International Journal of Human-Computer Studies 121 (2019) 1–3

Contents lists available at ScienceDirect

International Journal of Human-Computer Studies

journal homepage: www.elsevier.com/locate/ijhcs

Advances in computer-human interaction for recommender systems T

(AdCHIReS)

A R T I C LE I N FO

Keywords:
Human-computer interaction
Recommender systems

1. Introduction Recommender Systems in broader contexts, there is a need for struc-

tured and systematic approaches to engineer such complex computing
Recommender Systems produce suggestions of items or content that systems.
users have not considered yet, but that might be interesting for them.
Such recommendations are produced by analyzing what they pre- 2. Background
viously consumed (bought, watched, or listened), or by the identifica-
tion of similarities with other users. Users usually express their pre- As previously mentioned, numerical ratings are biased, since users
ferences by providing explicit ratings; however, these might be not very tend to express whether they loved or hated something, without giving
informative, since users might only indicate whether they liked some- a real quantitative and qualitative measure of what they experienced
thing or not, thus employing only the ratings at the end of the scale (this with an item. It has also been shown that an effective collection of
is also one of the reasons why Netflix employed a thumb-up/thumb- explicit preferences through ratings is not trivial, since the effort re-
down rating system). However, between these two extremes there lies a quired to the users affects the accuracy of the system (Cremonesi et al.,
set of different actions in the interface that might be interpreted as 2012), and different rating scales lead to different system
feedback, but that needs to be collected implicitly. Even if the literature performance (Sparling and Shilad, 2011). Moreover, studies have
provides different techniques for collecting implicit feedback, they are shown that choice-based approaches are more effective than rating-
usually tailored to specific types of applications. based ones (Willemsen, 2017). Therefore, an approach based on im-
From the user’s point of view, Recommender Systems remain a plicitly-collected data is more objective (Buder and Schwind, 2012).
black box that suggests content, but the users hardly understand why Accurately exploiting the interactions with the UI poses a set of
some items are included in the list. The relevance of this issue has in- problems for engineers and designers. First of all, it is necessary to filter
creased in recent years, as the introduction of approaches based on the huge amounts of data generated during the interactions for de-
latent features (such as Matrix Factorization or Deep Learning) has scribing the users’ behavior. In addition, the implicit tracking of the
made it very hard to connect user preferences with the recommended user’s actions may be a source of distrust for the application, which
items. Providing the users with an understandable representation of should help the users in understanding the rationale behind the re-
how the system represents them and allowing them to control the re- commendations. Having a coherent UI and making a recommender
commendation process would lead to benefits in how the re- system understandable by the users are considered as challenges in this
commendations are perceived and in the capability of the system to be research area, as highlighted by Konstan in his ACM “Intelligent User
persuasive. Such transparency is one of the multiple (and usually con- Interfaces”’ tutorial (Konstan, 2011). Recent literature (Calero Valdez
flicting) requirements of Recommender Systems. et al., 2016) has confirmed how Human-Computer Interaction in Re-
Beyond the classical engineering of Recommender Systems, based commender Systems should be characterized by aspects such as con-
on data processing, filtering, and sorting, the engineering aspects trollability, adaptivity, capability to generate trust in the users, visua-
should also cover aspects related to how users interact with these sys- lization, and communication of risks.
tems, including how to input data, how to define and evolve the user Different studies show that the capability to solicit responses from
model, how to present information to users, and how these can ma- the users is one of the key factors in their interaction with
nipulate such information. These engineering processes might benefit technologies (Fogg, 2002; Nass and Moon, 2000; Reeves and Nass,
from practice in specific areas, such as web configurators (which guide 1996). Moreover, studies conducted in the Recommender Systems re-
the users in the inspection of possible product variants) and safety search area, confirmed that various system characteristics influence the
critical interactive systems (where predictability and consistency over way the users evaluate it and perceive the recommendations (Nguyen
executions are prerequisite to certification). In order to deploy et al., 2007; Pu and Chen, 2007; Qiu, 2010). Several approaches

