As financial markets have evolved and become more digital, the ways of marketing, communication and customer service have also adapted to the times. Customers expect a higher standard from all industries, financial services included. On top of being a nuisance to unsatisfied customers, poor customer service costs industries globally ＄338.5 billion in potential revenue losses per year. The industry with the highest losses is financial services, with about ＄44 billion lost per year (Genesys, 2009). Systems utilizing customer data for the purpose of improving business relationships with customers through better customer communication and service are called Customer Relationship Management (CRM) models (Chen & Popovic, 2003). Part of CRM operations is lead management. Lead management in particular is the set of methodologies, systems and practices with the aim of helping to better service existing clients (retention) or discovering and bringing in potentially new clients (acquisition). This project is performed with the goal of improving the lead management system in the private banking department of ING. More specifically, the goal is to improve the identification of the best leads weekly from the pool of all available leads to be sent out to customer contact teams. The methodology consists of the integration of machine learning algorithms into the lead management infrastructure for the purpose of scoring leads in order to improve the selection process, leading to improved customer communication as well as revenue potential. Additionally, more information was put into the decision making by considering the performance and preference of the customer contact teams. Due to the usage of data modelling, a review of relevant compliance measures with regards to GDPR was performed, with an additional measure of decision explainability being proposed for this project as well as all projects using machine learning algorithms.Three different algorithms were tested, with the best one selected based on performance being a random forest model. The model was tested against the existing lead selection method for 6 weeks, and showed considerable and consistent improvement in performance from the third week onwards (up to 16%). The random forest was more flexible over the weeks and based on analysis of decision interpretability made on particular model decisions, benefitted largely from the inclusion of team performance. In fact, the team which was handling the lead proved to be the most important factor in the decision making of the model. Preference did not seem to have any particular impact on the performance of the leads and thus was omitted from the final model. Overall, based on the results of the testing, the use of machine learning algorithms was shown to significantly improve the performance of the lead management system, based on better lead selection and the consideration of team performance. For future research, it is suggested to implement machine learning techniques in the lead generation step (rather than after), in order to reduce the information restriction the algorithm faced in the case of this project.

The limitation of natural resources forces humanity to rethink its current habits of material use. Plastic product packaging, fossil fuels for transportation and energy generation are just a few examples of highly resource intensive processes in modern society. Urgency of this matter increases due to the fast-growing world population and big economies as India, China and Brazil. Understanding the current material flows and the system they flow through might generate knew knowledge of material use dynamics. This knowledge can help policy makers to stimulate material use efficiency. To generate the necessary knowledge this research takes an Urban Metabolism (UM) approach. In contrast to established material flow focused and analytical aggregated UM approaches, this research proposes a more comprehensive approach. Defining the UM not only as material flows and processes, but giving more weight to the Social, Economic, and Institutional system aspects as well as the role of actors. On top of the different system perspective the metabolism analysis is conducted on different time and spatial levels. The addition of disaggregated analytical levels enables the use of statistical analysis to discover material use patterns over time and identification of actor group related material-use behaviour. The statistical methods used are Time Series, Correlation and Geospatial Analysis in addition to the common Material Flow Analysis and a Life Cycle Assessment. The proposed UM approach is applied to the case of household waste management in Amsterdam. This research showes the potential of a deeper and holistic metabolism analysis enabled by increasing amount of available data. Opening up and troughing light into the up to this point Urban Black-Box.

Cardiovascular diseases are considered as one the deadliest disease and have also been the most prominent health burden around the world and particularly in the Netherlands. Enormously mitigation has been done to reduce the death burden, by improving the quality of health care services and research related to cardiovascular diseases. One prominent strategy to reduce it is to identify early symptoms of cardiovascular diseases among the potential population. Currently, the prevailing cardiovascular disease risk prediction guidelines that used by a general practitioner only taking into account straightforward factors into their risk factors, and significant improvement to the guidelines is needed to include more socio-economic factors into account since many expert realize the fallacy of the systems. This research expands the current cardiovascular risk estimation guidelines with socio-economic factors such as ethnicity, occupation, social deprivation, by utilizing Bayesian network modeling to understand better the nature of socio-economic factors related to cardiovascular disease risk among the Hague population in the Netherlands. This research is collaborative research between Leiden University of Medical Center (LUMC) as the problem owner, the data provider and knowledge expert and TU Delft as an analyst.

