process that led the group
and Vienna (at TU Delft and UNI Leiden); to their final choice;
• Participants were asked how many times they have
4. Overall attractiveness of the ten predefined destina-
visited each destination;
tions (e.g., ”Many destinations were appealing.”, ”I did
• Participants at the TU Wien rated, while other not like any of the destinations.”);
participants ranked the ten destinations (implica-
tions of this distinction are discussed in section 6). 5. Satisfaction with the group choice (e.g., ”I like the des-
tination that we have chosen”);
4. Ranking of decision criteria (i.e., budget, weather, dis-
tance, social activities, sightseeing and other). 6. Difficulty of the decision process (e.g., ”Eventually I
was in doubt between some destinations.”);
A five-point likert scale was used for the 50 questionnaire
statements related to the 17 tourist roles and the Big Five 7. Participant’s perceived identification and similarity with
Factors. To obtain the scores, i.e., the level to which a person the other group members (e.g., ”I see myself as a mem-
belongs to a certain tourist role or to a certain personality ber of this group”, etc.);
trait, ratings of the statements were summed and divided
by the number of related questionnaire statements. Tourist 8. Assessment of the task (i.e., participants were asked to
roles and personality traits are related to the user model of select the statements to which they agree regarding the
the picture-based recommendation engine (see section 5). organization of the task, their feedback and willingness
In the second phase group decision task took place. By to participate in the same or similar study).
now, only at the TU Wien, observational part of the study
was implemented. The report template for the observers’ A five-point likert scale was used to assess 4., 5., 6. and 7.
recordings was designed based on the Bales’s Interaction The overall structure of the data is shown in Figure 2.
Process Analysis (IPA) (i.e., a method to study small groups It visualizes the data as an Entity Relationship Diagram
and interactions among group members) [8]. Thus, the task (ERD). Different colors indicate different study phases, i.e.,
for observers was to audio record group discussion and to pink: pre-survey questionnaire, blue: groups meetings/ dis-
fill in the provided report template. The report template cussions and yellow: post-survey questionnaire. Central en-
consisted of the following sections: tity in the ERD is the group member, i.e., the decision maker
who is connected to all the other data dimensions (for the
1. Whether a plan for the group decision process was used observers, only the demographic data is collected).
or not and if yes the duration of the different deci-
sion process phases. We note that in [8] a four phases
structure for the decision making process is indicated 4. THE OUTPUT
as typical: 1) Orientation, 2) Discussion, 3) Decision In this section we summarize some concrete output ob-
and 4) Implementation and evaluation of the decision; tained by an initial analysis of the data [5]. However, this
is only one example how this type of studies can help to
2. Group members’ roles (e.g., leader, follower, initiator, obtain deeper insights into the interplay of individual pref-
information giver, opinion seeker); erences and group processes. Various other analyses can be
conducted making use of the rich information that has been
3. Group members’ behavior (i.e., twelve categories of be-
(see Section 3). To facilitate this, we plan to provide the
havior: Show solidarity/ “Friendly”; Show tension re-
data to the research community.
lease; Agree, Give suggestion/ opinion/ information;
Ask for suggestion/ opinion/ information; Disagree; 2
https://survey.aau.at/2012/index.php?sid=98597&lang=
Show tension) - For each group member, the observers de
Figure 2: Structure of the collected data
(Pink - First study phase; Blue - Second study phase; Yellow - Third study phase)
In a first step, we studied whether or not the users were less satisfied group typically all members show disagreement
satisfied with the outcome of the group decision making pro- during the decision making process.
cess, and we tried to understand the impact of the initial
preferences into that. The vast majority of users showed a
high satisfaction for the destination chosen by the group.
5. IMPLICATIONS FOR RECOMMENDER
Obviously this was particularly true for users, where the SYSTEMS
group selection matched their individual top choice. How- As mentioned previously, the proposed observational study
ever, also more than two-thirds of the users, for whom the is ultimately motivated by the goal of designing more effec-
group decision was not in accordance with their most pre- tive group recommender systems. This means that the sys-
ferred destination, were satisfied with the collective choice. tem should better predict, and therefore recommend, which
To some extent this might be related to the fact that the items the group will choose and will make the group mem-
users perceived the different destinations, which could be bers more satisfied. We will now discuss some important
chosen for the group tour, overall as very attractive. How- benefits that we expect the analysis of the data acquired by
ever, our analysis clearly indicated that the group decision observing users’ interactions in group decision making tasks
making process itself played a decisive role in this context: can bring to recommender systems.
