Academy Projects 2015

The following projects have been accepted for the Academy Assistants 2015.

Learning a (programming) language from mistakes

Although making mistakes is usually associated with negative emotions, making mistakes is by no means “evidence of failure to learn”. Actually, making mistakes may enhance the learning process and might even be essential for an effective learning process. This project focusses on learning from mistakes a new language, be it a foreign human natural or a programming one. Although there are many differences between these two types of languages, both are characterised by sets of rules that constrain the use of certain expressions or forms. Not being familiar with all the rules, learners will make mistakes. The projects sets out to explore possibilities to develop an efficient automatic training system to assist students in learning a new (programming) language. The system uses a more refined manner of assessing students’ knowledge and understanding, that extends beyond traditional assessment tools, with the ultimate goal to turn mistakes into positive learning events. Read more…


Natalia Silvis-Cividjian

Monique Lamers

Monique Lamers

Peer and company influence on consumer responses and brand perceptions in company social network sites

This project focuses on an analysis of a large MySQL database that contains the content of interactions on company or brand Facebook pages, collected through the Facebook API. The database is made available by Social Embassy, a leading Dutch social media agency, and contains the profiles of 2 million consumers and their interactions with to 35k brand posts of over 50 brands: 7 million likes and 2.5 million comments. Read more…


Peter Kerkhof


Aart van Halteren

The idea is to design a graph based prediction system for social networks where users comment on brand pages. The objective is to analyze and visualize users’ comments in a graph ​-based fashion. We hypothesise it is possible to predict users’ comments based on the graph’s structure.
Firstly, we build a graph representation of users commenting on pages. Nodes are brands pages, edges represent users cross ​posting over the nodes. Then, we consider the structure of users’ cross ​-comments across pages. In particular, we highlight some highly connected sets of pages (cliques) that show a good amount of cross- ​posting activity. We outline a comments
prediction system based on cliques, we claim it can predict the evolution of a clique in terms of users comments. Furthermore, we describe how the system can be validated using historical data.
The hypothesis is how users present in a clique q can be used to predict future comments within the same clique. Specifically, users that are in q and have commented on some pages of q are expected to comment on pages where they have not commented yet. The validation of such system can be done by considering historical data. Given the cliques at time t 0, for each clique we generate the predictions and we compare them against what we observe at time t 1. This is possible since the data has timestamps and spans over 3 years of social network activities.

Ethical Implications of Self-Improving Intelligent Robots

There is substantial concern in the general public and even some experts about the perceived threat of artificial intelligence (AI) running amok, improving itself to outstrip human intelligence and negating any possibility of shutting it down. This proposal sets out a research project to investigate and test the safety and controllability of such ‘smart’ AI, in particular in intelligent robots.
This project aims to investigate the technical possibilities and philosophical implications of ensuring that general AI is safe and controllable. Read more…


Jacqueline Heinerman


Scott Robbins

Our project started with very broad discussions on the interface of Artificial Intelligence and Philosophy. These discussions led to main conclusion that robot designers use many philosophical concepts to design their robots.Robots have ‘knowledge’ and ‘understanding’. They will have their own ‘motivations’ and ‘desires’, which will influence their autonomous ‘choices’. These words to describe contemporary robotics come with much philosophical baggage. This makes the debate about the ethical implications of these robots extremely difficult. For example, robots generating and acting upon their own desires indeed sounds like it could be a major ethical problem; however, what does the design of a robot that generates its own desires look like? How does a ‘desire’ get translated into code? Scott summarised these misconceptions as three epistemological gaps (Ethicist’s Gap, Roboticists’s Gap and the Public Gap) and presented his paper at the International Conference on Social Robotics (ICSR). When we established the three epistemological gaps, our main focus was describing an experiment that could bridge the gap between roboticists and ethicists. The experiment should pursue the following goals: (1) use a key scenario from ethics to describe the choices a robot could have, (2) have a simple scenario to inform the philosophers on the current techniques used for self-learning robots and (3) give insights into the ethical implications of self-improving robots.
We converged for the experiments to the classic Tunnel Problem scenario where the robot is a self-driving car and learns through evolutionary mechanisms. The car has to choose between saving the passenger and overrun a child or evade and kill the passenger by falling off the road. This experiment explains how a self-learning robot can be programmed and how it can learn. While this scenario can already call for many discussions, we extended our experiments to a two-car scenario including the dynamics of the well-known prisoners dilemma. We are in the process off analysing these results and thereafter submit the paper to a journal.