A nation’s main port is a crucial component that contributes to the economic development of a country. Therefore, the role of the port becomes more critical nowadays. However, due to the existing hub-and-spoke structure of the European multimodal transportation network, polarisation and regionalisation are visible in the network. The increasing regionalisation and polarisation that happens in the European region widen the inequality in the network. Moreover, the inequality in the network will impact the sensitivity and resistance of the whole network to targeted disruptions, such as infrastructure failures, natural disasters, or potential terrorist attack. Then, on the real aspect, the widening of inequality in the network also inhibits the well-distributed economic development in the region. In this study, various methods will be employed to have multiple points of view of the network itself, such as the multilayer network analysis, community detection, and Network-based Hub Port Assessment (NHPA). Furthermore, the study will be structured into three building blocks that shapes as a foundation to perform the final analysis and provide a multifaceted overview of the overall network, which is versatility analysis, community structure analysis, and quantification of collaboration-connectivity of the container hubs. Afterwards, three determinants as measurements of network performance are defined, to perform the criticality analysis. This study concludes that the development of connections between different communities can help to decrease the existing inequality in the network. Then, the appearance of new community structure in the network can help to decrease the inequality in the network, which lead to more well-distributed economic development in the region.

The dietary habits changed drastically since 1980 and the chemical composition of a lot of processed foods was altered to meet new requirements. This nutrition shift along with the industrialisation and the rise of the sedentary lifestyle, led to the spread of the obesity epidemic which still manifests in many countries, including European. Children are also affected from this epidemic, with the childhood overweight and obesity trends to grow. The OECD children average for 2017 reached a 15.5%. However, in each country different factors contribute to the manifestation of the phenomenon, composing its unique obesogenic environment. In the current thesis proposal, exploratory statistical modeling is employed, to identify the non-biological factors explaining children's weight for the Netherlands. The analysis is conducted with the use of a custom framework to capture factors which are linked with societal inequities. The outcome of this analysis is a group of factors which explain the weight in preschool children. Also, significant results for preschool children are derived by comparing different screening definitions for overweight and obesity, showing the clusters of risk-prone children. By using the insights from the analysis, data informed policies are proposed to aid the creation of a healthier urban environment for future population to thrive in. Data informed policies based on governmental data for the Netherlands can facilitate the decision making and make prevention and mitigation of the epidemic more effective, by targeting appropriately different population segments which are in need. Future research could focus on prediction models of obesity based on the significant factors identified in this study. Also, the expansion of the current models with more variables from the urban environment is needed to show more specific associations with urban features.

The urban population is globally increasing. Additionally, mental health problems are increasing. The determinants of mental health are found to be more present in urban environments. Due to the growing populations and the urban presence of mental health determinants, mental health problems are risking to further increase. This thesis develops a framework, in which the concepts that influence mental health are visualized. Based on this framework, the determinants of mental health are modelled for the case of Rotterdam.
Subsequently, the modelling results are translated into the policy system. The findings of this thesis are that the pathways towards mental health are complex, multivariate, interconnected and sometimes contradicting. In order to address this challenge, policy-makers should take an integral evidence-based approach. This thesis conducts a first exploration of the relation between urbanization and mental health. Furthermore, it takes the first step towards developing a policy-process that incorporates this knowledge and is able to act on it. Nevertheless, more research is needed about the urban determinants of mental health, in which data is used on individual level. Moreover, qualitative studies can research integrated evidence-based policy-making in more depth.

Simulation models are increasingly used to substantiate long-term planning policy decisions. The communication of model results from modellers to policy-makers has not received much attention by scholars. While, without correct interpretation and correct use of model results the quality of the model itself is irrelevant for the impact of the model on policy. This research analyses this gap in literature by looking at a system from model to policy, using the case of the 2017 office policy of Amsterdam.

Nighttime light remote sensing data is regarded as an important data that can be used to measure human socioeconomic activities. Because of the positive correlation between long time-series nighttime light data and economic development, long time-series nighttime light data can be used to construct GDP forecasting models to predict GDP. However, there are different temporal resolutions, measurement standards, and errors due to different sources of nighttime light data. How to eliminate the data errors and the differences between different data sources to construct a reliable long time-series nighttime light data is currently a major problem. In addition, different GDP forecasting models have different applicable conditions and forecasting accuracy. How to choose the optimal GDP forecasting model is another major problem. In this paper, data errors and differences are dealt with at first to obtain long time-series nighttime light data that can be quantitatively analyzed. Then, the advantages and disadvantages of different GDP forecasting models are analyzed and compared. The main contents of this paper are as follows.