group preferences are not just an aggregation of the initial First of all, group recommenders requires the design of
group members’ preferences but are rather constructed dur- ranking functions that can highlight which items a group
ing the process. This was also supported by the fact that must primarily look at. Ranking functions for group rec-
common aggregation strategies in group recommender sys- ommender are based on preference aggregation strategies.
tems were hardly able to predict the outcome of the group While we already mentioned that there is not a single best
decision making process. aggregation strategy that fits all recommendation tasks and
Next, we studied the choice satisfaction of the users in decision contexts, observational study data can be used to
more detail and identified relevant user and group char- choose and customize the aggregation function to the spe-
acteristics in this context. We found some significant and cific contextual conditions of the group. We conjecture that,
moderately high correlations between the individual choice having a family of candidate aggregation functions, one can
satisfaction and personality traits of a user. Also behavioral optimally choose the right one by fitting the observation
patterns during the discussion could be related to the sat- data. For instance, experimental results of the study showed
isfaction of a user as well as the difficulty of the task. To that the social role and personality of the group members
capture the satisfaction of a group, we studied the average influence group choices which was also confirmed in other
choice satisfaction of the group members. Statistical tests studies [9], [18], [19]. Hence, for instance, among a family of
identified significant differences between highly and less sat- multiplicative aggregation models one can fit the importance
isfied groups with respect to a number of factors. These weights of the group members depending on their roles and
factors captured, on the one hand, whether or not the group personality.
perceived the task as difficult. On the other hand, they were A second important usage of observational data is the con-
related to aggregated travel behavioral patterns as well as struction of a more dynamic model of recommendation that
personality traits of the group members. Furthermore, in integrate into the baseline user preference models preference
information derived by the observations of the discussion
Figure 3: Screen-shots of STSGroup, from left to right: (a) Group discussion, (b) Hint suggestions, (c) Group
suggestions.
Figure 4: Screen-shots of the picture-based recommendation engine PixMeAway
process. In fact, it is clear from our study that the final initial prototype implementing this idea is presented in [17].
output decision is not completely determined by the initial That mobile system, which is called STSGroup, allows group
preferences of the users. We conjecture that the observed dy- members to be engaged in a discussion where they can pro-
namic of the users-to-users interactions must be considered pose items that are thought to be suitable for their group
in order to better predict which items may suit the group and react to other group members’ proposals by giving feed-
at that precise point in time. We have for instance men- back such as likes, dislikes or favorites. They can also tag
tioned the observed correlation between the user activity in the proposed items with comments and emoticons as shown
providing information or criticizing options and the satisfac- in Figure 3a. The interactions between the members and
tion for the final choice. As we suggested in the paragraph the system during the group discussion are monitored and
above, also this data can be used to identify a better aggre- taken into account in order to actively provide group mem-
gation function. But, we also conjecture that this type of bers with appropriate directions and recommendations (see
information can be exploited to revise the initial user models Figure 3b and Figure 3c). The group recommendations are
learned by the system using the historical preference data of built up with explanations that are computed on the base
the users. For instance, if a content based model was fitted of the group members’ actions and contexts.
to the known ratings of a user, this model can then be revised A third, probably most fundamental issue, is related to
by considering the items that the user liked or criticized. An the ultimate goals of observational data and the scope of
a group recommender system. Should the recommender fit 4. How to match and compare individual preferences to
the data, i.e., suggest what the users in a given context are the preferences of the group as a whole?
supposed to choose, or should instead the system act as a
mediator, aimed at driving the group towards a more fair 5. How to address ratings/ ranking difference in different
choice? In the first case, as illustrated in the two para- study implementations?
graphs above, the system pleases the group and let it more
6. How to relate participants’ personalities to their pref-
smoothly and efficiently converge towards the decision that
erences?
the group may have taken even without the system inter-
vention. In the second case, the system is instead assuming So far, we were mainly using the average of the individual
that the fairness of a sound aggregation strategy should pre- scores when aggregating them at the group level [5]. How-
vail on the natural group dynamics and will stick to it. This ever, more sophisticated approaches will be applied in future
contraposition is not new in recommender systems: it relates work.
