Towards a Design Space for Personalizing the Presentation of - - PowerPoint PPT Presentation

Towards a Design Space for Personalizing the Presentation of Recommendations

Catalin-Mihai Barbu and Jürgen Ziegler

2nd Workshop on Engineering Computer-Human Interaction in Recommender Systems (EnCHIReS 2017) Lisbon, 26 June 2017

Recommender Systems

▪ Information systems that match users with items they might find interesting ▪ Several types

– Content-based – Collaborative Filtering (CF) – Knowledge-based – Hybrid

▪ Mature algorithms

Personalization

“A process of collecting and using personal information to uniquely tailor products, content and services to an individual” (Tuzhilin, 2000)

Why Personalize at All?

▪ Cater to different tastes, interests, and preferences ▪ Simplify user experience ▪ Support users’ decision making ▪ Increase trust in the recommended items (hopefully) Recommendations are only as good as users perceive them to be (perceived vs. objective accuracy)

Personalizing the Presentation of Recommendations

▪ Exploit user profile to identify important attributes

– Stated preferences (reviews, ratings) – Interactions (applied filters, discarded options)

▪ Adapt presentation to suit the user’s needs

– E.g., personalized summaries

Overview of Design Space

Design Space Dimensions: Modality

Design Space Dimensions: Modality

▪ Present information using appropriate modalities ▪ Design patterns Cremonesi et al. (2016) ▪ Changing modalities

Nanou et al. (2010)

Design Space Dimensions: Salience

Design Space Dimensions: Salience

▪ Draw users’ attention to important information

– Show relevant additional information directly

▪ De-emphasize less important features

– Or hide them completely

Zhao et al. (2016)

Design Space Dimensions: Comparison Functions

Design Space Dimensions: Comparison Functions

▪ Help users evaluate attribute values across recommendations

– Normalize / transform measurement units

Pu et al. (2007)

Design Space Dimensions: Interactive Control

Design Space Dimensions: Interactive Control

▪ Provide mechanisms to influence the output

– Preference elicitation (e.g., using tags)

Loepp et al. (2015)

Design Space Dimensions: Interactive Control

▪ Provide mechanisms to influence the output

– Interactive critiquing

Donkers et al. (2017)

Design Space Dimensions: Explanations

Design Space Dimensions: Explanations

▪ Facilitate the discovery of supporting evidence

– Objective info might be wrong / incomplete – Subjective info might be out of context

Vig et al. (2009) Chen et al. (2017)

Design Space Dimensions: Trust Cues

Design Space Dimensions: Trust Cues

▪ Determine the reliability of presented information

– Complement item descriptions with objective measurements – Determine credibility of user reviews

Yan et al. (2013)

Interactive Mechanisms

Goal: Support the user’s decision-making process ▪ Content

– Show personalized summaries – Allow users to “zoom” in / out – Manipulate the order in which sections are presented

▪ User profile

– Preview effect of changes on the presentation

Evaluation Criteria

▪ Suitability w.r.t. the user’s informational need

– Consequences of poor choices – Level of detail – User characteristics

Conclusions & Future Work

▪ Personalizing the presentation of recommendations can lead to increased transparency and control ▪ Proposed dimensions need to be validated empirically ▪ Quality of user data is a limiting factor

Thank you!

Questions?

References

Chen, Y. Interface and interaction design for group and social recommender systems. In Proc. RecSys '11, ACM (2011), 363–366. Cremonesi, P., Elahi, M., & Garzotto, F. User interface patterns in recommendation-empowered content intensive multimedia applications. Multimedia Tools and Applications, Springer (2016), 1–35. Donkers, T., Loepp, B., & Ziegler, J. Tag-enhanced collaborative filtering for increasing transparency and interactive

• control. In Proc. UMAP '16, ACM (2016), 169–173.

Loepp, B., Herrmanny, K., & Ziegler, J. Blended recommending: Integrating interactive information filtering and algorithmic recommender techniques. In Proc. CHI '15, ACM (2015), 975–984. Nanou, T., Lekakos, G., & Fouskas, K. The effects of recommendations’ presentation on persuasion and satisfaction in a movie recommender system. Multimedia Systems 16, 4-5 (2010), 219–230. Pu, P., & Chen, L. Trust-inspiring explanation interfaces for recommender systems. Knowledge-Based Systems, 20-6 (2007), 542–556. Tuzhilin, A. Report on the KDD2000 panel personalization and data mining: Exploring the synergies. ACM SIGKDD Explorations Newsletter, 2-2 (2000), 115–116. Vig, J., Sen, S., & Riedl, J. Tagsplanations: explaining recommendations using tags. In Proc. IUI '09, ACM (2009), 47– 56. Yan, Z., Liu, C., Niemi, V., & Yu, G. Exploring the impact of trust information visualization on mobile application usage. Personal and ubiquitous computing, 17-6 (2013), 1295–1313. Zhao, Q., Chang, S., Harper, F. M., & Konstan, J. A. Gaze prediction for recommender systems. In Proc. RecSys '16, ACM (2016), 131–138.