The “affordability of housing” is generally defined as affordable housing for those with median household income (Eurostat, 2018). If not addressed effectively by policymakers, the unaffordable housing gap is expected to affect 1.6 billion people around the world by the year 2025 (McKinsey, 2014). The unaffordable housing in Amsterdam specifically harms the financial and social well-being of the residents of Amsterdam. Therefore, this research aims to grant a contribution to the field by using the Bayesian modelling methods on private rental market prices to reduce unaffordable housing issue. To investigate the issue, the researcher analyses the literature on urban economic models, the use of models in policy making and, collects data from the national database and online rental housing agencies. With the use of hierarchical Bayesian modelling and exploration tools, the relations between house features and local characteristics are explored and two price prediction models are built by using local house features such as size, bedroom number, distance to city centre and district category. The researcher finds several demand profiles and detected a strong negative correlation between rental prices and some industrial and business locations, which might provide an opportunity for city planners to combine the development of these locations with house supply injections to create more affordable housing. Moreover, from the two prediction models, the first model investigated the average district expensiveness better than conventional metrics by increasing its prediction accuracy and ability to quantify uncertainty. Furthermore, the second model categorized the Amsterdam districts according to the preference profiles obtained by model parameters to find most suitable districts for middle-income households. In both models, the location parameter is found to have the highest impact on rent prices. The research provides informative demand profile findings and a descriptive plan on housing supply injections which can be useful for policy makers. Moreover, the policymakers can benefit from the use of advanced model techniques to better assess the spatial housing market according to the city needs. Furthermore, the research evaluates the effect of local factors on rent prices in order to customize development plans to meet the citizen’s needs more robustly. Lastly, besides benefiting from policymaker’s improved actions, middle-income tenants themselves can also use the research findings to make more informed affordable housing decisions.

The use of machine learning systems has great potential to better predict probabilities of default for credit underwriting. Despite this advantage, herewith there exists the substantial risk of discrimination. Moreover, machine learning models with the highest prediction-accuracy are often the least explicable (i.e. explainable). Nonetheless, explicability is needed to create accountability of automated credit decisions by machine learning systems. Furthermore, there exists a regulatory need for explicability of machine learning systems in the General Data Protection Regulation (GDPR) and the Consumer Credit Directive (CCD). Besides that, an ethical- and societal need exists for explicability. Within the exploration of literature, it becomes clear that research lacks on how to move from a high-level principle like explicability, towards a prospective assessment of a machine learning use case on this principle, it lacks a multi-disciplinary perspective, and it misses an assessment framework that can guide decision-makers within machine learning use cases, aligned with a multi-organizational development lifecycle. This research aims to design a prospective pragmatic assessment framework that can guide decision-makers, within machine learning applications in European credit underwriting cases from the point of view of explicability. To accomplish this, the Design Science Research Methodology (DSRM), complemented with the Value Sensitive Design (VSD) approach, is utilized. To this end, the Explicability Assessment Framework (EAF) was developed. This framework is adapted to the context- and explanation characteristics of the case, and aligns with the CRISP-DM development lifecycle. It was found in two case studies that the framework helps with the decision-making whether a machine learning system is sufficiently explicable or not. Lastly, a wide range of future research areas is identified that needs attention: empirical validation and expansion of the framework, the relevance for automated explanation creation, the scalability to other context and a large amount of explanations, and the practical perspective regarding adoption in the industry.

In the coming decades, more frequent and more extensive climate disasters such as coastal and river floods, droughts, extreme weather, and wildfires can be expected worldwide. Innovations will be required to face this grand challenge. The BRIGAID project developed a methodology consisting of a Test- and Implementation Framework and a set of practical tools. BRIGAIDs tools are offered to support efficient development and market introduction of promising innovations. The methodology in its present state still requires an extension to cover cybersecurity issues. It should be assessed what components are key in innovation projects and where and whether cybersecurity is relevant within the TIF. The proposed research will establish an assessment for the cybersecurity readiness of BRIGAID’s innovation projects. The goal here is to give the innovation projects an indication on their level of security and cybersecurity readiness. We selected GM4W and QoAir as representative innovation projects for a case study consisting of a cyber risk assessment. We find that key cyber components for innovation projects benefit the identification and mitigation of cyber threats. When assessing an innovation, the cyber components serve as a starting point of the assessment. We used SecRAM as the risk assessment method in this study and aimed to test whether the method applies to the risk assessment of innovation projects. We conclude that the SecRAM method serves its purpose and applies to the innovation projects in the context of this study. The risk assessments applied to different cases with contrasting structures and enabled us to identify and mitigate cyber threats effectively. The use of SecRAM also applied to the design of the TIF cybersecurity extension. The tool consists of questions that score the innovation projects based on confidentiality, integrity, and availability to raise concerns on the cybersecurity readiness. The acceptance and perceived usefulness of the tool need validation among innovators. Future research should extend the involvement of innovators and experts to the entire risk assessment phase.