to the question whether a recommender should only suggest Different dimensions of the study procedure can be varied
items predicted to be top choices for the user or inject in in order to grasp diverse insights into the group dynamics
the recommendations items that would make the list of rec- in this particular context. In the following we present some
ommendations more diverse, novel, sustainable, or simply of the variations and their potential implications:
more trendy. In order to address these fundamental ques-
tions, and understand which role the recommender should 1. Duration and timing of the study: In our implementa-
play, live user studies are unavoidable. tions, we noticed different behaviors of the students in
A fourth, very concrete implication of the study is related the study conducted over the three weeks period on the
to the picture-based approach introduced in [15, 16]. The one hand and the study conducted in one lecture ses-
pre-survey questionnaire and the picture-based approach lean sion on the other hand. In the first case students were
upon the same dimensions when capturing a user model, i.e., not explicitly referring to their initial, individual pref-
17 tourist roles and the Big Five Factors. The findings of erences, but were rather discussing their preferences
the observational study will be related to the picture-based in general. In the second case, students were compar-
approach model, which is illustrated in Figure 4, and then ing their initial preferences and their final choice was
generalized to a group recommender system. The proposed based on these comparisons.
research and related challenges are described in [4].
2. Diversity of the ten predefined destinations (e.g., coun-
try side tourism vs. big city tourism; mountain des-
6. DISCUSSION tination vs. sea side destination; hot climate desti-
In this section we summarize the contributions of the pa- nation vs. cold climate destination): Higher diversity
per and mention several challenges that have to be addressed could generate more conflicting preferences in groups
when analyzing the data. Furthermore, we discuss potential and more intense discussions and decision processes.
variations and generalizations of the observational study.
The main contributions of the paper are: 3. Locality of the ten predefined destinations: In our case
the ten destinations (but Amsterdam) were capitals
• A detailed description of the replicable study proce- in Europe and in an hour or two flight distance. By
dure and the instruments used for the data collection changing the locality of the chosen destinations would
that can provide insights into the actual group decision there be some differences in the observed decision pro-
making processes. cess? Furthermore, the locality and overall popular-
• The implementation of the study procedure in a con- ity of the ten chosen destinations were related to the
crete context of tourism and traveling. knowledge that the participants possessed about these
destinations. But, by using less known destinations,
• Experimental results showing that certain individual how would the unfamiliarity with the destinations in-
and group characteristics, which go beyond the initial fluence the decision process?
preferences of the individuals and their straightforward
aggregation, play an important role in the final choice 4. Groups size: The conducted data analysis showed dif-
of the group. ferences in groups’ satisfaction with respect to the group
size - smaller groups tend to be more satisfied with the
• The implications of the observational study for group group choice than the larger groups, which is quite in-
recommender systems and different aspect that should tuitive. Nevertheless, varying the group size in the
be considered when building such systems. study can provide insights in different aspects that
should be considered.
During the initial data analysis, we encountered several
challenges related to data measurements we used. These 5. Budget: Including budget into the group discussion in-
challenges are at the same time limitations of the study and creases the complexity of the task for the participants
need to be addressed in the future work: and it also enables more realistic setting of the decision
process in the context of traveling.
1. How to aggregate different individual scores, e.g., per-
sonality traits, at the group level? 6. Group decision task : If the group were to choose a
2. How to measure diversity among group members with point of interest that they actually had to visit to-
respect to the different data dimensions? gether right after the group discussion, then the group
members might pursue their preferences and interests
3. How to distinguish satisfied from not so satisfied groups? in a more natural manner and more persistently.
7. Domain: The same study could be carried out in a
different domain, such as music, movies, restaurant, [12] A. Jameson. More than the sum of its members:
etc. In this case it would be much easier to introduce challenges for group recommender systems. In
a realistic setting to participants, but the discussion Proceedings of the working conference on Advanced
process, in this case, would clearly be much different. visual interfaces, pages 48–54, 2004.
[13] J. Masthoff. Group recommender systems:
To summarize, in this paper we presented the observa- aggregation, satisfaction and group attributes. In
tional study implemented at several universities, the instru- F. Ricci, L. Rokach, and B. Shapira, editors,
ments used for the data collection and described the col- Recommender Systems Handbook, pages 743–776.
lected data. We stressed the implications of the study for Springer, 2015.
group recommender systems and our future work relying on [14] K. McCarthy, L. McGinty, B. Smyth, and M. Salamo.
the founding of this study. At the end, we outlined main con- The needs of the many: a case-based group
tributions, introduced challenges and limitations detected by recommender system. Advances in Case-Based
now. Reasoning, pages 196–210, 2006.
[15] J. Neidhardt, R. Schuster, L. Seyfang, and
7. REFERENCES H. Werthner. Eliciting the users’ unknown preferences.
[1] R. F. Bales. A set of categories for the analysis of In Proceedings of the 8th ACM Conference on
small group interaction. American Sociological Review, Recommender systems, pages 309–312, 2645767, 2014.
