EuADS Summer School – Data Science for Social Media

We are pleased to confirm that the Summer School is going ahead as an in-person event.

The European Association for Data Science organises a

Summer School on Data Science for Social Media

Wednesday June 15th to Friday June 17th 2022 in Kirchberg, Luxembourg.

The summer school is preceded by a public opening event on Tuesday June 14th 2022.

Please note a change in the summer school programme: Instead of Fionn Murtagh’s talk “
Qualitative and Quantitative Data Analytics in Data Science, with Correspondence
Analysis and Clustering.” Eyke Hüllermeier will give a talk titled “Learning user preferences from social media data”

Social media is ubiquitous in our modern world. Its ubiquity makes it an attractive object of study in various fields. Applications start with social science, over economics, computer science and go even to health science and disaster control.

On the other hand the validity of such data has been repeatedly brought into question and, given the rich portfolio of approaches and ideas this field, is certainly in danger of overpromising.

Regardless of these concerns, social media analysis will be an important method in future research in many different fields. Discussing its potentials and limitations, transfer of methods and ideas, is what this summer school is about.

The Summer School is primarily aimed at advanced PhD students, postdoctoral and early-career researchers with an interest and basic grounding in data science, machine learning, and/or statistics. It is also aimed at data science professionals working with social media data.

Public Event

The summer school will be preceded by a public event on Tuesday, June 14th starting 13:00 p.m.

At the heart of the public event is the Sabine-Krolak-Schwerdt-Lecture, in memoriam of EuADS’ founding president. This will be held by Mireille Hildebrandt (Vrije University Brussels, Belgium)

Agenda

13h00 – 14h00 Registration and Coffee
14h00 – 15h00 Opening and Welcome
Marc Hansen
Minister Delegate for Digitalisation
Peter Flach
EuADS President
15h30 – 17h00 Sabine Krolak-Schwerdt Public Lecture
The Whisper Challenge – The Rule of Proxies
Mireille Hildebrandt
Professor at Vrije University Brussels, Belgium
17h00 Welcome Reception

The Symposium on Tuesday is
free, but a registration is required:
contact@euads.org

The Whisper Challenge - The Rule of Proxies in Data Science

Mireille Hildebrandt (Vrije Universiteit Brussel)

To develop a learner algorithm we must tell the algorithm what and/or how to learn, which goes for supervised (labelling), reinforcement (machine-consumable goal) and unsupervised (choice of the training data, objective function). This requires a dedicated effort in translation, transforming relevant natural language concepts that inform the specific data science quest into variables, datasets and/or mathematical functions. The act of translation results in the use of proxies that stand for whatever concept we aim to operationalise. As machine learning and data science cannot work with human concepts one could say the domain is ruled by proxies and the output of an ML system will necessarily depend on the choice of the proxies. Subsequently the output must be rendered, explained and often ‘sold’ in human language, again requiring an act of translation. In this lecture I will investigate to what extent the rule of proxies is a game of ‘Chinese whispers’ and how this may affect the reconfiguration of real life challenges.

Speaker's Website

Topics and Presenters

The schedule for the summer school including speakers and topics:

Wednesday,
June 15th
9:30 a.m.
to 1 p.m.
Static and Dynamic Mapping Method for Uncovering Competitive Positions Bernd Skiera
(Goethe U Frankfurt, D)
  2:30 p.m.
to 6 p.m.
The evolution of digital behavioral trace and survey data in social networks Christoph Stadtfeld
(ETH Zürich, CH)
Thursday,
June 16th
9:30 a.m.
to 1 p.m.
Social media metrics: definitions and applications Zohreh Zahedi
(U of Leiden, NL, PGU University, IR)
  2:30 p.m.
to 6 p.m.
Learning user preferences from social media data Eyke Hüllermeier
(LMU Munich, D)
Friday,
June 17th
9:30 a.m.
to 1 p.m.
Responsible social-media based collective intelligence Eirini Ntoutsi
(FU Berlin, D)

For details see below!

Wednesday, June 15th

9:30 a.m. to 1 p.m.

2:30 p.m. to 6 p.m.

Static and Dynamic Mapping Method for Uncovering Competitive Positions

Prof Dr Bernd Skiera (Goethe University Frankfurt, Germany)

A market map provides managers with a static snapshot of the competitive positions of a market’s participants (such as products or brands). Today, most markets are rather large (e.g., comprising hundreds of products so that a comprehensive visualization of competitive market structures can be cumbersome and complex. Yet, reduction of the analysis to smaller representative product sets can obscure important information. The first part of the workshop outlines (i) data sources to derive consideration sets of consumers that reflect competition between products and (ii) approaches (e.g., building upon social network analysis) that integrate these data into a modeling and mapping approach to visualize competition in large markets and to identify distinct submarkets.

Yet, as markets tend to be in flux, knowledge about the trajectories of competitive positions of market’s participant over time would be more informative than a static snapshot. In contrast to static snapshots, trajectories create a forward-looking perspective on competition, reveal whether positions are converging or diverging and help managers evaluate the impact of their positioning efforts. Although data for market structure analysis is increasingly available in high frequency (see part 1 of the workshop), extant mapping methods are exclusively static, and do not reveal market participants’ trajectories. Therefore, I focus in part 2 of the workshop on dynamic mapping method that generate a sequence of cohesive maps that enable market analysts to track the trajectories of competitive positions over time.

Speaker's Website

The evolution of digital behavioral trace and survey data in social networks

Prof Dr Christoph Stadtfeld (ETH Zürich, Switzerland)

The increasing availability of digital behavioral trace (DBT) data promises novel social science studies that simultaneously scale up on a large number of study participants and zoom in on fine-grained individual behavioral actions. These data may, for example, stem from social media platforms, social sensor experiments, or wearable technologies such as smart phones or watches. DBT data offer a seemingly objective perspective on how people behave individually and socially – how they eat, sleep, travel, interact, socialise, and date. DBT network data often come in the form of relational events – time-stamped dyadic observations that can be represented as time-ordered edge lists. Several new models for the statistical analysis of relational events have been proposed over the past years. The first part of the course will cover the statistical analysis of DBT data with relational event models.