In this Master Thesis project, the objective is to study how can Supervised Machine Learning be used to detect text-based rumours for humanitarian activities in Twitter. A model was developed in this project in order to classify a tweet at question whether is a rumour or not and whether is relevant to humanitarian activities or not. The findings of this research were promising as the classification modules developed were able to score 75.8% in Recall classifying tweets to rumours and non-rumours and 96.6% in Recall classifying tweets to relevant to humanitarian activities and not relevant.

During the last decade, two developments on the Dutch housing market have received a lot of attention. One of these developments concerns the recovering house prices and specifically the strong price rises in cities. The other development is related to the growth of the private rental sector and particularly the buy-to-let market. This market consists of the private rental dwellings owned by reasonably small private landlords. These landlords predominantly buy owner-occupied dwellings to rent them out, hence the name buy-to-let. The growing buy-to-let market and the recent price rises have often been linked to each other. Some people claim that the buy-to-let landlords are driving up the house prices. Nevertheless, the relation between these two developments has not been properly studied yet for the Netherlands. Consequently, the objective of this study is to gain more insight in the Dutch buy-to-let market in general and specifically the relation between this market and the regional house price development. Mostly quantitative research methods are applied to answer this research question since especially quantitative information on the Dutch buy-to-let market is missing. House price models have for instance been made for Groningen, Breda and Zoetermeer. The research shows that the Dutch buy-to-let market is still quite small as the buy-to-let dwellings only comprise 5% of the total Dutch housing stock. By comparing the house price development of different municipalities with a varying share of buy-to-let, it can be seen that the recent house price increases are on average stronger for the municipalities with a high share of buy-to-let dwellings. It is however possible that these stronger increases are caused by other factors than buy-to-let. When comparing the results of the three regional house price models no big differences in the drivers are found. The results show that the determinants for the house price development are, despite the varying share of buy-to-let dwellings, in general the same for the three municipalities. Furthermore, a buy-to-let indicator could not explain some of the remaining unexplained variance. Based on these results, it can be concluded that none of the results actually point at an inflationary effect of buy-to-let investments. However, because only three municipalities were studied, no general conclusions can be drawn and the possible inflationary effect of buy-to-let landlords cannot be completely excluded.

About 85 percent of all volume trade goes by marine traffic nowadays which is not expected to get any less soon. In marine operations there are different operations with each having their own characteristic. Different trades for example include containers, tankers, bulk, or passenger vessels. But in the end, all of them have to enter a port, this is the place where many actors come together to service a vessel. Due to the many actors involved in the port call, which could run up to around ten actors, planning a vessel stay can get very tedious. Planning the port call can get especially hard if sequential services are not aligned with each other or if relevant information appears to be inaccurate or even missing.
Major complications in port calls are therefore the lack of information sharing. Often parties have a very poor insight in when a vessel is arriving or departing. In general actors in the port environment have been striving to optimise their own processes not including others affected. Very few research has been done in port operations to see what the effects are of data sharing and collaboration among actors. The question is therefore how information sharing among actors in a port call can affect the situation and to what extent. And in particular what information should be shared and with what interval.
To quantify the effects of data sharing three major components are included in the research approach with the port of Rotterdam as use case. The first part of the research focusses on qualitative aspects exploring actors and the port call event. Through this part of the research a better understanding of the port call process is gained which will be useful for the next steps. Also understanding which actors have a dominant role, benefit, or have a lot of power is important for further steps in the research. After a clear overview of the port call and most important actors a combination of data analysis and modelling is done. In this research a discrete simulation model is used to make an abstraction of the real world and use this for testing. Through a simulation parts of the port call process can be tested under different circumstances or inputs of interest. Outputs will then give an indication how the system will respond to particular changes. To get the model correctly running data from the Port of Rotterdam will be used for a correct parameterisation. Parameters would include statistics of port operations such as the number of vessels, handling time, and speed of the vessel.
After going through the previous mentioned steps results show that vessels can reduce their waiting time at anchorage by 35% and therefore their fuel consumption as well. One of the biggest gains would be realised if captains and terminals would start sharing information with each other about arrival and departure times. Ideally this would be done on an interval smaller than 2 hours. When vessels are aware of delays in the berth they can slow down to arrive just in time at the anchorage, or perhaps they can sail straight to the terminal. This information towards the captain is crucial as it can be used to adjust speed and thereby realise fuel savings. In the most optimal case the waiting time at anchorage could be reduced by 35%. Furthermore throughout the whole process more accurate information is required which will support actors in making a more robust planning and be able to plan farther ahead.
Two things need to be done from here, one is further research to consolidate these outcomes and see effects in other operations such as bulk or on more microscopic level such as inland shipping. Also research with regard to the implementation will be required to get everyone on board such as actors with fewer gains that are required to make this a success. the second is to get stakeholders together and make them realise that cooperation and sharing of data will have tremendous implications not only for the waiting times but also for CO2 emissions and robustness of operations.

