Open menu

Do you know how much time you spend on social media? Do you still remember your posts from ten years ago - and would seeing such a post perhaps bring back some dear memories? What topics do you post most about and how are these related?

If you are a Facebook user like I am, you might have posted something on Facebook about 1-2 times every week, with predictable peaks during holiday seasons, Christmas, conferences and other periods in which you travel, and see family, friends or colleagues. There are probably also some obvious recurring events, such as birthdays - and particular themes that interest me - such as vegetarian cooking, dogs and user modeling.

To find out my activities in the (recent and more distant) past that I found worthwhile to post on Facebook, I can scroll through my timeline. However, scrolling through that virtually endless timeline would be very time-consuming.  Alternatively, Facebook offers me the possibility to download all my data (post, comments, photos) in HTML or JSON format. However, if you have tried to do that before, you know that these files are almost just as useless as simply scrolling through the timeline.


Visualizing my Facebook posts and their relations

What I would like to do is to look for particular keywords (such as 'dog' or 'christmas') and see which other keywords, dates and persons are connected to this keyword. Together with two colleagues from Hof University of Applied Sciences (Bavaria, Germany), Claus Atzenbeck and Daniel Roßner, we developed scripts to translate my posts into a browsable graph. And this is a video of me browsing through the result, using their visualization tool 'Mother'.


What can you see in the video above? Via the term "Hypertext" I am reminded of the collocated UMAP and Hypertext 2016 conferences in Halifax, where I presented a paper and had a hike with Peter Brusilovsky. The hike with Peter is thematically connected to the many walks in Hildesheim, Germany, where I used to live and now in the Ooijpolder near Nijmegen, the Netherlands. Another thing that I apparently often do is congratulating friends with their birthdays, which then - via the keyword 'cake' - reminded me of the several rounds of cake that I served my colleagues to celebrate yet another award for the master Information Sciences, of which I am coordinator.

This all might seem obvious to reconstruct, but it requires quite some tweaking and deciding what thematic and temporal connections are important. Posts that are thematically too similar (in terms of words) are often too obviously related (e.g. birthday wishes with birthday wishes), so we also put in some temporal dimensions (e.g. posted in the same period, or at the same hour of day, or day of week), until at some point I observed that I typically post day reflections and photos during the early evening.

Once we had a visualization that was meaningful for me, we asked three further people to download their Facebook data and to interact with the visualization. Similar to me, they were reminded about ongoing themes and memorizable events from the past. The tool also prompted participants to think about "what happens with a Facebook profile after one's death".

What do our Facebook posts and comments tell about ourselves?

Most of my posts on Facebook still made sense to me, as I typically deliberately decide to share something with my friends or the world - for example announcements, beautiful pictures, life events, and holidays. In other words, my posts reflect how I want others to perceive me online.

In contrast, my commenting behavior turned out to be more diverse and far less planned or deliberated than my posting behavior. Many of my comments were simply brief comments to friends' posts ("Same for me", "Funny", "Get well soon"). Other comments could be classified as 'rants', invoked by a particular friends' post, a page post or a discussion in a Facebook group. Apparently, cycling in Germany - particularly in Hildesheim - was a topic that often caused me to write long comments.

Both Daniel and I observed that the tone in our comments often was more humorous (or meant to be humorous) than in our Facebook posts. Daniel said at some point: "Apparently, I find myself quite funny when commenting on Facebook." Perhaps our Facebook comments better reflect who we are and how we behave. They probably also better reflect our responses triggered by particular topics, events, photos or advertisements. But we also had the feeling that most of these comments were not relevant anymore and perhaps in hindsight even a bit embarrassing.

This leads to the question: what is the most 'honest' online Facebook identity? Is that my carefully constructed history of self-presentation via my posts, or would it be the more diverse and less polished set of my Facebook comments? I believe that my comments are a more honest reflection, but at the same time I would prefer you to get to know me based on my posts instead.

facebook fotos

This blog post is based on and inspired by the following article: Eelco Herder, Daniel Roßner and Claus Atzenbeck. Structuring and Exploring User Behavioral Patterns in Social Media Traces. Proc. UCAI'20 - Workshop on User-Centered Artificial Intelligence at Mensch und Computer 2020, presented on 9 September 2020.

eelcoherder 256px

Privacy Engineering, User Modeling, Personalization, Recommendation, Web Usage Mining, Data Analysis and Visualization, Usability, Evaluation

Dr. Ir. Eelco Herder
Associate Professor
Human-Centered Computing Group
Information and Computing Sciences
Utrecht University

Buys Ballotgebouw
Princetonplein 5
3584 CC Utrecht
The Netherlands