=Paper=
{{Paper
|id=Vol-1533/paper4
|storemode=property
|title=Evaluating User's Personality and Social Interactions for Groups Recommendations
|pdfUrl=https://ceur-ws.org/Vol-1533/paper4.pdf
|volume=Vol-1533
|authors=Francesco Barile,Francesco Cervone,Silvia Rossi
|dblpUrl=https://dblp.org/rec/conf/dmrs/BarileCR15
}}
==Evaluating User's Personality and Social Interactions for Groups Recommendations==
https://ceur-ws.org/Vol-1533/paper4.pdf
Evaluating User’s Personality and Social Interactions
for Groups Recommendations
Francesco Barile1 , Francesco Cervone2 and Silvia Rossi2
1
Dipartimento di Matematica e Applicazioni,
Universita’ degli Studi di Napoli “Federico II”, Napoli, Italy
francesco.barile@unina.it
2 Dipartimento di Ingegneria Elettrica e Tecnologie dell’Informazione,
Universita’ degli Studi di Napoli “Federico II”, Napoli, Italy
silvia.rossi@unina.it
Abstract. Common approaches to provide group recommendations are based
on the aggregation of the recommendations that are provided, for each group’s
member, by an individual recommendation system, using social choice functions.
These techniques do not consider factor like social interactions, roles, influences,
and group members’ personality that model real group’s decision-making pro-
cess. Recent approaches tried to include these factors by introducing social-based
weights in the aggregation of preferences or social aware utility functions. On the
basis of these evaluations we propose two possible approaches, one based on the
study of user’s personality, in particular, on the agreeableness factor, and a sec-
ond one based on the analysis of social interaction between group’s members on
a social network, to determine a dominance factor for each user in the group. The
conducted pilot studies show that these approaches can increase the goodness
of the recommendation provided and the satisfaction of group’s members with
respect to standard aggregation mechanisms.
Keywords: Group Decision Making, Social Choice, Group Recommendation, Small
Groups, Social Networks.
1 Introduction
Group recommendation approaches rely either on building a single group profile, result-
ing from the combination of the profiles of all the users, or on merging the recommenda-
tion lists generated for each individual users, at runtime, using different group decision
strategies. In this case, we talk about Social Choice functions. These strategies, accord-
ing to [14], can be classified as majority-based (mainly implemented as voting mecha-
nisms to determine the most popular choices among alternatives), consensus-based (that
try to average among all the possible choices and preferences), and role-based (that ex-
plicitly take into account possible roles and hierarchical relationships among members).
Examples of these techniques are illustrated in [9], the most common approaches are
based on the average satisfaction and least misery techniques.
Nevertheless, many of these techniques do not consider social relationships among
the group members [7], while the design and implementation of group recommendation
systems, and, more generally, of decision support systems, should take into account the
type of control in the group decision-making process, and the diversity and the dy-
namics of relationships, roles and mutual influences among the group members [7]. For
example, the decision of a group member whether or not to accept a given recommenda-
tion may depend not only on the own evaluation of the content of the recommendation,
but also on the beliefs about the evaluations of the other group members [2].
PolyLens [11] has been one of the first approaches to include social characteris-
tics within a group recommendation system. A more recent example is represented by
the work of [7], where the Authors started to evaluate the group members’ weights,
in terms of individual members’ importance or influence in a group, for movie recom-
mendations. The defined group consensus function relies on the concept of “expertise”
and “group dissimilarity”. Also, it introduces the idea of diversifying the social choice
strategy to use on the basis of the characteristics of the group. The Authors calculate a
“social value” on the basis of social interactions between group members and, then, use
this value to discriminate the strategy to apply for the group.
Another way to use interactions between group members is presented in [13]. Here,
the authors introduce the concept of empathetic utility on social networks: the satisfac-
tion of an individual depends from both his intrinsic utility and his empathetic utility
deriving from the happiness of his neighbors in the social network [13]. Based on this
idea, individual preferences are aggregated in a weighted social choice function that
takes into account local relationships with neighborhoods in the network. However, in
[13] the Authors do not specify how to evaluate such numerical relationships, while
they focus on computational aspects of scaling up with large networks of friends.
2 A Personality based Social Utility Function
In our opinion, a key factor that can influence the group decision in a realistic scenario
is user’s personality. Some people could rarely change their minds because they believe
that their decision is the best for everyone, or simply because they do not want to re-
duce their utility in favor of others. Other types of people instead, can be worried about
the satisfaction of all the other members, at the cost of the personal one. To involve
these elements, a study of users’ personalities through some models proposed in human
sciences area is necessary. One of the most common is the Five-Factor Model (FFM),
that summarize the behavioral features of a person into five factors, also called Big-
Five: openness, conscientiousness, extraversion, neuroticism and agreeableness [5]. In
particular, we focused on the role of the agreeableness factor in the definition of a util-
ity function that models altruistic behavior. While the role of personality has been ad-
dressed before, in literature, to improve the performance of Recommendation Systems
[10, 8], up to our knowledge, this is the first attempt to introduce personality factors in
group decision making through the use of social choice functions.
We start assigning a utility to the items for each user that takes into account both the
personal and the group satisfaction, depending on an altruistic factor (i.e., the agree-
ableness). It can be described as “the satisfaction of the user if the recommendation
system chooses that item for the group”. Once the new utilities are evaluated, the goal
is to recommend items that maximize the social welfare. We used a model developed
by [4] and a Mini-IPIP (Mini International Personality Item Pool) [6] to evaluate the
agreeableness level of individual users.