The last decade saw the rise of e-commerce trade and the shift of the manufacturing industry to the emerging economies, China first of all. In this context, the European Customs Authorities experienced an explosion of small parcels coming from e-commerce websites, often from China, and faced difficulties to detect fiscal frauds and security threats using their conventional risk management systems. To address this problem, the European project PROFILE brings together the customs administrations of Netherlands, Belgium, Sweden, Norway, and Estonia, aiming to provide the EU with a shared platform for: (1) accurately assessing customs risks; (2) optimizing operation and logistics by integrating multiple sources of information; (3) developing a shared data platform to share customs risk management (CRM) practices.

As part of this project, the Dutch Customs Administration (DCA) and International Business Machines (IBM) Corporation are collaborating to deploy the cutting-edge technologies of artificial intelligence to automatically cross-check the customs declarations coming from Chinese e- commerce against online information. Through a Design Science approach, I carried out this research for the Delft University of Technology, written in collaboration with IBM Netherlands, aiming to deliver a preparatory study for the developing team before the PROFILE project begins. This includes knowledge brokering between the Dutch Customs Administration and IBM Netherlands so that a more precise problem scope can be defined, and the requirements elicited. In particular, this research focuses on the first part of the project: the development of an adaptive web-crawler for e-commerce, able to compare the declarations documents against online information.

According to the Dutch Customs Administration, the web-crawling system should gather the description of the goods from declarations, search the product on the web, find its price of sale on the e-commerce platforms, compare it with the value declared in the declaration, and return a risk indicator of green/red flag to the targeting officer. The design process of this system follows approaches coming from the systems engineering discipline, starting with the requirement analysis, addressing them with the state-of-the-art big data analytics, and finally deriving the logical components of the system, whose design is presented through a logical architecture.

First, the application domain is investigated. When goods entry the Netherlands need an entry declaration. These goods arrive at the harbor of Rotterdam or airport of Schiphol, where some of these are imported into the country and become import/export, and others stop temporarily as transit waiting to be shipped somewhere else. The Dutch Customs Administration monitors these processes through risk management systems aiming to stop non-compliant goods. This research describes these practices, with a higher focus on the e-commerce risk targeting. About the e- commerce world, a study of the e-commerce processes behind an online purchase is also carried out through a real purchase on Chinese e-commerce. This was used to observe how the Chinese sender described the item, and how the Dutch Customs assessed the risk and decided on the duties to be paid. This led to reflect on the possible frauds scenarios and how to address them. Finally, the Dutch Customs also reported that the products descriptions are often vague and ambiguous, and a more accurate formulation of the problem is described.

Secondly, an in-depth literature on the fields of web-crawling and big data analytics techniques is carried out. The possible technologies that could be useful to address the requirements and the problem formulation are investigated. Starting with an analysis of the existing literature on the field of big data analytics, this research also covers the recent trends of machine learning and artificial intelligence. To avoid reporting a too big literature, the topics reported have been accurately chosen, for instance describing only the techniques for web analytics and text analytics.

This literature on big data analytics is further broken in two sub-topics, one more theoretical, which classifies the types of analytics methods and defines the technology of machine learning and natural language processing, including the last paradigms of deep learning and reinforcement learning, and one more practical, where guidelines for the design, development, and implementation of machine learning techniques are proposed. It is here that a theoretical framework to systematically reflect on the challenges of the field of big data analytics has been identified. This framework is then used to systematically collect the main technological challenges of the use case under analysis and translate them into non-functional requirements.

