The Human Factor in Digital Recommender Systems

1
Andrew Moore- School of Information Management. Viewing Types - Viewing Type-Gauging Interest - Viewing Type- Familiar Content Finding Aids -Finding Aid: Synopsis/Description - Finding Aid: Serendipity - Finding Aid: Visual Elements - Finding Aid: Positive Review, Non-Personal - Finding Aid: Positive Review, Personal - Finding Aid: Negative Review, Non-Personal - Finding Aid: Negative Review, Personal - Finding Aid: Common Values/Interests - Finding Aid: Thematic/Genre - Finding Aid: Common Talent - Finding Aid: Feeling/Emotional Context External Factor Effects - External Factors: Positive - External Factors: Negative - User Time Commitment User Experience - Negative Indicator: Personal Recommendation - Positive Indicator: Personal Recommendation - Negative Experience with Netflix - Positive Experience with Netflix - Recommender System: Positive - Recommender System: Negative Categories & Codes Future Investigation: - Delving deeper into viewing types and external factors. - Wider Array of interview participants. - Longer, more robust questionnaire - Focus Groups, to allow for additional depth of information. - Deeper investigation into users experiences with trust/belief. Points of Interest: - No code is strongly correlated with satisfaction & recommender systems. - Top three types of Finding Aids: i. Finding Aid: Common Thematic/Genre ii. Finding Aid: Common Talent iii. Finding Aid: Positive Review, Personal - The element of trust/belief is prevalent. Interviewees were more willing to accept recommendations if they felt they could trust the recommendation came without ulterior motives. - Most participants had differing ideas about how Netflix creates its recommendations for them. The Question: - What do Users think of Netflix recommender systems? Could research into their methods of search and opinions on the current system of recommendations provide new insights? - Current Scholarship discusses Recommender Systems primarily from Algorithmic/Systems- based standpoint Methods/Investigation: - 8 semi-structured interviews were planned and conducted within the School of Information Management. - Interviews asked participants 6 questions. - Interviewees discussed interactions with recommender systems, and criteria they use to determine value in recommendations. - A Coding Schema of 22 Codes in 4 separate categories was applied. Results: - Participants discussed opinions on recommender systems, the nature of serendipity in algorithmic recommendations and the makings of a good/bad recommendation. - Additional discussion took place on the nature of Netflix recommendations and how participants make content selection decisions References: Adomavicius, G., Tuzhilin, A. (2005). Towards the next generation of recommender systems: A survey of the state-of-the-art and possible extensions. IEEE Transactions on Knowledge and Data Engineering. 17(6), 734-749. DOI: 10.1109/TKDE.2005.99 de Gemmis, M., Lops, P., Semeraro, G., Musto, C. (2015). An investigation on the serendipity problem in recommender systems. Information Processing and Management, 51. 695-717. DOI: http://dx.doi.org/10.1016/j.ipm.2015.06.008 Konstan, J.A., Reidl, J. (2012). Recommender systems: from algorithms to user experience. User Modelling and User-Adapted Interaction 22(1). 101-123. DOI: 10.1007/s11257-011-9112-x Núñez-Valdéz, E.R., et. al. (2012). Implicit feedback techniques on recommender systems applied to electronic books. Computers in Human Behavior 28(2012), 1186-1193. DOI: 10.1016/j.chb.2012.02.001 Ortega, F., Bobadilla, J., Gutierrez, A. (2013). Incorporating group recommendations to recommender systems: Alternatives and performance. Information Processing and Management, 49, 895-901. Retrieved from: http://dx.doi.org/10.1016/j.ipm.2013.02.003 Park, D.H., Kim, H.K., Choi, I.Y., Kim, J.K., (2012). A literature review and classification of recommender systems research. Expert Systems with Applications 29(2012) 10059-10072. DOI: 10.1016/j.eswa.2012.02.038 Perugini, S., Goncalves, M.A., Fox, E.A. (2003). Recommender systems research: A connection-centric survey. Journal of Intelligent Information Systems, 23(2). 107-143. Retrieved from: http://link.springer.com/article/10.1023%2FB%3AJIIS.0000039532.05533.99 Victor, P., Cornelis, C., De Cock, M., Pinheiro da Silva, P. (2009). Gradual trust and distrust in recommender systems. Fuzzy Sets and Systems, 160(2009), 1367-1382. DOI: 10.1016/j.fss.2008.11.014 ”If I do take a recommendation seriously from somebody, I…take into consideration who the person is giving the recommendation is, and what I know of their tastes”. “In a setting like Netflix..I’m more looking to see the randomized things that show up on the screen”. “If I’m tired, or burnt out on schoolwork, then I watch something I’ve watched a million times before, so I don’t have to concentrate on it” “I do not like superhero movies; I’ve never watched one in my life, let alone on Netflix. But I’ve been emailed by them at least a couple times, just.. “Don’t forget! The Avengers is on Netflix”. “But I definitely do look at the front picture, whatever it is” “Making a Murderer is not related to Unbreakable Kimmy Schmidt” “I’m..more interested in reading another book that made me feel the same way, or had the same level of plot twist, or had the same depth of characters, and all these kind of things”. “For me it’s more word of mouth than it is ratings. I feel like the rating system doesn’t usually reflect my interests”.