15:257–263, 1950. ACM.
[2] P. Bekkerman, S. Kraus, and F. Ricci. Applying [16] J. Neidhardt, L. Seyfang, R. Schuster, and
cooperative negotiation methodology to group H. Werthner. A picture-based approach to
recommendation problem. In Proceedings of Workshop recommender systems. Information Technology &
on Recommender Systems in 17th European Tourism, 15(1):49–69, 2015.
Conference on Artificial Intelligence (ECAI’06), pages [17] T. N. Nguyen and F. Ricci. Supporting group decision
72–75, 2006. making with recommendations and explanations. In
[3] H. Blanco and F. Ricci. Inferring user utility for query Posters, Demos, Late-breaking Results and Workshop
revision recommendation. In Proceedings of the 28th Proceedings of the 24th Conference on User Modeling,
Annual ACM Symposium on Applied Computing, SAC Adaptation, and Personalization (UMAP 2016),
’13, Coimbra, Portugal, March 18-22, 2013, pages Halifax, Canada, 2016.
245–252, 2013. [18] L. Quijano-Sanchez, J. A. Recio-Garcia,
[4] A. Delic. Picture-based approach to group B. Diaz-Agudo, and G. Jimenez-Diaz. Social factors in
recommender systems in e-tourism domain. In group recommender systems. ACM Transactions on
Conference Proceedings of the 24th Conference on Intelligent Systems and Technology (TIST), 4(1):8,
User Modeling, Adaptation, and Personalization 2013.
(UMAP 2016), Halifax, Canada, 2016. [19] J. A. Recio-Garcia, G. Jimenez-Diaz, A. A.
[5] A. Delic, J. Neidhardt, N. Nguyen, F. Ricci, L. Rook, Sanchez-Ruiz, and B. Diaz-Agudo. Personality aware
H. Werthner, and M. Zanker. Observing group recommendations to groups. In Proceedings of the 3rd
decision making processes. In Proceedings of the tenth ACM conference on Recommender systems, pages
ACM conference on Recommender systems, RecSys’16, 325–328, NY, USA, 2009.
2016. [20] F. Ricci, L. Rokach, and B. Shapira. Recommender
[6] A. Delic, J. Neidhardt, and H. Werthner. Are sun systems: Introduction and challenges. In
lovers nervous? - research note at enter 2016 etourism Recommender Systems Handbook, pages 1–34. 2015.
conference. Bilbao, Spain, 2016. [21] M. Stettinger, A. Felfernig, G. Leitner, S. Reiterer,
[7] I. Fernández-Tobı́as, M. Braunhofer, M. Elahi, and M. Jeran. Counteracting serial position effects in
F. Ricci, and I. Cantador. Alleviating the new user the choicla group decision support environment. In
problem in collaborative filtering by exploiting Proceedings of the 20th International Conference on
personality information. User Model. User-Adapt. Intelligent User Interfaces, pages 148–157, GA, USA,
Interact., 26(2-3):221–255, 2016. 2015.
[8] D. Forsyth. Group Dynamics. Wadsworth Cengage [22] R. S. Tindale and T. Kameda. Social sharedness as a
Learning, 6th edition, 2014. unifying theme for information processing in groups.
[9] M. Gartrell, X. Xing, Q. Lv, A. Beach, R. Han, Group Processes and Intergroup Relations,
S. Mishra, and K. Seada. Enhancing group 3(2):123–140, 2000.
recommendation by incorporating social relationship [23] H. Werthner, A. Alzua-Sorzabal, L. Cantoni,
interactions. In Proceedings of the 16th ACM A. Dickinger, U. Gretzel, D. Jannach, J. Neidhardt,
international conference on Supporting group work, B. Pröll, F. Ricci, M. Scaglione, B. Stangl, O. Stock,
pages 97–106, FL, USA, 2010. and M. Zanker. Future research issues in IT and
[10] H. Gibson and A. Yiannakis. Tourist roles: Needs and tourism. J. of IT & Tourism, 15(1):1–15, 2015.
the lifecourse. Annals of tourism research, [24] H. Werthner and F. Ricci. E-commerce and tourism.
29(2):358–383, 2002. Communications of the ACM, 47(12):101–105, 2004.
[11] L. R. Goldberg. An alternative ”description of [25] A. Yiannakis and H. Gibson. Roles tourists play.
personality”: the big-five factor structure. Journal of Annals of tourism Research, 19(2):287–303, 1992.
personality and social psychology, 59(6):1216, 1990.