Co-exposure Maximization in Online Social Networks

Sijing Tu
Sijing Tu
Cigdem Aslay
Cigdem Aslay

NIPS 2020, 2020.

Cited by: 0|Bibtex|Views13
EI
Other Links: dblp.uni-trier.de|academic.microsoft.com
Weibo:
We address the problem of maximizing co-exposure in social networks

Abstract:

Social media has created new ways for citizens to stay informed on societal matters and participate in political discourse. However, with its algorithmically-curated and virally-propagating content, social media has contributed further to the polarization of opinions by reinforcing users’ existing viewpoints. An emerging line of research ...More

Code:

Data:

0
Introduction
  • Social media have created new ways for citizens to stay informed and participate in societal discourse.
  • Despite enabling users to access a variery of information, social media has been linked to increased societal polarization [22], by amplifying the phenomenon of echo chambers [4, 27], where users are only exposed to information from like-minded individuals, and of filter bubbles [35, 37], where algorithms only present personalized content that agrees with the user’s viewpoint.
Highlights
  • Social media have created new ways for citizens to stay informed and participate in societal discourse
  • Despite enabling users to access a variery of information, social media has been linked to increased societal polarization [22], by amplifying the phenomenon of echo chambers [4, 27], where users are only exposed to information from like-minded individuals, and of filter bubbles [35, 37], where algorithms only present personalized content that agrees with the user’s viewpoint
  • As a step of addressing the aforementioned challenges, we formally introduce co-exposure maximization (COEM) as the problem of assigning seed sets to each campaign such that the expected number of users co-exposed to campaigns under a stochastic information propagation model is maximized
  • We address the problem of maximizing co-exposure in social networks
  • We further propose TCEM, a scalable instantiation of our approximation algorithm that can efficiently estimate the expected co-exposure
  • We would like to extend our approach to account for the social advertising setting [2] in which the advertisers, with a limited monetary budget, are required to pay a monetary amount to the host for each engagement to their virally propagating campaign
Results
  • Co-exposure results for different networks are shown in Figure 1.
  • For the weighted-cascade model, the three local algorithms that use out-degree information can perform slightly better than TCEM; this happens for instance in WikiVote_wc
  • This result is consistent with the empirical observation that nodes with high out-degree obtain large expected spread under the weighted-cascade model [28], resulting in large expected co-exposure values.
  • The authors observe that the BalanceExposure algorithm is consistently outperformed by all the algorithms in all the settings
Conclusion
  • The authors address the problem of maximizing co-exposure in social networks.
  • The authors show that the problem is NP-hard and the objective function is neither submodular nor supermodular.
  • By exploiting a connection to a submodular function that acts as a lower bound to the objective, the authors devise an approximation algorithm with provable guarantee.
  • The authors further propose TCEM, a scalable instantiation of the approximation algorithm that can efficiently estimate the expected co-exposure.
  • It would be interesting to improve the approximation guarantee for the problem the authors define.
  • The authors would like to extend the approach to account for the social advertising setting [2] in which the advertisers, with a limited monetary budget, are required to pay a monetary amount to the host for each engagement to their virally propagating campaign
Summary
  • Introduction:

    Social media have created new ways for citizens to stay informed and participate in societal discourse.
  • Despite enabling users to access a variery of information, social media has been linked to increased societal polarization [22], by amplifying the phenomenon of echo chambers [4, 27], where users are only exposed to information from like-minded individuals, and of filter bubbles [35, 37], where algorithms only present personalized content that agrees with the user’s viewpoint.
  • Objectives:

    The authors study the problem of allocating seed users to opposing campaigns: by drawing on the equal-time rule of political campaigning on traditional media, the goal is to allocate seed users to campaigners with the aim to maximize the expected number of users who are co-exposed to both campaigns.
  • By drawing on the equal-time rule of political campaigning in the pre-digital era, the aim is to devise a seed-set allocation framework so that the expected number of users who are exposed to both campaigns, through the propagation of information in the social network, is maximized
  • Results:

    Co-exposure results for different networks are shown in Figure 1.
  • For the weighted-cascade model, the three local algorithms that use out-degree information can perform slightly better than TCEM; this happens for instance in WikiVote_wc
  • This result is consistent with the empirical observation that nodes with high out-degree obtain large expected spread under the weighted-cascade model [28], resulting in large expected co-exposure values.
  • The authors observe that the BalanceExposure algorithm is consistently outperformed by all the algorithms in all the settings
  • Conclusion:

    The authors address the problem of maximizing co-exposure in social networks.
  • The authors show that the problem is NP-hard and the objective function is neither submodular nor supermodular.
  • By exploiting a connection to a submodular function that acts as a lower bound to the objective, the authors devise an approximation algorithm with provable guarantee.
  • The authors further propose TCEM, a scalable instantiation of the approximation algorithm that can efficiently estimate the expected co-exposure.
  • It would be interesting to improve the approximation guarantee for the problem the authors define.
  • The authors would like to extend the approach to account for the social advertising setting [2] in which the advertisers, with a limited monetary budget, are required to pay a monetary amount to the host for each engagement to their virally propagating campaign
Related work
  • Our work relates to the emerging line of research on breaking filter bubbles in social media through information-propagation lens. There have been a number of studies on the effects of “echo chambers” [4, 26] and “filter bubbles” [4, 17, 22, 37]. In particular, it has been observed that news stories containing opinion-challenging information spread less than other news [26] and filtering of content by a social-network owner to increase user engagement can significantly increase societal polarization [17]. Recent approaches to breaking filter bubbles focus on making recommendations to individuals of opposing viewpoints [23, 24, 31], targeting users so as to reduce the polarization of opinions and bridge opposing views by considering opinion-formation models [16, 33, 34], or addressing these issues under information-propagation models [3, 25] as we do in our work.

