[Vers des systèmes de recommandation axés sur la nouveauté]
Nous recevons des recommandations à propos de tout et d'une manière omniprésente. Les systèmes de recommandation agissent comme des boussoles pour notre voyage dans des espaces conceptuels complexes, et nous comptons de plus en plus sur les recommandations pour fonder la plupart de nos décisions. Malgré leur efficacité et leur fiabilité extraordinaires, les systèmes de recommandation sont loin d'être parfaits. Ils présentent plutôt de sérieux inconvénients qui pourraient sérieusement réduire notre ouverture d'esprit et notre capacité à faire l'expérience de la diversité et à avoir des opinions contradictoires. Dans cet article, nous examinons attentivement les fondements mêmes des algorithmes de recommandation afin d'identifier les déterminants de ce que pourrait être la prochaine génération de systèmes de recommandation. Nous postulons qu'il est possible de surmonter les limites des systèmes de recommandation actuels en s'inspirant de la façon dont les gens recherchent les nouveautés et valorisent les nouvelles expériences. De ce point de vue, la notion de possible adjacent semble pertinente pour repenser les systèmes de recommandation d'une manière qui s'aligne mieux avec l'inclination naturelle des êtres humains vers des expériences nouvelles et agréables. Nous affirmons que cette nouvelle génération de prescripteurs pourrait aider à surmonter les écueils des technologies actuelles, à savoir la tendance au manque de diversité, à la polarisation, à l'émergence d'écho-chambres et à la désinformation.
We get recommendations about everything and in a pervasive way. Recommender systems act like compasses for our journey in complex conceptual spaces and we more and more rely on recommendations to ground most of our decisions. Despite their extraordinary efficiency and reliability, recommender systems are far from being flawless. They display instead serious drawbacks that might seriously reduce our open-mindedness and our capacity of experiencing diversity and possibly conflicting views. In this paper, we carefully investigate the very foundations of recommendation algorithms in order to identify the determinants of what could be the next generation of recommender systems. We postulate that it is possible to overcome the limitations of current recommender systems, by getting inspiration from the way in which people seek for novelties and give value to new experiences. From this perspective, the notion of adjacent possible seems a relevant one to redesign recommender systems in a way that better aligns with the natural inclination of human beings towards new and pleasant experiences. We claim that this new generation of recommenders could help in overcoming the pitfalls of current technologies, namely the tendency towards a lack of diversity, polarization, the emergence of echo-chambers and misinformation.
Mot clés : Systèmes de recommandation, Nouveautés, Adjacente possible, Zone de confort
Pietro Gravino 1 ; Bernardo Monechi 1 ; Vittorio Loreto 1, 2, 3
@article{CRPHYS_2019__20_4_371_0, author = {Pietro Gravino and Bernardo Monechi and Vittorio Loreto}, title = {Towards novelty-driven recommender systems}, journal = {Comptes Rendus. Physique}, pages = {371--379}, publisher = {Elsevier}, volume = {20}, number = {4}, year = {2019}, doi = {10.1016/j.crhy.2019.05.014}, language = {en}, }
Pietro Gravino; Bernardo Monechi; Vittorio Loreto. Towards novelty-driven recommender systems. Comptes Rendus. Physique, Volume 20 (2019) no. 4, pp. 371-379. doi : 10.1016/j.crhy.2019.05.014. https://comptes-rendus.academie-sciences.fr/physique/articles/10.1016/j.crhy.2019.05.014/
[1] Exposure to ideologically diverse news and opinion on facebook, Science, Volume 348 (2015) no. 6239, pp. 1130-1132
[2] Tweeting from left to right: is online political communication more than an echo chamber?, Psychol. Sci., Volume 26 (2015) no. 10, pp. 1531-1542 (PMID: 26297377)
