Young people perceptions of content personalisation on social media: a user study

Abstract

The way in which social networking sites are constructed —providing personalised interfaces through recommendation systems— disrupts how users access and process information as well as, more fundamentally, the processes that form their opinions. In this context, we take on an interdisciplinary perspective mixing media studies, science and technology studies and computer science to examine the relevance of the "filter bubble" theory. This theory asserts that customizing the recommendation and classification of content based on users' past behaviour reduces the proportion of conflicting opinions that are visible to them, thus facilitating the creation of so called “echo chambers” (Sunstein, 2009) or “filter bubbles” (Pariser, 2011).However, several studies invite us to question the validity of this theory (Bruns, 2019; Dahlgren, 2021; Dubois & Blank, 2018; Moeller and Helberger, 2018; Mutz and Young, 2011; Zuiderveen Borgesius et al., 2016). More than trying to identify specific bubbles or quantify them, our work aims at understanding how teenagers and young adults perceive these mechanisms. To understand the relationship between these systems and young users (15-25 years old), we have analysed the discourses of teachers, media experts and young users on this topic gathered during 21 interviews, 19 focus groups and 13 in praxis interviews in Brussels. Our results show that young users have various degrees of understanding of the different mechanisms operating for each platform. If most perceive filtering mechanisms to some extent such as content classification on their Instagram feeds, others don’t and feel that they will find “more diverse” information by scrolling through social networking sites than by watching television or listening to the radio. The fact that the workings of these technologies are completely known to them also brings up fantasized representations such as cross-platform data transfers (for example how watching something on Youtube will influence what appears on other platforms). When it comes to evaluating these personalisation mechanisms, young users have conflicted views. If a few denounce them and feel literally trapped “in a bubble”, most do not have strong opinions and some even enjoy customizationpossibilities which allow them to enjoy the content they crave. Teens then also highlight the possibilities to influence the content offered to them. They develop strategies and tactics (de Certeau, 1990) intending to guide the platforms in personalising content. These are not all aimed at diminishing bubble effects, as some try to accentuate them. This can range from subscribing to, or unsubscribing from, different types of accounts on the platforms, liking content on purpose or directly providing information to the algorithms through feedback the modules of the platforms. These results contribute to deepen the understanding of the roles of young users within those socio-technical infrastructures.

Full text

Young people perceptions of content personalisation on social media:
a user study
Arnaud Claes
Marie Dufrasne
Sylvain Malcorps
Thibault Philippette
Victor Wiard

ALG-OPINION, a project funded by Innoviris
o2017 : Call from Innoviris around artificial intelligence, digitalisation and
automation.Proposal by USL-B and UCLouvain on the notion of the filter bubble
popularised by the activist Eli Pariser in 2012.
oUnsatisfaction with this theory because we wanted to:
▪Avoid an approach exagerating the potency of the algorithm and downplaying the agency of the
user to focus on the relationshipbetween the users and the technical system.
▪Assess (empirically) news personalisationas part of a broader media ecosystem.
oThe project is built around a triple objective:
▪To observe and analyse the informational practices of young people on social networks and collect
their opinions on the influence of recommender systems. [WP1]
▪To develop a media education activity allowing to improve critical thinking around digital
practices involving recommender systems. [WP2]
▪To evaluate a recommendation platform that is supposed to give an intelligible interface and allow
more autonomy of choice for the user (the "innovation" side of the project). [WP3]

The Project Design
WP0
Coordination
WP4
Valorisation
WP1
Social evaluation of the effects of
recommander systems on young
people's opinion
WP3
Technical evaluation of the
effectsof recommander systems
and design of educational tools
WP2
Information literacy education for
active research online
Improved uses
Improved design practices
Today's presentation

Young people and
the algorithmic imaginary

Methods: data collected
Protocol 1: Focus on the "algorithmic imaginary" (Bucher, 2017)
I.e.: ways of thinking about what algorithms are, should be,
function, …
•27 interviews (9 teachers, 9 experts, 9 teenagers or young adults)
•19 Focus groups (4 to 7 teenagers between 15 and 18 y.o./FG)
•Brussels, Belgium
Protocol 2: Online in praxis interviews
I.e.: invite young adults to perform simple tasks on their smartphone
and confront their imagined social media practices to their actual practices
•N = 15 young adults (18-21 y.o.)
•Online only (due to the COVID-19 situation)
•Brussels, Belgium

Research questions
(1) What are the teenagers, young adults, teachers and experts' perceptions of the
agency of social media platforms on their everyday lives? (dangers,
challenges, opportunities, ...)
(2) How do teenagers and young adults perceive the impact of social media platforms on
their own opinion-making processes and social media practices?
(3) How do teenagers and young adults evaluate personalisation mechanisms on social
media platforms and act towards it?