To evaluate this approach, we conducted a user study on movie recommendations,
where we compare the results of our social function with a simple Least Misery strategy
(LM). Results showed that for small groups a LM performs slightly better. In particu-
lar, for two people groups LM is the best choice; in the other cases the two methods
are comparable and show similar performances. Our utility function improves its effec-
tiveness proportionally to the group size: the larger is the group, the greater will count
altruism in the final decision.
3 Dominance Weighted Social Choice Functions
According to [9], users involved in real interaction seem to care about fairness and to
avoid misery. In [12], we decided to use a fairness strategy and one based on average
satisfaction and to weigh such functions with a measure of the influence of each user on
the other group’s members, and, consequently, on the group’s final decision. To make
this, we evaluate the weight of the relationship between pairs of users from the analy-
sis of the interactions on an OSN. We are interested in the analysis of the strength and
the directionality of online social interactions in order to gather useful information on
intra-group relationships, and use the strength of the different pairwise relationships in
an aggregated manner in order to evaluate the power/dominance of each single mem-
ber on the whole group. To compute the users’ ranking, we decided to use a simple
“non-semantic” approach defined in our previous work [3]. Such popularity values are
obtained implementing an extension of the well-known PageRank algorithm [1] start-
ing from the users’ interactions on the social network facebook.com. These dominance
values are used as weights for the average satisfaction strategy, while, for the fairness
strategy, we proposed to use the dominance values to provide a ranking and to sort the
users.
To evaluate our approach, we conducted two pilot studies with real users involved
in the task of planning a trip in a city, and compare the results of our weighted functions
with respect to their standard not-weighted implementations. Results showed that the
weighted functions had better performances. In particular, in the first user study with
binary selections (e.g., no rating provided), a bigger improvement was noted for the
average satisfaction function, which also performed better than the fairness that suf-
fers more of random choices. On the contrary, in the second one, which involved POI
rankings, fairness strategy has a bigger acceptance rate and appreciation evaluation.
4 Conclusions
In this work, we introduced two approaches that include factors like user’s person-
ality, social roles and influences in the process of preferences aggregation for group
recommendations. We evaluated these approaches through pilot user studies that show
encouraging results, in particular, for the dominance weighted social choice functions.
Nevertheless, a more deep analysis is necessary, involving a greater number of users
and groups of larger dimensions.
Acknowledgments
The research leading to these results has received funding from the Italian Ministry of
University and Research and EU under the PON OR.C.HE.S.T.R.A. project (ORgani-
zation of Cultural HEritage for Smart Tourism and Real-time Accessibility).
References
1. Brin, S., Page, L.: The anatomy of a large-scale hypertextual web search engine. Comput.
Netw. ISDN Syst. 30(1-7), 107–117 (1998)
2. Cantador, I., Castells, P.: Group recommender systems: New perspectives in the social
web. In: Recommender Systems for the Social Web, Intelligent Systems Reference Library,
vol. 32, pp. 139–157. Springer (2012)
3. Caso, A., Rossi, S.: Users ranking in online social networks to support pois selection in small
groups. In: Extended Proceedings of the 22nd Conference on User Modelling, Adaptation
and Peronalization - UMAP 2014. pp. 5–8 (2014)
4. Charness, G., Rabin, M.: Understanding social preferences with simple tests. Quarterly jour-
nal of Economics pp. 817–869 (2002)
5. Costa, P.T., MacCrae, R.R.: Revised NEO Personality Inventory (NEO PI-R) and NEO Five-
Factor Inventory (NEO FFI): Professional Manual. Psychological Assessment Resources
(1992)
6. Donnellan, M.B., Oswald, F.L., Baird, B.M., Lucas, R.E.: The mini-ipip scales: tiny-yet-
effective measures of the big five factors of personality. Psychological assessment 18(2), 192
(2006)
7. Gartrell, M., Xing, X., Lv, Q., Beach, A., Han, R., Mishra, S., Seada, K.: Enhancing group
recommendation by incorporating social relationship interactions. In: Proc. of the 16th ACM
International Conference on Supporting Group Work. pp. 97–106. ACM (2010)
8. Hu, R., Pu, P.: Enhancing collaborative filtering systems with personality information. In:
Proceedings of the Fifth ACM Conference on Recommender Systems. pp. 197–204. RecSys
’11, ACM (2011)
9. Masthoff, J.: Group recommender systems: Combining individual models. In: Recommender
Systems Handbook, pp. 677–702. Springer US (2011)
10. Nunes, M.A.S., Hu, R.: Personality-based recommender systems: An overview. In: Proceed-
ings of the Sixth ACM Conference on Recommender Systems. pp. 5–6. RecSys ’12, ACM,
New York, NY, USA (2012)
11. O’Connor, M., Cosley, D., Konstan, J.A., Riedl, J.: Polylens: A recommender system for
groups of users. In: Proc. of the 7th European Conf. on CSCW. pp. 199–218 (2001)
12. Rossi, S., Caso, A., Barile, F.: Combining users and items rankings for group decision sup-
port. In: Trends in Practical Applications of Agents, Multi-Agent Systems and Sustainability,
Advances in Intelligent Systems and Computing, vol. 372, pp. 151–158. Springer Interna-
tional Publishing (2015)
13. Salehi-Abari, A., Boutilier, C.: Empathetic social choice on social networks. In: 13th Inter-
national Conference on Autonomous Agents and Multiagent Systems. pp. 693–700 (2014)
14. Senot, C., Kostadinov, D., Bouzid, M., Picault, J., Aghasaryan, A., Bernier, C.: Analysis of
strategies for building group profiles. In: User Modeling, Adaptation, and Personalization,
LNCS, vol. 6075, pp. 40–51 (2010)