Finally, the last part of the literature describes what a web-crawler is and what web- crawling/web-craping means. This later extends to the concepts of focused web-crawling and smart, intelligent, adaptive web-crawling, where machine learning techniques are deployed to improve performance. The literature concludes by providing related works of machine learning techniques implemented in smart web-crawling of the e-commerce websites and stating the knowledge gap that needs to be bridged to address the use case under analysis.

After the application domain and the literature review, the knowledge from these previous phases combines in a continuous iterative process according to the design science methodology (Hevner, 2014). Through unstructured interviews with the DCA and IBM experts, the requirements elicitation is carried out. The approach by Armstrong and Sage (2000) deriving from the field of systems engineering is used. The main objective of the system to be developed is broken down into a series of sub-activities that must be carefully structured to formulate the requirements. About the non-functional requirements, instead of reflecting on the different domains – technological, environment, law compliance, etc. – as it is proposed by the same systems engineering approach mentioned earlier, this research uses the framework identified in the literature review about the main challenges of big data project (Sivarajah, 2016).

To derive the components of the architecture from the requirements and customer needs, the methodology proposed by Suh (1998) called Axiomatic Design has been used, mapping the requirements into architectural components in a rigorous manner. In this way, the design domains proposed by this methodology – customer, functional, physical and process domains – are taken as the reference point for the design process: first, the business needs are identified, then these are translated into requirements, which are mapped into design features. The process domain is left out of this research and will be addressed by the IBM development team in Ireland.
The design cycle leads to the design of a web-crawling system represented through a service- oriented architecture (SOA). Its block diagram and black-box description of each application service are provided. Furthermore, the architecture functionality is described with an architecture walk-through and a sequence diagram in the unified modeling language (UML). The result is an innovative real-time web-crawling system to identify the value of a given product on the e-commerce websites. It deploys natural language process models to filter the non-relevant search results, and other machine learning models to best matching the remaining relevant results with a given item description.

The design and architecture description of this innovative web-crawling system is the main artifact of this research, while the mixed methodology of systems engineering methodologies and big data frameworks is another important scientific contribution.

The theory of the Habermasian public sphere is applied to Twitter. To do so, the traditional literature on public opinion is combined with more recent literature on social media. After the literature research the main knowledge gap is identified. This gap is the lacking relation between the philosophical investigation on public opinion on Twitter and an empirical investigation on actual tweets. This thesis combines the fields of philosophy and data science to contribute to the knowledge gap. Palm oil is used as a case study. The main research question is:

How can public opinion on Twitter be characterized, on the case study of palm oil?

The findings of this research suggest how Twitter, in its current form, does not create the optimal environment for forming something that is approaching public opinion due to five characteristics: the exclusion of more than half the world population (a), the domination by a few users (b1) (often with strategic interest (b2)) , the pollution through spam (c), Twitter’s role (d) and the
characteristics of tweets (e).

a. The exclusion of more than half the world population from social media is at least based on the lack of internet access, new media literacy and cultural capital.

b.1. The dominating key players are users who get the most attention from other users, policymakers and journalists. This can either be on a specific topic, or transcending topics. The findings from this research suggest that organizations like Greenpeace, BBC Earth and the Roundtable on Sustainable Palm Oil have an above average role in guiding the thinking of the public and therewith steering public opinion. It is difficult to find a user that is saying
something completely different than what the defined influencers are saying. These few key players play an above average role in defining public opinion on Twitter, and the influence of this role is not limited to Twitter.

b.2. The tweets analyzed show how the dominant key players often act with a strategic interest in mind. The public sphere should, according to Habermas, be free of power and ideology and the only common interest should be the interest in good opinion formation.

c. The pollution through spam is argued to play a serious role in influencing public opinion on Twitter. The exact effects of fake accounts and hyperactive automated users is unclear but it can be argued, as can be seen from the examples, that they cause a stream of misinformation and therewith influence public opinion.

d. It cannot be proved what Twitter is doing, and if and how they have influenced the debate on palm oil, since the algorithm is nontransparent. Examples like Morozov’s accusation of censorship in Twitter’s ‘trending algorithm’ shows the influence they can have

e. tweets are not just 280 characters. They are often carefully considered combinations of hashtags, calls for action, snippets of information, interpreted
data and images or videos. These characteristics all help in convincing the reader of the point the user is trying to make.