https://doi.org/10.1016/j.ijhcs.2018.07.008

Available online 26 July 2018

1071-5819/ © 2018 Published by Elsevier Ltd.
 International Journal of Human-Computer Studies 121 (2019) 1–3

embody the interactions of the users with the device in the re- recommendation algorithm and is based on the insight that properties
commendation process. TasteWeights (Bostandjiev et al., 2012) is a describing the items the user previously liked, as well as the suggestions
hybrid approach that considers classic recommendation approaches and she received, can be effectively used to explain the recommendations,
an interactive interface to elicit preferences from the end user. He et al. using the Linked Open Data semantics. The validation work shows the
(2016) evaluate the transparency of 24 Recommender Systems that effectiveness of the explanation in three different recommendation
embody interactive visualization techniques to support user interaction. domains (books, music, and movies).
In Pu et al. (2011), present ResQue, a user-centric framework to mea- Calero Valdez and Ziefle (2018), in their article “The Users’ Per-
sure the user experience with a Recommender System. Knijnenburg spective on Privacy-Utility Trade-offs in Health Recommender Sys-
et al. (2011) studied five interaction methods of the users with a Re- tems”, investigate the importance of the techniques for preserving
commender System (the implicit one is among them) and showed that privacy in sharing personal health data. Users disagreed with sharing
different users have a different preferred method of interaction. data for commercial purposes regarding mental illnesses and with high
de-anonymization risks. They showed little concern when data is used
3. The special issue for scientific purposes and is related to physical illnesses.
Fayollas et al. (2018), in their article “Engineering Issues Related to
The nine accepted articles in this special issue cover many of the the Development of a Recommender System in a Critical Context: Ap-
aforementioned themes, with innovative techniques for taking into plication to Interactive Cockpits”, explore the requirements for de-
account user interactions in Recommender Systems. The research con- ploying Recommender Systems in critical contexts. They propose a
tributions advance the state of the art considering different interaction generic architecture for decomposing Recommender Systems and
aspects and scenarios for the recommendation, ranging from content identify a set of requirements that must be fulfilled for each component.
suggestion to safety-critical operation support, by considering different Finally, they concretely illustrate the approach in the context of future
aspects such as interactivity, mood, privacy, and social activities. aircraft alerting systems.
In their article “User Perception of Sentiment-Integrated Critiquing Andjelkovic et al. (2018), in their article “MoodPlay: Interactive
in Recommender Systems”, Chen et al. (2018) propose a novel criti- Music Recommendation based on Artists’ Mood Similarity”, introduce a
quing method that extracts feature sentiments, together with a model novel interface for recommending music and artists, according to both
that incorporates both static item values and sentiments for computing content characteristics and mood, together with a novel recommenda-
the product utility. Through different experiments they show that the tion model. The evaluation shows that the interface increases the re-
approach increases different user’s perceptions, such as the preference commendation acceptance and understanding in users.
certainty, the product knowledge, the decision confidence, and the In conclusion, this special issue covers many aspects of the inter-
perceived information usefulness. action with Recommender Systems, discussing the relevance, the ad-
Loepp et al. (2018), in their article “Interactive Recommending with vantages and the complexity of linking the underlying recommendation