Identifying implicit stereotypical view in natural language through automatic linguistic analyses

Stereotypical views about social groups (e.g., Muslims, Germans, women, immigrants) play a pervasive role in how we perceive and interact with people. Stereotypes emerge from the way we communicate about categorized people and their behavior both privately and in the media. It is valuable to learn about the exact (linguistic) means through which such (negative) stereotypical views become shared within communities.
The present project aims to merge complementary approaches, and thereby develop a methodology for using NLP1 to automatically identify content and strength of stereotypes about specific groups shared within communities. Read more…


Camiel Beukeboom


Antske Fokkens

The goal of our NIAA project was to develop annotation guidelines to identify different types of implicit bias in natural language. In order to develop these guidelines, we first created an online survey, asking participants to write about gender differences in a controlled context. Social psychological control questions were included to test for biased attitudes. This is one of the first studies to integrate all aspects of implicit bias together in one dataset, creating an important resource. Based on this dataset, we developed an innovative codebook. For instance, we developed a category label model that is the first attempt to classify different types of labels of social categories. We furthermore improved an existing model of abstraction in behavior descriptions. Most insights were gained during the development of these annotation guidelines. Ideally, these guidelines translate insights gained in social and communication psychology to clear, linguistic guidelines that could eventually be used to implement an automatic system. The discussions of these guidelines revealed fundamental differences in perspectives in both fields, which has led to important insights that might not have been discovered in a project with only one of the disciplines.
As the development of the guidelines required substantial discussions, we were only able to annotate a small portion of the collected data set. On the basis of this, two publications are planned:

  • Analysis of dataset using annotation guidelines against tested bias (survey)
  • A report on the insights gained in developing the guidelines.This project laid the groundworks for further research, such as applying the annotation guidelines different data containing different social categories and implementing an automatic system.


Modelling Perspectives in Philosophy: A Computational Experiment on Quine’s Word & Object

In Modelling Perspectives a MA student in philosophy and a MA student in computational linguistics take the first steps toward developing a sound method to extract and interpret information about perspectives as expressed in philosophical texts in a computational way. We investigate appropriate calibration of an already existing model combining sentiment/opinion mining and event factuality (Van Son et al. 2014) for application to philosophical texts. We focus on an English corpus including one of the most influential philosophical books of the 20th century, Word & Object (1960) by W. V. O. Quine. Read more…


Arianna Betti


Piek Vossen

Cascades and Avalanches in Twitter Communication Networks

The project Cascades and Avalanches in Twitter Communication Networks focuses on the structure and content of communication networks on Twitter, with the aim of identifying mechanisms that lead to sudden information cascades and social avalanches (hypes). We approach hypes as mass synchronization of attention, for example, when specific topics (as identified by shared hashtags) hype on Twitter. In particular, we will test models of preference synchronization in social networks with Twitter data. We will observe the dynamics of the changing communication networks via cooccurring hashtags, targeted users and changing content on Twitter. Read more…


Lina Hellsten


Ines Lindner

Building an Intelligent System to Reduce Impulsive Snacking by Providing Tailored and Contextualised Feedback

To control the “obesity epidemic”, health messages promote intentions to reduce unhealthy snacking. While general messages often fail to influence intentions, tailoring messages to the needs of individuals can increase healthy intentions1. Unfortunately, our “toxic environment” full of calorically-dense foods confronts people with temptations, making them forget their healthy intentions. Consequently, people often act on impulse and make unhealthy snack choices. This project therefore investigates whether an intelligent system can reduce unhealthy snacking by providing nudges/notifications via a mobile phone app tailored towards the person’s needs, time of the day and his/ her physical environment (location). Read more…


