ODISSEI Conference for Social Science in the Netherlands 2023

This talk will explore the societal implications of the ongoing shift toward linked and enriched social data in research. What are the risks and concerns of the new data sources and analytical practices involved, and are current disciplinary and legal rules sufficient safeguards? Principles of data justice will be explored as benchmarks for the beneficent use of population data.

Making sensitive data available via SANE
Lucas van der Meer, ODISSEI, Erasmus University Rotterdam, School of Social and Behavioural Sciences (EUR-ESSB)
Privacy, copyright, and competition barriers limit the sharing of sensitive data for scientific purposes. We propose the Secure Analysis Environment (SANE): a virtual container in which the researcher can analyse sensitive data, and yet leaves the data provider in complete control. By following the Five Safes principles, SANE will enable researchers to conduct research on data that up until now are hardly available to them. SANE comes in two variants. Tinker SANE allows the researcher to see, manipulate and play with the data. In Blind SANE, the researcher submits an algorithm without being able to see the data and the data provider approves the algorithm and output. SANE uses concepts from the CBS Remote Access Environment, ODISSEI Secure Supercomputer and SURF Data Exchange, to build a generic off-the-shelf solution to be used by any sensitive data provider and researcher. SANE can be used by researchers in any discipline, as illustrated by the involvement of consortia in both the social sciences (ODISSEI) as well as humanities (Clariah). Potential sensitive data providers include the Dutch Chamber of Commerce (KvK), ING,, National Library of the Netherlands (KB) and Netherlands Institute for Sound and Vision (NISV).

Showcase of the DANS Data Station for the Social Sciences and Humanities
Ricarda Braukmann, Data Archiving and Networked Services (DANS); Jetze Touber, Data Archiving and Networked Services (DANS)
DANS is the Dutch national centre of expertise and repository for research data. Our domain-specific data archiving and publishing services enable researchers and data stewards to make their data FAIR and share them for reuse where possible. In this presentation, we would like to showcase our new repository: the DANS Data Station Social Sciences and Humanities (SSH). The release of the DANS Data Station SSH is part of the transition from our previous system EASY to a new repository system based on the open source software Dataverse. The Data Station has a number of new features and improvements that make it easier to archive, publish and find data:
– When depositing data, researchers have the option to restrict access to particular files if they cannot be openly shared. The data station also allows for new versions of a dataset if the research is updated.
– Detailed information about the datasets can be added in various metadata fields and we support SSH-specific vocabularies such as the European Language Social Science Thesaurus (ELSST) and vocabularies from CESSDA, the Consortium of European Social Science Data Archives.
– Datasets are assigned with Persistent Identifiers and the metadata is automatically transferred to the ODISSEI Portal and European data portals to make data findable for others.
– Finding relevant data in the Data Station is supported through full text and advanced search and various filter options. PDFs, images and videos can be previewed directly.
In this presentation, we will demo these and other features of the Data Station and guide the audience through the process of archiving and finding data for reuse. We will elaborate on the importance of archiving and publishing data in a trustworthy digital repository like our Data Station and how this supports FAIR data and Open Science.

Searching data using of the ODISSEI Portal
Angelica Maria Maineri, ODISSEI, Erasmus University Rotterdam, School of Social and Behavioural Sciences (EUR-ESSB); The ODISSEI Portal Team
Datasets are often scattered across different institutional, national, and international repositories. In the case of CBS microdata, the only public information about available datasets used to be in pdf files, organised thematically, accessible only via the CBS website. This fragmented landscape limits the findability of new data sources a researcher might not be accustomed to, and makes the process of selecting an appropriate data source time-consuming and inefficient. The goal of the ODISSEI Portal is to enable and facilitate the discovery of social sciences datasets across various data providers in the Netherlands within a single interface. The Portal currently collects metadata from social science datasets available at DANS (the Dutch national centre of expertise and repository for research data), DataverseNL and the International Institute for Social History (IISH), from the LISS archive, and from the microdata catalogue available at Statistics Netherlands (CBS). Metadata of these collections is harmonised and enriched using a thesaurus which enables multilingual search, while the user interface allows users to search across all datasets available in these collections to find data relevant for their research. The prototype ODISSEI Portal is publicly available (https://portal.odissei.nl/) in order to be able to get feedback from the community on the design features and functionalities. During the demo presentation, after a brief introduction we will show the latest version of the Portal to the ODISSEI community. A virtual “suggestion box” will be made available to all participants in the session.

ODISSEI SoDa support and fellowship.
Erik-Jan van Kesteren, Utrecht university

The ODISSEI SoDa team helps social scientists with data intensive and computational research. If you work at an ODISSEI organization, we will even do this for free! Visit our booth at the conference so you can:

• chat with us about your research, see how we can help you.
• sign up for our upcoming workshops (e.g., supercomputing with the OSSC).
• look at some projects we have done in the past.
• try out creating synthetic data with our metasyn software.
• ask us all about our new fellowship opportunity!