Tag-Enhanced Matrix Factorization (TagMF)”, enhance a latent factor algorithms with the user interface. The results show both an increment
model for providing recommendations with semantic tags, used for in the recommendation performance and higher overall usability for the
explaining the semantics of the latent factors to the users. In this way, end-users.
users can explicitly express their interests, understand and control their The Guest Editors
preferences, and criticize the recommendation results. A user study Lucio Davide Spano, University of Cagliari
shows that the approach increases the perceived quality of the re- Ludovico Boratto, Eurecat
commendations. ⁎
Lai et al. (2018), in their article “A Social Recommendation Method Lucio Davide Spanoa, Ludovico Boratto ,b,
a
Based on the Integration of Social Relationship and Product Popu- Department of Mathematics and Computer Science, University of Cagliari,
larity”, propose a social recommendation method that integrates in- Italy
b
teractions, trust relationships, and product popularity, in order to in- Data Science and Big Data Analytics, Eurecat, Technology Center of
crease the recommendation performance in social networks and better Catalonia, Spain
support their purchase decision making. E-mail address: [email protected] (L. Boratto),
Liu et al. (2018), in their article “Real-time Social Recommendation
Based on Graph Embedding and Temporal Context”, introduce an ef- References
ficient model that exploits social network information for providing
real-time recommendations, based on dynamic graph-based embed- Andjelkovic, I., O’Donovan, J., Parra, D., 2018. Moodplay: Interactive music re-
ding. The technique, while maintaining a real-time efficiency, over- commendation based on artists’ mood similarity”. Int. J. Hum-Comput. Stud. Special
Issue on Advances in Computer-Human Interaction for Recommender Systems.
comes the state-of-the-art recommendation results on two large data-
Bostandjiev, S., O’Donovan, J., Höllerer, T., 2012. Tasteweights: A visual interactive
sets. hybrid recommender system. Proceedings of the Sixth ACM Conference on
Cardoso et al. (2018), in their article “IntersectionExplorer, a Multi- Recommender Systems. ACM, New York, NY, USA, pp. 35–42. https://doi.org/10.
1145/2365952.2365964.
Perspective Approach for Exploring Recommendations”, propose a
Buder, J., Schwind, C., 2012. Learning with personalized recommender systems: a psy-
scalable visualization, interleaving the output of several recommenda- chological view. Comput. Hum. Behav. 28 (1), 207–216. https://doi.org/10.1016/j.
tion engines with human-generated data, for increasing the exploration chb.2011.09.002.
experience and enhancing the potential to find relevant items. They Calero Valdez, A., Ziefle, M., 2018. The users’ perspective on privacy-utility trade-offs in
health recommender systems. Int. J. Hum-Comput. Stud. Special Issue on Advances in
applied the visualization in a conference paper recommendation sce- Computer-Human Interaction for Recommender Systems.
nario, showing that the combinations of machine-produced re- Calero Valdez, A., Ziefle, M., Verbert, K., 2016. Hci for recommender systems: The past,
commendations with bookmarks of users and tags resulted in an in- the present and the future. Proceedings of the 10th ACM Conference on
Recommender Systems. ACM, New York, NY, USA, pp. 123–126. https://doi.org/10.
creased likelihood of finding relevant papers in explorations. 1145/2959100.2959158.
Musto et al. (2018), in their article “Linked Open Data-based Ex- Cardoso, B., Sedrakyan, G., Gutirrez, F., Brusilovsky, P., Parra, D., 2018.
planations for Transparent Recommender Systems”, focus on the pro- Intersectionexplorer, a multi-perspective approach for exploring recommendations.
Int. J. Hum-Comput. Stud. Special Issue on Advances in Computer-Human Interaction
blem of explaining the recommendations to users, in order to support for Recommender Systems.
the creation of a mental model. The approach is independent of the