Michel Klein


Guido van Koningsbruggen

The purpose of our study was to design an app that would help individuals in reducing their unhealthy snacking behaviors. The original plan was to test the app in a 2 (Context: Location-Aware vs. Random) X 2 (Message: Tailored vs. Random) between-participants experiment. We conducted a pre-test to determine the spots where people mostly got their unhealthy snacks. These were the spots where we would place beacons, which would enable the app to send messages at the right time.
Unfortunately, due to difficulties during the programming phase, we were not able to finish the app in the indicated time frame. At this moment, we do have a functioning app. However, due to the upcoming summer break we were not able to test on larger scale. Therefore we revised our research question, and decided to run a feasibility study. For this we will get 20 employees from the VU to test the app for seven days. Using this data we will write a protocol paper, so that future research can elaborate on the work that we have done so far. At the end of July, we hope to get our protocol paper published. Despite the fact that we were not able to test our original research question, we are glad to have a fully functioning app, and we hope that future research will continue our work.

Treatment of information security tools under the Wassenaar Arrangement

This proposal seeks to evaluate the treatment of information security research tooling under the Wassenaar Arrangement. As a result of developments outlined below, many tools that the information security research community commonly produces and uses, and that are readily distributed amongst the members of this community are now, or will soon be, in scope of the export control regime maintained as a result of the Wassenaar Arrangement. Not surprisingly, the Arrangement has led to much concern among hackers, security experts and scientists alike. Read more…


Arno Lodder


Herbert Bos

The Wassenaar Arrangement is a multilateral export control regime that involves over 40 participating states, including the United States, Russia, EU countries (and, thus, The Netherlands) and tries to find methods and structures to control the export of military and dual-use goods.
In December 2013, the Wassenaar Arrangement added new controls that aim to limit the export of “intrusion software”, making it more difficult for oppressive regimes to import these goods. In reaction to these new controls, the information technology industry and information security researchers raised their concerns about the overly broad reach of the “intrusion software” rules. Examples of these concerns include the controls having a chilling effect on innovation and operational ability to respond to security issues.
The goal of the project “Treatment of information security tools under the Wassenaar Arrangement” was to gather insight into the historical and present day context of the Wassenaar Arrangement, assess the new controls regarding “intrusion software” and propose changes that would alleviate the concerns regarding the new controls.
The contributions of the project towards solving the previously mentioned issues include:

  1. An essay introducing the problems caused by the Wassenaar Arrangement’s rules regarding “intrusion software”.
  2. A panel discussion at the NCSC One conference by supervisors R. van den Hoven van Genderen and H. J. Bos.
  3. An advisory report and memorandum requested by the Dutch Cyber Security Council, containing:
    1. Historic and legal context of the Wassenaar Arrangement
    2. Detailed explanation of the issues regarding harmonization of national implementations between participating states.
    3. Detailed explanation of the problematic controls in the Wassenaar Arrangement;
    4. Recommendations on how changes in the wording of the Wassenaar Arrangement could resolve specific issues.

The Dutch Cyber Security Council acknowledged the problems that we addressed.
They are working on a solution in line with the content of our advisory report.


Playing it by ear: serious gaming for better hearing

To improve the hearing quality of hearing impaired individuals, regular training is important. Unfortunately, patients (especially children) are quickly bored by traditional training programs. Additionally, their social and communicative skills, including cognitive control of emotion and motivation, are often less developed. The aim of this project is to explore the potential benefit of gamification to make (computer-based) training applications for hearing impaired children more appealing, and therefore more effective. Mechanisms that will be investigated include fantasy (e.g., using role play), rewards (e.g., the possibility to obtain points or items), and challenge (e.g., controlling the difficulty level).

