Affect- and Personality-based Recommender Systems Part II: - PowerPoint PPT Presentation

Affect- and Personality-based Recommender Systems Part II: Acquisition, Usage in Recommender Systems Marko Tkalčič, Free University of Bozen-Bolzano ACM Summer School on Recommender Systems 2017 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 1/53

Table of Contents Recap Acquisition of Personality Usage of personality in recommender systems Acquisition of Emotions Usage of emotions in recommender systems Conclusion Hands-on User Study Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 2/53

Emotions vs Personality Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 3/53

Measuring Emotions and Personality TIPI: I see myself as: 1. Extraverted, enthusiastic. 2. Critical, quarrelsome. 3. Dependable, self-disciplined. 4. Anxious, easily upset. 5. Open to new experiences, complex. 6. Reserved, quiet. 7. Sympathetic, warm. 8. Disorganized, careless. 9. Calm, emotionally stable. 10. Conventional, uncreative. Self-reporting is intrusive There is a need for unobtrusive ways of measuring E&P Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 4/53

Report • N = 42 • 32M, 8F, 2 unknown Ext Agr Con EmS Ope Age Mean 3,61 3,75 5,08 3,27 5,06 28,72 SD 1,51 1,15 1,21 1,44 1,22 4,97 Male norms 21-30 Mean 3,73 4,50 4,57 4,64 5,49 SD 1,54 1,20 1,39 1,46 1,13 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 5/53

Table of Contents Recap Acquisition of Personality Usage of personality in recommender systems Acquisition of Emotions Usage of emotions in recommender systems Conclusion Hands-on User Study Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 6/53

Personality Detection Our online behaviour is influenced by our personality. Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 7/53

Personality Detection Our online behaviour is influenced by our personality. Hence, our traces in social media should reflect our personality. Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 7/53

Personality Detection Our online behaviour is influenced by our personality. Hence, our traces in social media should reflect our personality. It is enough to acquire personality once. • Twitter (Golbeck et al., 2011, Quercia et al., 2011) • Facebook, (Golbeck et al., 2011, Kosinski et al., 2013) • Instagram (Skowron et al., 2016) Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 7/53

Personality from Twitter • N=335 • features • log(num of followers) • log(num of followees) • log(num of times listed) • Klout score (num of clicks, num of retweets) • TIME (twitter and facebook followers) • M5 rules regressor • personality on the scale 1-5 References Quercia, D., Kosinski, M., Stillwell, D., and Crowcroft, J. (2011). Our twitter profiles, our selves: Predicting personality with twitter. In Proceedings - 2011 IEEE International Conference on Privacy, Security, Risk and Trust and IEEE International Conference on Social Computing, PASSAT/SocialCom 2011 (pp. 180–185). IEEE. https://doi.org/10.1109/PASSAT/SocialCom.2011.26 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 8/53

Kosinki - personality from FB • personality prediction from Facebook References Kosinski, M., Stillwell, D., and Graepel, T. (2013). Private traits and attributes are predictable from digital records of human behavior. Proceedings of the National Academy of Sciences of the United States of America, 110(15), 5802–5. https://doi.org/10.1073/pnas.1218772110 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 9/53

Kosinki - personality from FB • Prediction accuracy as Pearson correlation coefficient • all correlations significant at p < 0 . 001 • transparent bars = questionnaire’s baseline accuracy (test–retest reliability) Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 10/53

Kosinki - personality from FB Selected most predictive likes for openness Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 11/53

Personality from Instagram • N=113 (AMT) • 22398 pictures • BFI • features • low-level image features (Hue-Value-Saturation) • filters • presence of people References Skowron, M., Ferwerda, B., Tkalčič, M., and Schedl, M. (2016). Fusing Social Media Cues : Personality Prediction from Twitter and Instagram. WWW’16 Companion, 2–3. https://doi.org/10.1145/2872518.2889368 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 12/53

