People recommender systems are widely used to suggest connections between users in social networks but are becoming pervasive across other areas (e.g., expert finding [SpDe13], education [VaMcGr16], dating [PRAK13], and employment [ATKS15]). Analyses done on Twitter [SuShGo16] and Social Blue [DaGeMi10] data indicate that these recommender systems impact the structure of social networks, favoring individuals who are already popular. In that context, researchers have questioned whether social recommender systems may exacerbate existing disparities [FCBC22; StRiCh18], e.g., increasing the social capital gap between majority and minority groups.
In the absence of recommender systems, previous work explored how the growth mechanisms of social networks (e.g., homophily, rich get richer, and minority majority partitions) can lead to a “glass ceiling” effect where specific groups have less “success” in the network [AKLM15; KGWS18]. This theoretical work resonates with research showing bias against particular groups on platforms like Twitter [NGLD16], Fiverr [HWGM17], and AirBnB [EdLuSv17].
Recommendation algorithms may intensify the disadvantage of minority groups, an effect which [StRiCh18] call the “algorithmic glass ceiling.” Under a specific random walk recommender system and strong assumptions, [StRiCh18] find that differentiated homophily (“two groups exhibiting varying propensity to favor their own peers”) can lead recommender systems to exacerbate the “glass ceiling” effect already present in social networks. The notion of differentiated homophily also helps to explain the apparent contradiction that the glass ceiling can act upon groups that are majorities (women on Instagram).
[FCBC22] introduce a model to simulate the feedback loop between users and contact recommendation algorithms (e.g., who to follow). They find that link recommenders exacerbate “rich-get-richer” effects and provide disparate exposure to minority homophilic groups. This work largely confirms simulation-based results from previous work considering one round of recommendations [FBBC20], and theoretical insights from [StRiCh18].
Empirical evidence measuring the magnitude of the effect of the “algorithmic glass ceiling effect” is lacking — perhaps because it may be difficult to disentangle it from the “natural” glass ceiling effect that can occur as social networks grow [AKLM15]. In that sense, this literature would benefit from empirical experimental or quasi-experimental work that attempts to measure the effects of social recommendation in the wild. Also, proposed solutions to the “algorithmic ceiling” are not deeply explored (although some ideas are mentioned in [StRiCh18] and [KGWS18]) and seem hard to apply in the deep-learning-based recommender systems often employed in large social networks [CoAdSa16]. Upcoming work could also measure the extent to which these approaches can have a real-world impact.
References
[AKLM15] Avin, Chen ; Keller, Barbara ; Lotker, Zvi ; Mathieu, Claire ; Peleg, David ; Pignolet, Yvonne-Anne: Homophily and the glass ceiling effect in social networks. In: Proceedings of the Conference on Innovations in Theoretical Computer Science, 2015
[ATKS15] Almalis, Nikolaos D ; Tsihrintzis, George A ; Karagiannis, Nikolaos ; Strati, Aggeliki D: FoDRA—A new content-based job recommendation algorithm for job seeking and recruiting. In: Procceedings of the International Conference on Information, Intelligence, Systems and Applications : IEEE, 2015
[CoAdSa16] Covington, Paul ; Adams, Jay ; Sargin, Emre: Deep neural networks for youtube recommendations. In: Proceedings of the 10th ACM conference on recommender systems, 2016, S. 191–198
[DaGeMi10] Daly, Elizabeth M ; Geyer, Werner ; Millen, David R: The network effects of recommending social connections. In: Proceedings of the ACM Conference on Recommender Systems, 2010
[EdLuSv17] Edelman, Benjamin ; Luca, Michael ; Svirsky, Dan: Racial discrimination in the sharing economy: Evidence from a field experiment. In: American Economic Journal: Applied Economics Bd. 9 (2017), Nr. 2, S. 1–22
[FBBC20] Fabbri, Francesco ; Bonchi, Francesco ; Boratto, Ludovico ; Castillo, Carlos: The effect of homophily on disparate visibility of minorities in people recommender systems. In: Proceedings of the International AAAI Conference on Web and Social Media, 2020
[FCBC22] Fabbri, Francesco ; Croci, Maria Luisa ; Bonchi, Francesco ; Castillo, Carlos: Exposure Inequality in People Recommender Systems: The Long-Term Effects. In: Proceedings of the International AAAI Conference on Web and Social Media, 2022
[HWGM17] Hannák, Anikó ; Wagner, Claudia ; Garcia, David ; Mislove, Alan ; Strohmaier, Markus ; Wilson, Christo: Bias in online freelance marketplaces: Evidence from taskrabbit and fiverr. In: Proceedings of the ACM Conference on Computer Supported Cooperative Work and Social Computing, 2017
[KGWS18] Karimi, Fariba ; Génois, Mathieu ; Wagner, Claudia ; Singer, Philipp ; Strohmaier, Markus: Homophily influences ranking of minorities in social networks. In: Scientific reports Bd. 8, Nature Publishing Group (2018), Nr. 1
[NGLD16] Nilizadeh, Shirin ; Groggel, Anne ; Lista, Peter ; Das, Srijita ; Ahn, Yong-Yeol ; Kapadia, Apu ; Rojas, Fabio: Twitter’s glass ceiling: The effect of perceived gender on online visibility. In: Proceedings of the International AAAI Conference on Web and Social Media, 2016
[PRAK13] Pizzato, Luiz ; Rej, Tomasz ; Akehurst, Joshua ; Koprinska, Irena ; Yacef, Kalina ; Kay, Judy: Recommending people to people: the nature of reciprocal recommenders with a case study in online dating. In: User Modeling and User-Adapted Interaction Bd. 23, Springer (2013), Nr. 5, S. 447–488
[SpDe13] Spaeth, Alexandre ; Desmarais, Michel C: Combining collaborative filtering and text similarity for expert profile recommendations in social websites. In: Procceedings of the International Conference on User Modeling, Adaptation, and Personalization, 2013
[StRiCh18] Stoica, Ana-Andreea ; Riederer, Christopher ; Chaintreau, Augustin: Algorithmic Glass Ceiling in Social Networks: The effects of social recommendations on network diversity. In: Proceedings of the World Wide Web Conference, 2018
[SuShGo16] Su, Jessica ; Sharma, Aneesh ; Goel, Sharad: The effect of recommendations on network structure. In: Proceedings of the International Conference on World Wide Web, 2016
[VaMcGr16] Vassileva, Julita ; McCalla, Gordon I ; Greer, Jim E: From small seeds grow fruitful trees: How the PHelpS peer help system stimulated a diverse and innovative research agenda over 15 years. In: International Journal of Artificial Intelligence in Education Bd. 26, Springer (2016), Nr. 1, S. 431–447