⁎
Corresponding author.

2
 International Journal of Human-Computer Studies 121 (2019) 1–3

Chen, L., Yan, D., Wang, F., 2018. User perception of sentiment-integrated critiquing in tag-enhanced matrix factorization (tagMF). Int. J. Hum-Comput. Stud. Special Issue
recommender systems. Int. J. Hum-Comput. Stud. Special Issue on Advances in on Advances in Computer-Human Interaction for Recommender Systems.
Computer-Human Interaction for Recommender Systems. Musto, C., Narducci, F., Lops, P., De Gemmis, M., Semeraro, G., 2018. Linked open data-
Cremonesi, P., Garzottto, F., Turrin, R., 2012. User effort vs. accuracy in rating-based based explanations for transparent recommender systems. Int. J. Hum-Comput. Stud.
elicitation. Proceedings of the Sixth ACM Conference on sRecommender Systems. Special Issue on Advances in Computer-Human Interaction for Recommender
ACM, New York, NY, USA, pp. 27–34. https://doi.org/10.1145/2365952.2365963. Systems.
Fayollas, C., Bouzekri, E., Canny, A., Barboni, Y., Ericand, Deleris, Palanque, P., Martinie, Nass, C., Moon, Y., 2000. Machines and mindlessness: Social responses to computers. J.
C., 2018. Engineering issues related to the development of a recommender system in Social Issues 56, 81–103.
a critical context: Application to interactive cockpits. Int. J. Hum-Comput. Stud. Nguyen, H., Masthoff, J., Edwards, P., 2007. Persuasive effects of embodied conversa-
Special Issue on Advances in Computer-Human Interaction for Recommender tional agent teams. Proceedings of the 12th International Conference on Human-
Systems. computer Interaction: Intelligent Multimodal Interaction Environments. Springer-
Fogg, B.J., 2002. Persuasive Technology. Using computers to Change What We thiznk and Verlag, Berlin, Heidelberg, pp. 176–185. http://dl.acm.org/citation.cfm?id=
do. Morgan Kaufmann Publishers, Elsevier, San Francisco. 1769590.1769610.
He, C., Parra, D., Verbert, K., 2016. Interactive recommender systems: A survey of the Pu, P., Chen, L., 2007. Trust-inspiring explanation interfaces for recommender systems.
state of the art and future research challenges and opportunities. Expert Syst. Appl. Know.-Based Syst. 20 (6), 542–556. https://doi.org/10.1016/j.knosys.2007.04.004.
56, 9–27. https://doi.org/10.1016/j.eswa.2016.02.013. Pu, P., Chen, L., Hu, R., 2011. A user-centric evaluation framework for recommender
Knijnenburg, B.P., Reijmer, N.J.M., Willemsen, M.C., 2011. Each to his own: How dif- systems. Proceedings of the Fifth ACM Conference on Recommender Systems. ACM,
ferent users call for different interaction methods in recommender systems. New York, NY, USA, pp. 157–164. https://doi.org/10.1145/2043932.2043962.
Proceedings of the Fifth ACM Conference on Recommender Systems. ACM, New York, Qiu, L., 2010. Designing social interactions with animated avatars and speech output for
NY, USA, pp. 141–148. https://doi.org/10.1145/2043932.2043960. Product Recommendation Agents in electronic commerce. Ph.D. thesis. doi: https://
Konstan, J.A., 2011. Tutorial / HCI for recommender systems: A tutorial. Proceedings of doi.org/10.14288/1.0092949.
the 16th International Conference on Intelligent User Interfaces. ACM, New York, NY, Reeves, B., Nass, C., 1996. The Media Equation: How People Treat Computers, Television,
USA, pp. pp.463. https://doi.org/10.1145/1943403.1943501. and New Media Like Real People and Places. Cambridge University Press, New York,
Lai, C.-H., Lee, S.-J.L., Huang, H.-L., 2018. A social recommendation method based on the NY, USA.
integration of social relationship and product popularity. Int. J. Hum-Comput. Stud. Sparling, E.I., Shilad, S., 2011. Rating: How difficult is it? Proceedings of the Fifth ACM
Special Issue on Advances in Computer-Human Interaction for Recommender Conference on Recommender Systems. ACM, New York, NY, USA, pp. 149–156.
Systems. https://doi.org/10.1145/2043932.2043961.
Liu, P., Zang, L., Gulla, J.A., 2018. Real-time social recommendation based on graph Willemsen, M.C., 2017. Understanding user preferences and goals in recommender sys-
embedding and temporal context. Int. J. Hum-Comput. Stud. Special Issue on tems. Proceedings of the CEUR Workshop Proceedings. 1945. pp. 1–2. Keynote at the
Advances in Computer-Human Interaction for Recommender Systems. workshop.
Loepp, B., Donkers, T., Kleemann, T., Ziegler, J., 2018. Interactive recommending with