Results : Divergent discourses about algorithms (RQ1)
Teachers:
•Dystopian discourses (cyberharassment, conspiracy theories, public health issue, ...)
•Paradoxical discourses and clichés
Teenagers and young adults:
•Utopian discourses (love for technology, a way to be interconnected, to stay together)
« Quand on clique sur une vidéo, ça va analyser ce que tu regardes, et ça va nous recommander
ce genre de vidéos. Et ça te dérange ça ?Pas vraiment, je me dis que Youtube doit savoir nos goûts
afin de nous proposer des vidéos qu’on aime. »
•Alienating discourses (Resignation: “What can I do?”)
« Sur Facebook c'est cramé à 100%, je vais chercher des billets d'avion sur internet et je vais avoir
plein de promos sur la destination que j'ai choisie et même des articles qui parlent de ça, c'est un
truc de dingue. Sur Instagram carrément aussi. »
•Clichés
Experts:
•More nuanced: refer to the context, demonstrate a reflexive stance, are relativistic

Results: From "filter bubbles" (Pariser, 2011)
to intermingled and porous bubbles (RQ2)
•Existing but limited effects of
recommender systems on the
opinion-making process
(see Moeller & Helberger, 2018)
•Complex media ecosystem made
of platforms, peers, relatives,
technologies, authority figures
(teachers, social media
influencers, etc.), non-and-legacy
medias, etc.
•--> Socio-technical network with
"weak links" through which a
diversity of news and opinions
pass (active-passive attitudes)
•IT is this whole ecosystem that
impacts the opinion-making
process of teenagers and young
adults

Results : Teenagers and young adults have an implicit knowledge of
personalisation mechanisms (recommendation algorithms) on social media
platforms (RQ3)
•Conscious that the content that is presented to them is tailored to their activities (likes,
subscriptions, networks or comments)
•Aware that the operations they perform online have a consequence on the recommended content
and sponsored messages that they see: "I checked those shoes this morning and now, they are
suggesting them to me, as if by chance. I think that by now, anyone knows that it’s not by accident"
•The personalization of the social media feeds can sometimes be considered as :
oBeneficial: "[On Instagram]the stories that are at the end, I never watch them. Because it’s a waste of time. Well, not a
waste of time but I only look at essential things from my friends. For example, often, on Instagram, they post the stories
of my close friends first, so I see them first”
oAnnoying: "I always see the same things; it becomes a bit… annoying"
oEven intrusive: "[About online advertising and the marketing of products] I’m really under the impression that all the
stores selling articles that when you click on the pictures, there is a link, they are much more highlighted than photos of
ordinary people on Instagram"
--> Develop tactics to temper or accentuate the effects of the algorithm on their news feed

Results : Main tactics to temper or accentuate the effects of
algorithm on their news feed (RQ3)
•"Follow/Unfollow" and "Like/Dislike" tactics:
Increase and/or decrease the diversity of information they receive by clicking
•"Feeding the algorithm" tactics:
Intentionally inform the platform to show more or less of specific content categories
Intentionally under or over consume specific content over time (mainly on Youtube and TikTok)
•"Avoidance" tactics:
Avoid using the technology, either by not using an app or a device (even temporarily)
•BUT : little knowledge of the settings section of the social media platforms they use

Recommender System:
Autonomy by Design

Coming soon: Alveho

Coming soon: Alveho

In summary
(1) What are the teenagers, young adults, teachers and experts' perceptions of the agency of social media
platforms on their everyday lives? (dangers, challenges, opportunities, ...)
•Relative and general awareness
(2) How do teenagers and young adults perceive the impact of social media platforms on their own opinion-
making processes and social media practices?
•The impact of social media platforms on the opinion-making process needs to be understoon in
light of the broarder media ecosystem the young adults evolve in
(3) How do teenagers and young adults evaluate personalisation mechanisms on social media platforms and
act towards it?
•Via simple but yet powerful tactics
Triple diversity
- Diversity of use (platforms, technologies, routines)
- Differences between users
-Diversité des acceptions de ce qu’est l’information, l’actualité, les médias, le journalisme

Conclusion
oSynthétiser ici les éléments essentiels des deux parties de présentation (pratique des
jeunes et développement de la plateforme)
oAborder les perspectives du WP2 sur l'éducation à la construction d'un écosystème
informationnel en ayant meilleure conscience des mécanismes de recommandation.
("Towards an information literacy education for active research online")
•Triple diversity
•- Diversity of use (platforms, technologies, routines)
•- Differences between users
•-Diversité des acceptions de ce qu’est l’information, l’actualité, les médias, le
journalisme