If you want to know more about what we do and how we do it, visit https://odissei-soda.nl.

Exploring the significance of employment for the chronically ill. Showcase of the value of linking CBS microdata to survey data from the National Panel of the Chronically Ill and Disabled and the Dutch Healthcare Consumer Panel.
Annette Scherpenzeel, Netherlands Institute for Health Services Research (Nivel); Anne Brabers, Netherlands Institute for Health Services Research (Nivel)
The prevalence of chronic conditions is expected to rise in the coming years due to population aging and factors like sedentary lifestyles. Our study focuses on the labor participation and quality of life of persons with chronic conditions compared to the general population. For this aim, we linked survey data from the National Panel of the Chronically Ill and Disabled (NPCD) and the Dutch Healthcare Consumer Panel (CoPa, a sample of the general population) with registration data on employment history from Statistics Netherlands (CBS). Both panels are managed by the Nivel.
Our findings showed that persons with chronic conditions, who have not yet retired, were less likely to be working over a four-year period than those without such conditions. Moreover, currently employed persons with chronic conditions experienced more unemployment and illness/disability benefit episodes in their recent history compared to those without chronic illnesses. Labor participation was lowest among persons with cardiovascular diseases, followed by diabetes and lung diseases. Employed persons with chronic conditions more often felt a sense of societal inclusion than non-working counterparts, although this varied depending on the specific type of disease.
In addition to the research findings, we evaluated the process of linking data from the Nivel panels and the CBS, as well as the added value of the linked data. A significant proportion of Nivel panel members consented to the data linkage, and nearly all panel members’ data were successfully linked to the microdata. Our analysis provided insights that could not be obtained from CBS microdata or panel data alone, such as insights into the relationship between chronic conditions, work history, and the sense of belonging in society. This marked the first linkage of Nivel panel data with CBS microdata, a succesful starting point for future projects.

Less happy with the same: The role of migrant composition in shaping the migrant gap in neighborhood satisfaction in the Netherlands
Weiyi Cao, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG)

Lower levels of housing and neighborhood satisfaction have been found to be related to constraints in income, housing tenure, household type, housing attributes, and neighborhood conditions. However, socioeconomic profiles and objective housing and neighborhood characteristics cannot explain all the variations in the satisfaction gap between migrant and native households. This paper sheds new light on the role of neighborhood attributes, specifically the migrant composition, in shaping the nativity gap in neighborhood satisfaction in the Netherlands by investigating whether there exists an interactive effect of migrant status by migrant composition. The study runs ordinal logistic regressions using the 2018 wave of the comprehensive national survey on housing conditions and satisfaction (WoON) in the Netherlands. The result shows that migrant households are similarly satisfied with the living environment when living in a home and neighborhood of the same quality as non-migrant households. However, when the interaction term between the migrant households dummy and neighborhood migrant composition is included, the key results indicate that the nativity gap in neighborhood satisfaction varies as the migrant composition changes: migrant households express lower satisfaction with the neighborhood’s living environment than their native counterparts when the share of migrants in the neighborhood is less than 17%. Constraints in migrant households’ accessibility to particular amenities and social interactions are possible explanations for this satisfaction gap in native-majority neighborhoods. The paper contributes to the housing satisfaction literature by explaining the nativity gap in residential satisfaction with neighborhood characteristics, which has implications for immigrants’ demands for particular amenities, constraints in social interactions, and their integration into the settlement society.

Paying more to live in native neighborhoods: migrant price differentials in the home-buying market of the Netherlands
Weiyi Cao, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG); Nico Heerink, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG); Eveline van Leeuwen, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG)

Few studies investigated racial or ethnic price differentials in the European context. This paper investigates the patterns of intra- and inter-neighborhood price differentials between migrant and native households buying similar housing in similar neighborhoods in the Netherlands by using the pooled national surveys of Housing Research (WoON) 2015 and 2018 combined with micro-data. Parallel to U.S. studies that have distinguished neighborhoods as Black, integrated, and White neighborhoods, this study categorizes neighborhoods as native, integrated, and migrant ones. Results suggest that some patterns of price differentials found in the Netherlands resemble those in the U.S., yet others are different. First, the negative associations between the migrant share in the neighborhood and housing prices, interpreted as effects of prejudice against ethnic minorities, are also present in all three types of native, integrated, and migrant neighborhoods in the Netherlands; yet the negative relation is much stronger in native neighborhoods than in integrated and migrant neighborhoods. Second, the tipping points pushing prices upward when neighborhood migrant share reaches certain cut-off points, which are interpreted as the result of segregation in the U.S. empirical literature, are not found in this study. In addition, we further test whether intra-neighborhood price differentials due to supplier discrimination vary (1) across three types of neighborhoods of different migrant shares and (2) across homebuyers of different income groups, both of which have been rarely done in previous studies. The key results indicate that migrant households who buy housing in native neighborhoods on average pay a significantly higher price of 9.86%. For high-income migrant households, the price premium declines slightly to 7.88%. While both differentials are only found in native neighborhoods, the magnitudes are higher than those found in U.S. studies, calling for studies to further examine the issue of discrimination against foreign-born population in the housing market of the Netherlands.