    Aslay et al [3] study the related problem of diversifying exposure to information that is propagating in a social network. Their problem formulation assumes that the leanings of users and news articles are quantified in the interval [−1, 1] and are known. The goal is to find an assignment of articles to seed users to maximize the total diversity over all users in the network. The diversity of a user is defined to be a function that takes as input the leanings of the set of news articles that the user is exposed as well as the learning of the user. Therefore, users who are exposed to only one article, which has a different leaning from their own, still contribute to the value of the diversity objective. Translating this formulation to our setting, by considering two articles with leanings −1 and 1, implies that their objective function can potentially achieve a relatively high value while the co-exposure being equal to 0. Thus, their work does not guarantee that co-exposure is maximized. Moreover, Aslay et al [3] also propose an extension to random reverse-reachable sets [9] for scalable estimation of their objective function. For this task, they sample random sets defined over user-article pairs while our sampling domain is user-user pairs as we explain in Section 5. Due to the difference in the objective functions, hence, the estimation task, the sample-complexity results and the sample of random sets obtained for one problem cannot be used to solve the other.
Funding
  • This research is supported by the Academy of Finland projects AIDA (317085) and MLDB (325117), the ERC Advanced Grant REBOUND (834862), the EC H2020 RIA project SoBigData (871042), and the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by the Knut and Alice Wallenberg Foundation
Reference
  • Cigdem Aslay, Wei Lu, Francesco Bonchi, Amit Goyal, and Laks VS Lakshmanan. Viral marketing meets social advertising: ad allocation with minimum regret. Proceedings of the VLDB Endowment, 8(7):814–825, 2015.
    Google ScholarLocate open access versionFindings
  • Cigdem Aslay, Francesco Bonchi, Laks VS Lakshmanan, and Wei Lu. Revenue maximization in incentivized social advertising. Proceedings of the VLDB Endowment, 10(11):1238–1249, 2017.
    Google ScholarLocate open access versionFindings
  • Cigdem Aslay, Antonis Matakos, Esther Galbrun, and Aristides Gionis. Maximizing the diversity of exposure in a social network. In 2018 IEEE International Conference on Data Mining (ICDM), pages 863–868. IEEE, 2018.
    Google ScholarLocate open access versionFindings
  • Eytan Bakshy, Solomon Messing, and Lada Adamic. Exposure to ideologically diverse news and opinion on facebook. Science, 2015.
    Google ScholarLocate open access versionFindings
  • Nicola Barbieri and Francesco Bonchi. Influence maximization with viral product design. In Proceedings of the 2014 SIAM International Conference on Data Mining, pages 55–63. SIAM, 2014.
    Google ScholarLocate open access versionFindings
  • Nicola Barbieri, Francesco Bonchi, and Giuseppe Manco. Topic-aware social influence propagation models. Knowledge and information systems, 37(3):555–584, 2013.
    Google ScholarLocate open access versionFindings
  • Shishir Bharathi, David Kempe, and Mahyar Salek. Competitive influence maximization in social networks. In International workshop on web and internet economics, pages 306–311.
    Google ScholarLocate open access versionFindings
  • Andrew An Bian, Joachim M Buhmann, Andreas Krause, and Sebastian Tschiatschek. Guarantees for greedy maximization of non-submodular functions with applications. In Proceedings of the 34th International Conference on Machine Learning-Volume 70, pages 498–507. JMLR. org, 2017.
    Google ScholarLocate open access versionFindings
  • Christian Borgs, Michael Brautbar, Jennifer Chayes, and Brendan Lucier. Maximizing social influence in nearly optimal time. In Proceedings of the twenty-fifth annual ACM-SIAM symposium on Discrete algorithms, pages 946–957. SIAM, 2014.
    Google ScholarLocate open access versionFindings
  • Allan Borodin, Yuval Filmus, and Joel Oren. Threshold models for competitive influence in social networks. In WINE, pages 539–550, 2010.
    Google ScholarLocate open access versionFindings
  • Gruia Calinescu, Chandra Chekuri, Martin Pál, and Jan Vondrák. Maximizing a monotone submodular function subject to a matroid constraint. SIAM Journal on Computing, 40(6): 1740–1766, 2011.
    Google ScholarLocate open access versionFindings
  • Parinya Chalermsook, Atish Das Sarma, Ashwin Lall, and Danupon Nanongkai. Social network monetization via sponsored viral marketing. ACM SIGMETRICS Performance Evaluation Review, 43(1):259–270, 2015.
    Google ScholarLocate open access versionFindings
  • Wei Chen, Yajun Wang, and Siyu Yang. Efficient influence maximization in social networks. In Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 199–208, 2009.
    Google ScholarLocate open access versionFindings
  • Wei Chen, Chi Wang, and Yajun Wang. Scalable influence maximization for prevalent viral marketing in large-scale social networks. In Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 1029–1038, 2010.
    Google ScholarLocate open access versionFindings
  • Wei Chen, Alex Collins, Rachel Cummings, Te Ke, Zhenming Liu, David Rincon, Xiaorui Sun, Yajun Wang, Wei Wei, and Yifei Yuan. Influence maximization in social networks when negative opinions may emerge and propagate. In Proceedings of the 2011 siam international conference on data mining, pages 379–390. SIAM, 2011.
    Google ScholarLocate open access versionFindings
  • Xi Chen, Jefrey Lijffijt, and Tijl De Bie. Quantifying and minimizing risk of conflict in social networks. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 1197–1205, 2018.
    Google ScholarLocate open access versionFindings
  • Uthsav Chitra and Christopher Musco. Analyzing the impact of filter bubbles on social network polarization. In Proceedings of the 13th International Conference on Web Search and Data Mining, pages 115–123, 2020.
