Abstract
We analyze peer influence of production and consumption of projects in the Scratch community, an online platform developed by MIT Media Lab, where users collectively learn to program by creating and sharing projects. Scratchers can follow others’ activities on the platform; in the followers network, we investigate if Scratchers’ production popularity (determined by others) and consumption preference (self determined) are influenced by whom they follow on the platform (peers). Several mechanisms established in the literature like homophily, selection, peer influence, own behavioural tendency, reciprocated ties, and particular contexts can lead to observations of behavioural clustering in a social network like Scratch, and therefore isolating peer influence from other mechanisms is a challenging task. In this study, we measure peer influence in the Scratch community after accounting for such alternative confounding mechanisms. There are two key steps we follow to estimate peer influence of a behaviour. First, at a given time, we create experimental and control groups such that the peers’ behaviour under investigation can be justified as a random assignment. To do so we exactly match Scratchers’ personal and network attributes in both groups such that Scratchers in the experimental group have peers with higher degree of the behaviour under study compared to the control group, and all other attributes of Scratchers are balanced across both groups. Second, conditional on all activities up to this time (as captured by the attributes), we measure peer influence as the difference in Scratchers’ personal behavioural changes in subsequent periods across the two groups.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Various observable measures (love-it, download, and comment) convey information about popularity of a project. These measures are correlated because they are determined, often at the same time, after a project is viewed. Favorites count is not observable as a consumption count on project page, multiple comments can be made on a project by a single user, and downloads count has data issues (the count is supposed to be one per user, but multiple count was found for some users). So we choose love-it as our measure of popularity: one user can love any project once only.
- 2.
Major sources are identified by communities of projects that are consumed together. This is described later in the study.
- 3.
It is very unlikely that a Scratcher remembers the exact history of previous activities of his peers. Markov nature of decision making is a very plausible assumption in the scenario of Scratch community, and has been widely adopted in the social networks literature [27].
- 4.
The definition of treatment, peers’ behavioural state at t, follows from the Markov nature: it is a measure that summarizes peers’ behaviour upto t and neglects the historical pattern of its evolution.
- 5.
Empirically significant confounders can be determined by their statistical significance in logistic regression of treatment variable on personal and peers’ characteristics.
- 6.
It is important to note that a Scratcher can know which projects his peers are favoriting via activity feed, however the projects which receive the favorite clicks do not show such counts on the project page. The love-it counts (and all other forms of consumption except favorites) on the other hand are shown on the project pages, and is public information; this forms the difference between favorites and love-its.
- 7.
Matching exactly, especially with \(N_i^t\) variables was found to be very costly, and so only few variables were used. Post-matching balance of covariates is however not compromised.
- 8.
- 9.
We use favorites because peers’ favorites are visible as activity feeds.
- 10.
We performed communities detection using other algorithms as well, for example, we found 171 communities using fast greedy algorithm [7]. The results that we discuss are independent of the choice of algorithm.
- 11.
Bipartite projection is done prior to peer influence analysis and is conceptually independent from such analysis. It is a method used solely to cluster projects and identify major sources. Filtering the projection for weights more than 2 is done for computational ease and is inconsequential for the peer influence analysis for consumption behaviour.
- 12.