Transcript of The Human Factor in Digital Recommender Systems

Page 1: The Human Factor in Digital Recommender Systems

Andrew Moore- School of Information Management.

Viewing Types- Viewing Type-Gauging Interest- Viewing Type- Familiar Content

Finding Aids-Finding Aid: Synopsis/Description- Finding Aid: Serendipity- Finding Aid: Visual Elements- Finding Aid: Positive Review, Non-Personal- Finding Aid: Positive Review, Personal- Finding Aid: Negative Review, Non-Personal- Finding Aid: Negative Review, Personal- Finding Aid: Common Values/Interests- Finding Aid: Thematic/Genre- Finding Aid: Common Talent- Finding Aid: Feeling/Emotional Context

External Factor Effects- External Factors: Positive- External Factors: Negative- User Time Commitment

User Experience- Negative Indicator: Personal Recommendation- Positive Indicator: Personal Recommendation- Negative Experience with Netflix- Positive Experience with Netflix- Recommender System: Positive- Recommender System: Negative

Categories & Codes

Future Investigation:- Delving deeper into viewing types and external factors.- Wider Array of interview participants.- Longer, more robust questionnaire- Focus Groups, to allow for additional depth of information. - Deeper investigation into users experiences with trust/belief.

Points of Interest:- No code is strongly correlated with satisfaction & recommender systems.- Top three types of Finding Aids:i. Finding Aid: Common Thematic/Genreii. Finding Aid: Common Talentiii. Finding Aid: Positive Review, Personal- The element of trust/belief is prevalent. Interviewees were more willing to accept recommendations if they felt they could trust the recommendation came without ulterior motives. - Most participants had differing ideas about how Netflix creates its recommendations for them.

The Question: - What do Users think of Netflix recommender systems? Could research into their methods of search and opinions on the current system of recommendations provide new insights?- Current Scholarship discusses Recommender Systems primarily from Algorithmic/Systems-based standpoint

Methods/Investigation:- 8 semi-structured interviews were planned and conducted within the School of Information Management. - Interviews asked participants 6 questions.- Interviewees discussed interactions with recommender systems, and criteria they use to determine value in recommendations.- A Coding Schema of 22 Codes in 4 separate categories was applied.

Results:- Participants discussed opinions on recommender systems, the nature of serendipity in algorithmic recommendations and the makings of a good/bad recommendation. - Additional discussion took place on the nature of Netflix recommendations and how participants make content selection decisions

References:Adomavicius, G., Tuzhilin, A. (2005). Towards the next generation of recommender systems: A survey of the state-of-the-art and possible extensions. IEEE Transactions on Knowledge and Data Engineering. 17(6), 734-749. DOI: 10.1109/TKDE.2005.99de Gemmis, M., Lops, P., Semeraro, G., Musto, C. (2015). An investigation on the serendipity problem in recommender systems. Information Processing and Management, 51. 695-717. DOI: http://dx.doi.org/10.1016/j.ipm.2015.06.008Konstan, J.A., Reidl, J. (2012). Recommender systems: from algorithms to user experience. User Modelling and User-Adapted Interaction 22(1). 101-123. DOI: 10.1007/s11257-011-9112-xNúñez-Valdéz, E.R., et. al. (2012). Implicit feedback techniques on recommender systems applied to electronic books. Computers in Human Behavior 28(2012), 1186-1193. DOI: 10.1016/j.chb.2012.02.001Ortega, F., Bobadilla, J., Gutierrez, A. (2013). Incorporating group recommendations to recommender systems: Alternatives and performance. Information Processing and Management, 49, 895-901. Retrieved from: http://dx.doi.org/10.1016/j.ipm.2013.02.003Park, D.H., Kim, H.K., Choi, I.Y., Kim, J.K., (2012). A literature review and classification of recommender systems research. Expert Systems with Applications 29(2012) 10059-10072. DOI: 10.1016/j.eswa.2012.02.038Perugini, S., Goncalves, M.A., Fox, E.A. (2003). Recommender systems research: A connection-centric survey. Journal of Intelligent Information Systems, 23(2). 107-143. Retrieved from: http://link.springer.com/article/10.1023%2FB%3AJIIS.0000039532.05533.99Victor, P., Cornelis, C., De Cock, M., Pinheiro da Silva, P. (2009). Gradual trust and distrust in recommender systems. Fuzzy Sets and Systems, 160(2009), 1367-1382. DOI: 10.1016/j.fss.2008.11.014

”If I do take a recommendation seriously from somebody, I…take into consideration who the person is giving the recommendation is, and what I know of their tastes”.

“In a setting like Netflix..I’m more looking to see the randomized things that show up on the screen”.

“If I’m tired, or burnt out on schoolwork, then I watch something I’ve watched a million times before, so I don’t have to

concentrate on it”

“I do not like superhero movies; I’ve never watched one in my life, let alone on Netflix. But

I’ve been emailed by them at least a couple times, just.. “Don’t forget! The Avengers is on

Netflix”.

“But I definitely do look at the front

picture, whatever it is”

“Making a Murderer is not related to Unbreakable

Kimmy Schmidt”

“I’m..more interested in reading another book that made me feel the same way,

or had the same level of plot twist, or had the same depth of characters, and all

these kind of things”.

“For me it’s more word of mouth than it is ratings. I feel like the rating system doesn’t usually

reflect my interests”.