 Monique Lamers

Monique Lamers


Tibor Bosse

Despite the most sophisticated hearing aids, hearing impaired people experience difficulties in speech understanding, particularly in difficult hearing situations like noisy environments. Especially in the trajectory of rehabilitation, training is important. However, after years of being tested, (young) patients are quickly bored by traditional training programs. To make training applications more appealing and, therefore more effective, gamification can be an adequate method. However, with hearing impaired children showing less engagement than their normal hearing peers, it is an open question if and how gamification of speech perception training can be of added value for this target group.
The project Playing it by ear made some important steps towards answering this question. In particular, a prototype has been developed of a serious game for speech perception training. The game consists of three mini-games that use existing methods from audiology testing, as they focus on discrimination and identification of minimal pairs of phonemes in non-words and words, and the understanding of words both in isolation and in sentences. Using an adaptive mechanism, stimuli are presented in different noise levels depending on the performance of the gamer, using different types of noise. This all takes place in the scenery of an appealing setting for children: space. An initial version of the game has been presented as a poster presentation at the opening of the Language and Hearing Centre Amsterdam. In addition to the game, an experiment has been designed enabling future researchers to evaluate the usability as well as the effectiveness of the game in comparison to several baseline conditions.

Consumer Behaviour in Digital News Consumption

While the consumption of online news is increasing and analytical tools allow website owners to know which sections of their websites are being visited, it is unknown which topics on all Dutch news websites are being visited during the whole day. Do news users, e.g., read news about soccer in the morning at, articles about politics at during lunch break and gossip at in the evening?


Martijn Kleppe


Sandjai Bhulai

Expert versus public understanding of science

The central aim of this research project is to investigate expert versus lay understanding of scientific concepts as a window onto broader questions in the philosophy of science: Are such differences between experts and laypeople a matter of different degrees of understanding of the relevant concepts, or is the fundamental difference about different kinds of understanding? Answers to these questions can have important implications for science education in schools, popular media, and the formation of science policy for public consumption..


Henk W. de Regt


Alan Cienki

We consider understanding to be one of the central aims of science. However, in the philosophy of science there is still no consensus on an accurate notion of understanding. During our project we strived to shed light on this debate by investigating the potential difference between expert and lay understanding of science by studying the role of metaphor in multimodal forms of communication. For far from being mere rhetorical flourishes, metaphors are known to be used to construct and hypothesise about abstract theories and to provide, in general, sense-making frames of reality. This metaphor analysis therefore offers not only an insight into the possible differences between forms of scientific understanding, but also provides empirical basis as how to create and increase public understanding of science.This hypothesis was operationalised through Henk de Regt’s account of understanding being not merely a form of knowledge that can be captured in rules and facts, but is a skill that involves tacit knowledge. De Regt argues that in order to achieve scientific understanding, one as to be able to use a scientific theory, which is only possible when the theory is intelligible to the particular person. Based on previous literature on metaphor analysis, we hypothesised that metaphors, expressing mentalrepresentations, contribute to this intelligibility of, by making a mental representation of an abstract theory accessible, and therefore directly relate to understanding.
Our research showed that experts, having obtained a mental representation of the abstract scientific theory, use metaphors as ordinary scientific concepts referring to a precise object, mechanism or process. The metaphoricity of these metaphors is subsequently lost in expert communication. On the contrary, in expert to layman communications we found the metaphors used to retain this metaphoricity and hence a metaphorical openness. This confirmed our hypothesis that metaphoricity contributes to the intelligibility of a scientific theory, and hence understanding, by making a mental representation accessible. But it also showed that this metaphorical openness to have varying implications and ramifications to the form and content of this conceptual understanding. On the basis of our findings through metaphor analysis, we therefore concluded that understanding of science indeed can be conveyed to the public by providing metaphors as a figurative tool to obtain a mental representation of an abstract scientific theory, but that the form and depth-content of the mental representation, and hence understanding, varies widely.