Linguistic features • (Farnadi et al., 2016) compared Facebook, Twitter and Youtube • small differences between classifiers • major differences between linguistic features: • 81 LIWC features • 66 SPLICE features • 2 SentiStrength features • 14 MRC • 8 NRC • LIWC (Linguistic Inquiry and Word Count) features best predictors References Farnadi, G., Sitaraman, G., Sushmita, S., Celli, F., Kosinski, M., Stillwell, D., . . . De Cock, M. (2016). Computational personality recognition in social media. User Modeling and User-Adapted Interaction, (Special Issue on Personality in Personalized Systems). https://doi.org/10.1007/s11257-016-9171-0 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 13/53

IBM Watson Demo time :-) https://personality-insights-livedemo.mybluemix.net/#your-twitter-panel References Srivastava, Abhishek. "Gray Sheep, Influential Users, User Modeling and Recommender System Adoption by Startups." Proceedings of the 10th ACM Conference on Recommender Systems. ACM, 2016. Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 14/53

Table of Contents Recap Acquisition of Personality Usage of personality in recommender systems Acquisition of Emotions Usage of emotions in recommender systems Conclusion Hands-on User Study Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 15/53

Personality for mood regulation • high on openness, extraversion, and agreeableness more inclined to listen to happy music when they are feeling sad. • high on neuroticism listen to more sad songs when feeling disgusted (neurotic people choose to increase their level of worry) References Ferwerda, B., Schedl, M., and Tkalcic, M. (2015). Personality and Emotional States : Understanding Users ’ Music Listening Needs. In A. Cristea, J. Masthoff, A. Said, and N. Tintarev (Eds.), UMAP 2015 Extended Proceedings. Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 16/53

Personality and music browsing styles • personality is correlated with music browsing styles References Ferwerda, B., Yang, E., Schedl, M., and Tkalčič, M. (2015). Personality Traits Predict Music Taxonomy Preferences. In Proceedings of the 33rd Annual ACM Conference Extended Abstracts on Human Factors in Computing Systems - CHI EA ’15 (pp. 2241–2246). https://doi.org/10.1145/2702613.2732754 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 17/53

Personality and Ratings • movielens • N=1840 References Karumur, R. P., Nguyen, T. T., and Konstan, J. A. (2016). Exploring the Value of Personality in Predicting Rating Behaviors. In Proceedings of the 10th ACM Conference on Recommender Systems - RecSys ’16 (pp. 139–142). New York, New York, USA: ACM Press. https://doi.org/10.1145/2959100.2959140 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 18/53

Personality as user similarity • new user problem • N = 52 • images = 70 • neighborhood-based RS: Euclidian distance − 1 References Tkalčič, M., Kunaver, M., Košir, A., and Tasič, J. (2011). Addressing the new user problem with a personality based user similarity measure. In F. Ricci, G. Semeraro, M. de Gemmis, P. Lops, J. Masthoff, F. Grasso, J. Ham (Eds.), Joint Proceedings of the Workshop on Decision Making and Recommendation Acceptance Issues in Recommender Systems (DEMRA 2011) and the 2nd Workshop on User Models for Motivational Systems: The affective and the rational routes to persuasion (UMMS 2011). Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 19/53

Personality as user similarity • N = 113 • 646 songs • TIPI • user similarities (Pearson CC) • item-based • personality-based References Hu, R., and Pu, P. (2010). Using Personality Information in Collaborative Filtering for New Users. In Proceedings of the 2nd ACM RecSys’10 Workshop on Recommender Systems and the Social Web (pp. 17–24). Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 20/53

Personality and Diversity • within subject N=52 • movies • diversity per: genre, director, country, release time, actor • rules from a previous study • High Level of Openness is linked to high need for diversity w.r.t. —actor/actress • Low Level of Conscientiousness is correlated with high need for the overall diversity References Wu, W., Chen, L., and He, L. (2013). Using personality to adjust diversity in recommender systems. Proceedings of the 24th ACM Conference on Hypertext and Social Media - HT ’13, (May), 225–229. Chen, L., Wu, W., and He, L. (2013). How personality influences users’ needs for recommendation diversity? CHI ’13 Extended Abstracts on Human Factors in Computing Systems on - CHI EA ’13, 829. https://doi.org/10.1145/2468356.2468505 Marko Tkalčič, RecSys2017SummerSchool-Part2-AcquisitionUsage 21/53