    Google ScholarLocate open access versionFindings
  • Abhimanyu Das and David Kempe. Submodular meets spectral: greedy algorithms for subset selection, sparse approximation and dictionary selection. In Proceedings of the 28th International Conference on International Conference on Machine Learning, pages 1057–1064, 2011.
    Google ScholarLocate open access versionFindings
  • Abhimanyu Das and David Kempe. Approximate submodularity and its applications: subset selection, sparse approximation and dictionary selection. The Journal of Machine Learning Research, 19(1):74–107, 2018.
    Google ScholarLocate open access versionFindings
  • Samik Datta, Anirban Majumder, and Nisheeth Shrivastava. Viral marketing for multiple products. In Data Mining (ICDM), 2010 IEEE 10th International Conference on, pages 118– 127. IEEE, 2010.
    Google ScholarLocate open access versionFindings
  • Nan Du, Yingyu Liang, Maria-Florina Balcan, Manuel Gomez-Rodriguez, Hongyuan Zha, and Le Song. Scalable inuence maximization for multiple products in continuous-time diffusion networks. The Journal of Machine Learning Research, 18(1):30–74, 2017.
    Google ScholarLocate open access versionFindings
  • Seth Flaxman, Sharad Goel, and Justin M Rao. Filter bubbles, echo chambers, and online news consumption. Public opinion quarterly, 80(S1):298–320, 2016.
    Google ScholarLocate open access versionFindings
  • Kiran Garimella, Gianmarco De Francisci Morales, Aristides Gionis, and Michael Mathioudakis. Factors in recommending contrarian content on social media. In Proceedings of the 2017 ACM on Web Science Conference, pages 263–266, 2017.
    Google ScholarLocate open access versionFindings
  • Kiran Garimella, Gianmarco De Francisci Morales, Aristides Gionis, and Michael Mathioudakis. Reducing controversy by connecting opposing views. In Proceedings of the Tenth ACM International Conference on Web Search and Data Mining, pages 81–90, 2017.
    Google ScholarLocate open access versionFindings
  • Kiran Garimella, Aristides Gionis, Nikos Parotsidis, and Nikolaj Tatti. Balancing information exposure in social networks. In Advances in Neural Information Processing Systems, pages 4663–4671, 2017.
    Google ScholarLocate open access versionFindings
  • R. Kelly Garrett. Echo chambers online?: Politically motivated selective exposure among internet news users. JCMC, 14(2), 2009. ISSN 1083-6101. doi: 10.1111/j.1083-6101.2009. 01440.x. URL http://dx.doi.org/10.1111/j.1083-6101.2009.01440.x.
    Locate open access versionFindings
  • R Kelly Garrett. Echo chambers online?: Politically motivated selective exposure among internet news users. Journal of Computer-Mediated Communication, 14(2):265–285, 2009.
    Google ScholarLocate open access versionFindings
  • David Kempe, Jon Kleinberg, and Éva Tardos. Maximizing the spread of influence through a social network. In Proceedings of the ninth ACM SIGKDD international conference on Knowledge discovery and data mining, pages 137–146. ACM, 2003.
    Google ScholarLocate open access versionFindings
  • Arijit Khan, Benjamin Zehnder, and Donald Kossmann. Revenue maximization by viral marketing: A social network host’s perspective. In 2016 IEEE 32nd international conference on data engineering (ICDE), pages 37–48. IEEE, 2016.
    Google ScholarLocate open access versionFindings
  • Jure Leskovec and Andrej Krevl. SNAP Datasets: Stanford large network dataset collection. http://snap.stanford.edu/data, June 2014.
    Findings
  • Q Vera Liao and Wai-Tat Fu. Expert voices in echo chambers: effects of source expertise indicators on exposure to diverse opinions. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pages 2745–2754, 2014.
    Google ScholarLocate open access versionFindings
  • Wei Lu, Wei Chen, and Laks VS Lakshmanan. From competition to complementarity: Comparative influence diffusion and maximization. Proceedings of the VLDB Endowment, 9(2), 2015.
    Google ScholarLocate open access versionFindings
  • Antonis Matakos, Sijing Tu, and Aristides Gionis. Tell me something my friends do not know: Diversity maximization in social networks. Knowledge and Information Systems, 62(9): 3697–3726, 2020.
    Google ScholarLocate open access versionFindings
  • Cameron Musco, Christopher Musco, and Charalampos E Tsourakakis. Minimizing polarization and disagreement in social networks. In Proceedings of the 2018 World Wide Web Conference, pages 369–378, 2018.
    Google ScholarLocate open access versionFindings
  • Dimitar Nikolov, Diego FM Oliveira, Alessandro Flammini, and Filippo Menczer. Measuring online social bubbles. PeerJ Computer Science, 1:e38, 2015.
    Google ScholarLocate open access versionFindings
  • Naoto Ohsaka and Yuichi Yoshida. Monotone k-submodular function maximization with size constraints. In Advances in Neural Information Processing Systems, pages 694–702, 2015.
    Google ScholarLocate open access versionFindings
  • Eli Pariser. The Filter Bubble: What the Internet Is Hiding from You. Penguin UK, 2011.
    Google ScholarLocate open access versionFindings
  • Ajit Singh, Andrew Guillory, and Jeff Bilmes. On bisubmodular maximization. In Artificial Intelligence and Statistics, pages 1055–1063, 2012.
    Google ScholarLocate open access versionFindings
  • Youze Tang, Xiaokui Xiao, and Yanchen Shi. Influence maximization: Near-optimal time complexity meets practical efficiency. In Proceedings of the 2014 ACM SIGMOD international conference on Management of data, pages 75–86. ACM, 2014.
    Google ScholarLocate open access versionFindings
  • Youze Tang, Yanchen Shi, and Xiaokui Xiao. Influence maximization in near-linear time: A martingale approach. In Proceedings of the 2015 ACM SIGMOD International Conference on Management of Data, pages 1539–1554, 2015.
    Google ScholarLocate open access versionFindings
  • Isabel Valera and Manuel Gomez-Rodriguez. Modeling adoption and usage of competing products. In ICDM, 2015. doi: 10.1109/ICDM.2015.40. URL https://doi.org/10.1109/ ICDM.2015.40.
    Locate open access versionFindings
Full Text
Your rating :
0

 

Tags
Comments