Studies that merely rely on DBT data have some obvious blind spots. Individual behavior is to a large extent based on how individuals perceive their environment, their relationships, and themselves. Such perception data can be well collected through traditional surveys. Survey data also have known challenges such as cognitive burdens, necessary time investments by participants, and measurement biases. Traditional social network data often stem from surveys and may represent who individuals perceive as friends, whom they like or dislike, and whom they trust. Dynamic network data collected through surveys can, for example, be statistically analysed with stochastic actor-oriented models. These models will be briefly discussed in the second part of the course.

Speaker's Website

Thursday, June 16th

9:30 a.m. to 1 p.m.

2:30 p.m. to 6 p.m.

Social media metrics: definitions and applications

Dr Zohreh Zahedi (University of Leiden, Persian Gulf University)

Social media metrics (altmetrics) refers to metrics derived from social media platforms (such as Facebook, Twitter, Wikipedia, mainstream news websites, etc.). These metrics offer possibility of studying the relations and interactions between social media users, scholarly contents, and different actors. Altmetrics data aggregators provide access to social media metrics differ in terms of methodological choices in collecting, updating, tracking, and reporting metrics. This course focuses on defining and interpreting social media metrics, data possibilities and challenges, social media metrics data analysis and their uses and applications.

Speaker's Website

Learning user preferences from social media data

Prof Dr Eyke Hüllermeier (LMU Munich)

The topic of “preferences” has attracted increasing attention in artificial intelligence and machine learning in the recent past, where it has emerged as an interdisciplinary research field with close connections to operations research, social choice, and the decision sciences. Roughly speaking, preference learning is about methods for learning preference models from explicit or implicit information about the preferences of an individual or a group of individuals, and the use of such models for predicting preferences in new situations. Approaches relevant to this field range from learning special types of preference models, such ranking functions, to collaborative filtering techniques for recommender systems. The goal of this talk is to provide a brief introduction to the field of preference learning and, moreover, to elaborate on the use of preference learning in the context of social media, notably for learning user preferences from social media data.

Speaker's Website

Friday, June 17th

9:30 a.m. to 1 p.m.

Responsible social-media based collective intelligence

Prof Dr Eirini Ntoutsi (FU Berlin, Germany)

The Web offers enormous benefits for information sharing, collective organization and distributed activity with great impact in all areas of our lives. However, along with the benefits come also negative consequences like hate speech, fake news, surveillance, etc. Ambivalences lie at the heart of the Web and we must deal responsibly with these ambivalences to amplify the benefits and counter the negative effects. Towards this direction, as data scientists we should work towards responsible analysis of data collected via the Web. While we all agree that the huge amounts of data generated in the Web offer paramount opportunities for data-science related applications and are the pre-condition for the success of modern machine learning methods, we cannot ignore the fact that data collection comes with assumptions, and moreover, further assumptions are made during the analysis pipeline which of course have great impact on the extracted knowledge. In this talk, we will focus on such assumptions, including data sampling, redundancies, proxy-labeling, temporality and bias), their effect on the learning process and how to build effective models under such assumptions.

Speaker's Website
 

Fees and Registration

For EuADS members the fee for participating in the summer school is 250€. 100€ if they are PhD students.

For non-members the fee is 300 €. 150 € if they are PhD students. In both cases it includes a one-year EuADS membership for free.

To ensure an interactive experience the number of participants is limited, so early registration is strongly recommended. Please register by

1. Sending an email with your personal details to contact@euads.org, with reference to EuADS Summer School 2022 Data Science for Social Media.

2. Transferring the amount to the Banque et Caisse d’Epargne de l’Etat, Luxembourg (BIC: BCEELULL; IBAN: LU47 0019 4655 6967 1000).

Once the personal details and registration fee have been received, you will receive an email confirming your participation.

Your place can only be guaranteed if your email reaches us by 31st of May, 2022.

Pandemic

Many activities have gone digital in the past 1,5 years to an extend previously considered impossible. Some activities have, on the other hand, proven difficult to digitalise. Effective networking and informal exchange of ideas, a central part of the format of a summer school, is one such activity.

For that reason we conduct the summer school as an in-person event under the Luxembourg COVID19 regulations.

We consider it important that the speakers of the summer school are present on the venue side and available to questions and discussion e.g. during coffee breaks and similar. Should some speakers, however, be unable to travel to Luxembourg, we will make every effort for a remote talk to the plenum on the conference venue.

Venue:

Conference and Training Centre at the Chambre de Commerce Luxembourg
7, Rue Alcide de Gasperi
L-2981 Luxembourg Kirchberg

Accommodations:

From the following hotels you can walk to the location:
Meliá Luxembourg
Coque Hôtel
Hôtel Novotel Luxembourg Kirchberg
Sofitel Luxembourg Europe Hotel

Organisers:

  • Serge Allegrezza (STATEC, Luxembourg; EuADS Treasurer)
  • Matthias Böhmer (U Luxembourg, Luxembourg)
  • Reinhold Decker (U Bielefeld, Germany; EuADS Vice-President)
  • Andreas Geyer-Schultz (KIT, Germany)
  • Nils Hachmeister (U Bielefeld, Germany; EuADS Vice-President)
  • Marc Pauly (STATEC, Luxembourg)
  • Denise Schroeder (STATEC, Luxembourg)
  • Myra Spiliopoulou (U Magdeburg, Germany)

Sponsors