[3] Novelty, complexity, and hedonic value, Percept. Psychophys., Volume 8 (1970) no. 5, pp. 279-286
[4] https://blogs.cornell.edu/info2040/2012/09/20/last-fm-music-reccomendation-incorporating-social-network-ties-and-collaborative-filtering/ (blogs.cornell.edu. Last.fm – Music Recommendation incorporating social network ties and collaborative filtering)
[5] From hits to niches?: Or how popular artists can bias music recommendation and discovery, Proceedings of the 2nd KDD Workshop on Large-Scale Recommender Systems and the Netflix Prize Competition, ACM, 2008, p. 5
[6] Is seeing believing?: How recommender system interfaces affect users' opinions, Ft. Lauderdale, FL, USA, 5–10 April 2003, ACM, New York (2003), pp. 585-592
[7] Report of the Special Rapporteur on the promotion and protection of the right to freedom of opinion and expression https://www.ohchr.org/EN/Issues/FreedomOpinion/Pages/ContentRegulation.aspx
[8] The spreading of misinformation online, Proc. Natl. Acad. Sci., Volume 113 (2016) no. 3, pp. 554-559
[9] Filter bubbles, echo chambers, and online news consumption, Public Opin. Q., Volume 80 ( 03 2016 ) no. S1, pp. 298-320
[10] For the web https://fortheweb.webfoundation.org/
[11] Unveiling political opinion structures with a web-experiment, COMPLEXIS 2016, Rome, Italy, 22–24 April, 2016, Science and Technology Publications, Lda, Setúbal, Portugal (2016), pp. 39-47
[12] Crossing the horizon: exploring the adjacent possible in a cultural system, Paris, France, 27 June–1 July 2016 (2016)
[13] Will the global village fracture into tribes? Recommender systems and their effects on consumer fragmentation, Manag. Sci., Volume 60 (2013) no. 4, pp. 805-823
[14] What recommenders recommend–an analysis of accuracy, popularity, and sales diversity effects, UMAP 2013, Rome, Italy, 10–14 June 2013 (S. Carberry; S. Weibelzahl; A. Micarelli; G. Semeraro, eds.), Springer (2013), pp. 25-37
[15] Investigations, Oxford University Press, 2000
[16] The Bellkor solution to the Netflix grand prize, Netflix Prize Doc., Volume 81 (2009), pp. 1-10
[17] Collaborative filtering with temporal dynamics, KDD '09, Paris, France, 28 June, 28–1 July 2009, ACM, New York (2009), pp. 447-456
[18] Amazon.com recommendations: item-to-item collaborative filtering, IEEE Internet Comput., Volume 1 (2003) no. 1, pp. 76-80
[19] Content-based recommender systems: state of the art and trends, Recommender Systems Handbook, Springer, 2011, pp. 73-105
[20] Birds of a feather: homophily in social networks, Annu. Rev. Sociol., Volume 27 (2001) no. 1, pp. 415-444
[21] Waves of novelties in the expansion into the adjacent possible, PLoS ONE, Volume 12 (2017) no. 6
[22] Exploring the filter bubble: the effect of using recommender systems on content diversity, WWW '14, Seoul, Korea, 7–11 April 2014, ACM, New York (2014), pp. 677-686
[23] Education of a model student, Proc. Natl. Acad. Sci. USA, Volume 109 (2012) no. 6, pp. 1868-1873
[24] The long tail of destination image and online marketing, Ann. Tour. Res., Volume 38 (2011) no. 1, pp. 132-152
[25] The Filter Bubble: What the Internet Is Hiding From You, Penguin, UK, 2011
[26] Optimal learning paths in information networks, Sci. Rep., Volume 5 (2015)
[27] Item-based collaborative filtering recommendation algorithms, WWW'01, Hong Kong, 1–5 May 2001, ACM, New York (2001), pp. 285-295
[28] Collaborative filtering recommender systems, The Adaptive Web, Springer, 2007, pp. 291-324
[29] President Obama's farewell address https://www.nytimes.com/2017/01/10/us/politics/obama-farewell-address-speech.html
[30] Reinforcement Learning: An Introduction, MIT Press, 2018
[31] The dynamics of correlated novelties, Sci. Rep., Volume 4 (2014), p. 5890
[32] The impact of youtube recommendation system on video views, IMC '10, 1–3 November 2010, ACM, New York (2010), pp. 404-410
Cité par Sources :
Commentaires - Politique