Scratch-Wiki: https://en.scratch-wiki.info/wiki
References
Aiello, L.M., Barrat, A., Schifanella, R., Cattuto, C., Markines, B., Menczer, F.: Friendship prediction and homophily in social media. ACM Trans. Web 6(2), 9:1–9: 33 (2012)
Aral, S., Muchnik, L., Sundararajan, A.: Distinguishing influence-based contagion from homophily-driven diffusion in dynamic networks. PNAS 106(51), 21544–21549 (2009)
Aral, S., Walker, D.: Tie strength, embeddedness, and social influence: a large-scale networked experiment. Manag. Sci. 60(6), 1352–1370 (2014)
Atkinson, M.D., Fowler, A.: Social capital and voter turnout: evidence from saint’s day fiestas in mexico. Br. J. Polit. Sci. 44(1), 41–59 (2014)
Blondel, V.D., Guillaume, J.L., Lambiotte, R., Lefebvre, E.: Fast unfolding of communities in large networks. J. Stat. Mech. Theory Exp. 2008(10), P10008 (2008)
Centola, D.: The spread of behavior in an online social network experiment. Science 329(5996), 1194–1197 (2010)
Clauset, A., Newman, M.E.J., Moore, C.: Finding community structure in very large networks. Phys. Rev. E 70(6), 066,111 (2004)
Eckles, D.: Identifying peer influence effects in observational social network data: an evaluation of propensity score methods. Technical Report, Stanford University (2010)
Eckles, D., Bakshy, E.: Bias and high-dimensional adjustment in observational studies of peer effects. Technical Report, MIT (2010)
Feld, J., Zölitz, : U.: Understanding peer effects - on the nature, estimation and channels of peer effects. Technical Report No ROA-RM-2016/1. University, Maastricht (2016)
Hallinan, M.T., Williams, R.A.: Students’ characteristics and the peer-influence process. Sociol. Educ. 63(2), 122–132 (1990)
Hill, B.M., Monroy-Hernández, A.: A longitudinal dataset of five years of public activity in the scratch online community. Sci. Data 4, 170002 (2017)
Ho, D., Imai, K., King, G., Stuart, E.: Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Polit. Anal. 15(3), 199–236 (2007)
Huckfeldt, R., Sprague, J.: Networks in context: the social flow of political information. Am. Polit. Sci. Rev. 81(4), 1197–1216 (1987)
Imai, K., Keele, L., Tingley, D.: A general approach to causal mediation analysis. Psychol. Methods 15(4), 309–334 (2010)
King, G., Nielsen, R.: Why propensity scores should not be used for matching. Technical Report, Harvard University (2016)
Kramer, A.D.I., Guillory, J.E., Hancock, J.T.: Experimental evidence of massive-scale emotional contagion through social networks. PNAS 111(24), 8788–8790 (2014)
Lewis, K., Gonzalez, M., Kaufman, J.: Social selection and peer influence in an online social network. PNAS 109(1), 68–72 (2012)
Manski, C.F.: Identification of endogenous social effects: the reflection problem. Rev. Econ. Stud. 60(3), 531–542 (1993)
Mason, W., Watts, D.J.: Collaborative learning in networks. PNAS 109(3), 764–769 (2012)
McPherson, M., Smith-Lovin, L., Cook, J.M.: Birds of a feather: homophily in social networks. Annu. Rev. Sociol. 27(1), 415–444 (2001)
Newman, M.E.J.: Mixing patterns in networks. Phys. Rev. E 67(2), 026,126 (2003)
Reihaneh, R., Elatia, S., Takaffoli, M., Zaïane, O.R.: Collaborative learning of students in online discussion forums: a social network analysis perspective. Educational Data Mining: Applications and Trends, pp. 441–466. Springer, Cham (2014)
Sacerdote, B.: Peer effects with random assignment: results for dartmouth roommates. Q. J. Econ. 116(2), 681–704 (2001)
Shalizi, C.R., Thomas, A.C.: Homophily and contagion are generically confounded in observational social network studies. Sociol. Methods Res. 40(2), 211–239 (2011)
Snijders, T.A.: Statistical models for social networks. Annu. Rev. Sociol. 11(37), 131–53 (2011)
Snijders, T.A., van de Bunt, G.G., Steglich, C.E.: Introduction to stochastic actor-based models for network dynamics. Soc. Netw. 32(1), 44–60 (2010)
Steglich, C., Snijders, T.A.B., Pearson, M.: Dynamic networks and behavior: separating selection from influence. Sociol. Methodol. 40(1), 329–393 (2010)
Tingley, D., Yamamoto, T., Hirose, K., Keele, L., Imai, K.: Mediation: R package for causal mediation analysis. J. Stat. Softw., Articles 59(5), 1–38 (2014)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
A Appendix
A Appendix
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Samantray, A., Riccaboni, M. (2019). Peer Influence in Large Dynamic Network: Quasi-experimental Evidence from Scratch. In: Aiello, L., Cherifi, C., Cherifi, H., Lambiotte, R., Lió, P., Rocha, L. (eds) Complex Networks and Their Applications VII. COMPLEX NETWORKS 2018. Studies in Computational Intelligence, vol 813. Springer, Cham. https://doi.org/10.1007/978-3-030-05414-4_24
Download citation
DOI: https://doi.org/10.1007/978-3-030-05414-4_24
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-05413-7
Online ISBN: 978-3-030-05414-4
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)