•Diapo "any questions" + contacts

Bibliographie
•Cardon, D. (2015). A quoi rêvent les algorithmes : nos vies à l’heure des big data, Paris, Le Seuil.
•Claes, A., & Philippette, T. (2020). Defining a critical data literacy for recommender systems: A media-grounded
approach. Journal of Media Literacy Education, 12(3), 17-29. https://doi.org/10.23860/JMLE-2020-12-3-3
•Claes, A. et al. (2021). Algorithmes de recommandation et culture technique : penser le dialogue entre éducation et design.
tic&société, 15(1).https://doi.org/10.4000/ticetsociete.5915
•Dauphin, F. (2012). Culture et pratiques numériques juvéniles: Quels usages pour quelles compétences?. Questions vives.
Recherches en éducation,7(17), 37-52.
•Dufrasne, M. (2018). Etat des lieux de la participation citoyenne en Région bruxelloise. Rapport de recherche en suivi de la
Journée d’étude sur la démocratie participative, Parlement francophone bruxellois, Bruxelles, 23 janvier 2014.
•Fastrez, P. (2010). Quelles compétences le concept de littératie médiatique englobe-t-il ? Une proposition de définition
matricielle. Recherches en Communication, (33), « Les compétences médiatiques des gens ordinaires (I) », 35-52.
•Fastrez, P., & Philippette, T. (2017). Un modèle pour repenser l’éducation critique aux médias à l’ère du numérique.
tic&société, 11(1), 85-110.
•Mercenier, H., Wiard, V., Dufrasne, M. (2021) Teens, social media and fake news: A user’s perspective. In López-García, G.
Politics of disinformation : the influence of fake news on public sphere.
•Pariser, E. (2011). The filter bubble: What the Internet is hiding from you. Penguin UK.

Follow us
https://www.algopinion.brussels/ @AlgoCurieux

Slides additionnels
-Sur Facebook, je ne sais pas combien j'ai d'amis mais je ne vois certainement pas tout. Et ça te dérange ou ça ne te
dérange pas? Euh, oui, euh, non, après je pense que le concept me dérange. Le concept qu'on aiguille un peu tout ce
que je fais mais après est-ce que ça me dérange vraiment? Pas tant que ça.
-Je clique sur l’onglet politique et à chaque fois en fait j'ai des différents articles de différents médias. Par exemple là
j'ai Le Soir Bruxelles, la police intervient au parc du cinquantenaire, OK et quand tu cliques dessus ça arrive
directement sur l'article.
-Je sais... Instagram, c'est vraiment un monde où, entre guillemets, on va gratter des likes. Et donc tu vas aller liker
juste avant de poster une photo.
-(pour voir des choses sur le foot) Faudrait que je me mette à liker ce genre d'information là et puis la commenter et
la partager.
-Youtube, ça fonctionne comme ça non ? Enfin, en fonction de si on regarde la vidéo en entier ou pas. Sur Youtube on
peut voir si la vidéo, je l'ai déjà vu ou pas. Des fois je vais regarder la vidéo une 2e fois pour bien montrer à Youtube
que oui la vidéo je l'ai vu et que donc je veux qu'on propose ce genre de ce genre de contenu donc.

Results : Teenagers and young adults have an implicit knowledge of personalisation mechanisms
(recommendationalgorithms) on social media platforms (RQ3)
Activité
Actions
Opérations
S’informer
Suivre ou aimer (une personne,
une page, un post, un hashtag)
Cliquer (Tap)
Commenter
Double clics
Partager (sur sa page ou par
messages directs)
Dérouler
Publier (textes, photos, vidéos,
stories)
Zoomer (avant ou arrière)
Décider de ne pas suivre ou
aimer ou décliner une invitation
Ne rien faire
Effectuer une recherche
Voir / regarder

Les jeunes et les bulles de filtres : Connaissance implicite du fonctionnement de la
personnalisation (et qui fluctue selon les individus) et sentiments ambivalents envers les
algorithmes
Les effets des filtres et des réseaux sociaux sur les opinions des jeunes sont atténués par :
-Les éléments contextuels (i.e. l’influence du contexte familial, des ami-e-s, des proches et des
enseignant-e-s hors ligne)
-Les pratiques de recherches actives des jeunes
-L'écosystème médiatique global (i.e. la place des médias "traditionnels" dans l'accès à
l'information)
Les écosystèmes informationnels ne se limitent pas aux réseaux socio-numériques, les bulles sont
donc socio-techniques plutôt que simplement technologiques.
Il y a une connaissance des grandes questions d’actualité et des enjeux (élections, Notre Dame de
Paris, attentats, marches pour le climat)

Résultats : Diversité des usages (d’un jeune à l’autre)
•«YouTube je l'utilise comme source d'information et comme divertissement. (…) Si j'ai quelque chose
à faire et que je dois me renseigner, j'irai taper une recherche, mais si c'est un jour de repos, j’irai voir
ce qu'on me recommande. »
•«Les stars aussi ont beaucoup d’influence sur Instagram. Des fois des stars partagent des photos
quand il y a des événements, je regarde si ça m’intéresse. Je regarde aussi un peu les comptes, il y a
des comptes d’actualité sur Instagram. »
•«Sur Instagram j’ai des amis, la famille, et des stars. Pour la nouvelle Zélande j'avais appris via Bella
Hadid [ndlr : 24 millions d’abonnés]. Je sais pas si vous connaissez, elle est mannequin. »