Encouraging Diversity- and Representation-Awareness in Geographically Centralized Content

Eduardo Graells-Garrido / Telefónica R&D ChileMounia Lalmas / Yahoo, UKRicardo Baeza-Yates / UPF, Catalonia / Univ. Of Chile

“Las cuatro reinas de Chile”Ian Pierce, 2014

Twitter is global.

But we also know that there are cognitive and systemic biases that shape the behavior of users.

What is the effect of those biases and what can we do about it?

Leetaru et al., 2013.

Our context: Chile, a centralized country.Economic/political/media powers are concentrated in Santiago (the capital).

Región Metropolitana (RM) is the capital region.

Twitter activity is centralized - RM receives more tweets from other locations than expected due to population distribution.

Research Questions

Does centralization affect how people perceive information, and how people behave when browsing informational content in micro-blogging platforms?

If so, how can we encourage non-centralized exploration?

Our Proposal

To find if centralization affects how people perceive timelines, create a geographically diverse timeline.

● Proposed Method “PM”: Information Entropy + Sidelines (enforces location).● Baseline “DIV”: Information Entropy only.● Baseline “POP”: Most popular tweets (mostly tweets from Santiago/RM).

After reading timelines side-by-side, which one is more:

- diverse?- interesting?- informative?

Participants answeredusing a Likert scalefrom -3 to 3.

(for detailed results on all variables see the paper)

Being from a central or peripheral location makes a difference.

For peripheral/NOT-RM users, there was no perception of the diversity present by design on the algorithm!

Experimental SetupSnowball SamplingWithin-subjectsN = 125 (RM = 87, NOT-RM = 38)Ordinal Logistic RegressionModel includes statistical interactions

Main ResultStatistical interaction between location and condition POP/PM.RM participants find PM more diverse than POP (OR = 3.17).NOT-RM do not.

Users do not see the diversity in the timelines because they cannot identify themselves (in the location sense), even though they are present.

There is a diversity and representation awareness problem.

How to make users aware of their representation in the timeline, as well as the diversity inherent in it?

Previous work indicates that clustered representations help users to become aware of the diversity in news aggregators.

We follow that approach. But in previous work the number of clusters has been small - 2 or 3. In our case, we have 15 clusters!

How to depict 15 clusters without introducing positional bias on the screen?

Clustered Tweets by Location Standalone Tweets

Inspired by newsmap.jp, we use treemaps to depict differences in a tweet’s geographical origin, as well as giving every location a balanced amount of exposure.

We also allow users to filter locations by selecting a specific region. Doing so will show only tweets about the specified location.

“In the wild” study

Purpose - to evaluate user involvement with the application as proxy of diversity and representation awareness.

Diversity - do users click on content related to different locations?

Representation - do users choose to see only their location using the filters?

Interestingness - how many interactions with content do users make?

We used a social bot (@todocl) to generate timelines every hour and broadcast them, mentioning featured users, and retweeting their tweets.

This allowed us to get users and spread the word.

Experimental SetupBetween-subjects design.N = 321 (RM = 193, NOT-RM = 128)Negative Binomial and Logistic Regressions

Main Results

treemap increases:- # of interaction events

OR = 1.87- # of locations interacted with

OR = 2.08- filter likelihood

OR = 2.13

Users interacted with more content, from more locations, and filtered locations also! (diversity)

Being from RM:- increases locations interacted with

OR = 2.47- decreases filter likelihood

OR = 0.61

* NOT-RM increases representation awareness - they find themselves!

Discussion / What To Take Away

Centralization has effects on information perception and user behavior.

Algorithms are not enough! We need to find new ways of showing information to users (not necessarily new techniques - but new contexts).

Clustered representations work to enhance diversity awareness - but how to display clusters depends on cultural and individual differences.

Final Words

Not every culture has the same notion of relevance and importance in content. Even within a country there are differences. Let’s make algorithms and user interfaces aware of them!

Future Work

Test other visualization-based UIs. Do only hierarchical visualizations work?

Qualitative Evaluation. Why does visualization work in this case? Is it more engaging? Does it enable a different perception of information?

Replication. Chile is one country and it is very different to others. What about other centralized countries?

Thank you!Do you have any questions?

Contact Us:eduardo.graells@telefonica.com

@carnby

Reproduce Our Experiments!The source code of the project (Twitter crawler, django application, filtering algorithm and visual interface) is available at:

https://github.com/carnby/aurora

Acknowledgements

Dany Pajarito, Denis Parra, Luz Rello, Sergio Salgado.

Background

Presentation of Diversified Information (in biased scenarios)

- [Park et al., 2009] NewsCube - clustered representations make people aware of diversity in information in terms of media bias.

Diversification of Timelines (from an information perspective)

- [De Choudhury et al., 2013] Information-Entropy diversification of timelines.

- [Munson et al., 2009] Sidelines - filter news by enforcing sides to be present.

Cultural Differences

- [García-Gavilanes et al., 2014] People from different countries have different social behavior.

Treemaps to Visualize News

- [Weskampf, 2004] newsmap.jp uses treemaps to visualize news headlines.

Our context: Chile, a centralized country.Economic/political/media powers are concentrated in Santiago (the capital).

Región Metropolitana (RM) is the capital region.

Chart: flow of mentions and RTs between administrative regions.

This affects Web users as content is not geographically diverse (mostly related to/from Santiago). Content from other locations is hidden and hard to find.

(I was at WWW when I searched for this. “Everywhere” displays relevant tweets from Santiago only.)