Recent concerns about filter bubbles, fake news and echo chambers are symptomatic of the disruptive potential of digital recommendations, but also demonstrate the need for responsible recommender design. News recommender systems, driven by data and machine learning, automatically select the content of newsletters, personalized news apps, or social media news feeds. Recommender systems can make or break filter bubbles, and as such can be instrumental in creating either a more closed or a more open internet. Much will depend on the design of news recommenders. Are they merely designed to generate clicks and short term engagement? Or are they programmed to balance short term engagement and relevancy with the legitimate longer term interest of users in diverse information and not losing out on important information?
To unlock the full potential of news recommenders, we need to create innovative metrics and measures that help to strike the right balance between short-term engagement and diversity. The key to such an endeavour lies in the effective operationalization of a rather imprecisely-defined societal concept – diversity - into measures and metrics that can actually inform the design of recommenders. The aim of this proposal is therefore to translate research into diverse recommender design done at the University of Amsterdam into novel ways with which publishers and news aggregators can improve the performance of their algorithmic recommenders. We will develop a toolkit to automatically measure and externally judge the diversity of recommendations through the analysis of the recommended content. We will do so in cooperation with RTL and Blendle - an innovative Dutch online news platform that offers access to articles from more than 100 publishers in the Netherlands, Germany and Belgium, and is a forerunner in experimenting with algorithmic recommendations.
This project is made possible with a funding from the SIDN fonds.
- develop automated tools for analyzing content to measure diversity, and help to create a toolkit that automatically annotates media content on these measures;
- to this end, cooperate closely with the team from Blendle and RTL, and the researchers from the University of Amsterdam (media, communications science, data science) to operationalize several dimensions of diversity (e.g. topic diversity, frame diversity, discourse diversity, representativeness, etc.);
- experiment with the dimensions of diversity combined with different recommender metrics and develop a set of key performance indicators (KPIs);
- do real-life testing of the models developed;
- help to co-oordinate workshops and (international) outreach activities.
You should have:
- an advanced degree (e.g., master or PhD degree) in data science, or communication or other social science with outstanding data mining and analysis skills;
- active research knowledge of, and expertise in, modern search and text mining methods;
- experience in programming in Python (2.7/3.x);
- experience with Natural Language Processing (NLP) (e.g. Topic Modelling), Machine Learning experience is a plus;
- excellent oral and written communication skills;
- team spirit, willingness, and commitment to do autonomous work in a multidisciplinary team, including communication scholars, legal scholars, and data scientists.
Database handling experience, Git skills, Bash Scripting are considered pros.
The position preferably starts between 1 July and 1 September 2018. It concerns, in principle, a one-year period after an initial test period.
The monthly gross salary, depending on your knowledge and experience, will range between €2,588 and €4,757 (Scale 10/11) based on a full-time basis.
We offer a pension scheme, a holiday allowance of 8% per year, and an end of year allowance of 8,3% per year and flexible employment conditions. Our conditions are based on the Collective Labour Agreement for Dutch Universities.
For further information, please contact:
To apply for this position, please send at least the following documents in English by email to firstname.lastname@example.org:
- a motivation letter, including a brief discussion about how you see your fit with the topic and objectives of the RPA;
- a curriculum vitae, including publication list and overview of academic activities and achievements thus far (e.g., conference visits, courses taken, awards);
- an academic writing sample in English (e.g., journal article, PhD dissertation) and
- a list of names that could be consulted as a reference (no reference letters required).
Please send all your documents in PDF or doc(x) format. We will consider only complete applications.
The deadline for applications is 31 May 2018 at 24:00 PM CET. Please state vacancy number 18-218 in the subject line of your application.