Offline trumps online: name recognition effects in the Five Star Movement 2012 Primary Election
Giovanni Cassani, Tilburg University, Tilburg School of Humanities and Digital Sciences (TiU-TSH); Francesco Marolla, Tilburg University, Tilburg School of Social and Behavioural Sciences (TiU-TSB); Maria Lucia Miotto, Erasmus University Rotterdam, School of Social and Behavioural Sciences (EUR-ESSB)

We study the impact of (i) online engagement on the MeetUp platform and (ii) previous candidacies in local elections on the electoral prospects of the candidates in the 2012 online primaries of the Italian populist party Five Star Movement (FSM).
While previous research has produced mixed findings regarding the influence of extensive digital media use on intra-party democracy, our study addresses a critical research gap by examining how digital tools and previous candidacies shape the distribution of power within a political party.
Leveraging data from the FSM’s online galaxy, we derive the online network of party members and explore the relationship between their network centrality (to quantify levels of online engagement) and their result in the 2012 primary election. We hypothesise that higher network centrality correlates with better electoral performances, especially considering the emphasis the FSM put on grass-root online activism. We further control for documented influences on the voting process by the voting platform design, especially ballot order effects. We also look at whether candidates who ran in previous local elections (but lost) were advantaged due to name recognition.
Our results show that online activity had no correlation with electoral outcomes, with no differences between candidates who were or were not on MeetUp. Network centrality was also uncorrelated with the election outcome. Contrarily, the few candidates who already ran for local office received a consistent advantage, which remains significant after controlling for the influence of the voting platform.
While we considered a single platform to quantify online activity and plan to extend our study to other arenas where the party’s political discourse was articulated, our results suggest that online campaigning efforts were not rewarded, whereas electoral results were heavily conditioned by flaws in the platform design and party endorsements in local elections.

Assessing bidirectional relationships between noise annoyance and health: a cross-lagged panel analysis
Lion Cassens, TU Delft – Faculty of Technology, Policy and Management (TUD-TPM); Sander van Cranenburgh, TU Delft – Faculty of Technology, Policy and Management (TUD-TPM); Simeon Calvert, TU Delft – Faculty Of Civil Engineering and Geosciences (TUD-CEG); Maarten Kroesen, TU Delft – Faculty of Technology, Policy and Management (TUD-TPM)

Noise pollution causes severe health implications, including hearing loss, insomnia, and cardiovascular diseases. Many of these health effects are mediated through increased annoyance and stress levels. With an estimated 22 million cases of high and chronic noise annoyance in Europe, the relationship between noise annoyance and health becomes even more important. While it is often assumed that noise annoyance leads to health implications, some studies have shown that health implications can increase our sensitivity to noise. Due to this increase in noise sensitivity, deteriorating health conditions may lead to increased self-reported noise annoyance. Yet, estimations of health effects from noise almost exclusively rely on cross-sectional data, which cannot account for this possible reverse causality. This study assesses potential bidirectional relationships between noise annoyance and health outcomes using random intercept cross-lagged panel models (RI-CLPM) on LISS panel data. RI-CLPMs are structural equation models specifically designed to analyse panel data regarding the influence that multiple variables have on each other over time. This makes them well suited for an analysis of the bidirectional relationships between noise annoyance and health outcomes. The LISS panel consists of a representative sample of Dutch households and surveys, among others, the health of household members and their housing conditions, including self-reported noise annoyance from neighbours, aviation, and street traffic. By analysing this data, we clarify to what extent health implications affect noise annoyance over time and vice versa.

From Finding to Re-using Confidential Data
Freek Dijkstra, SURF; Emma Schreurs, University of Amsterdam – Faculty of Social and Behavioural Sciences (UvA-FMG)

The Research Data Exchange (RDX) allows researchers to share confidential data in a controlled and secure manner, while also adhering to legal requirements and institutional policies. The RDX is a prototype that integrates existing data repositories with algorithm-to-data solutions. The prototype was build by SURF in collaboration with the UvA. At first glance, it simply provides glue that allows a seamless workflow from ODISSEI data portal to ODISSEI secure analysis environment. Yet, it is more profound: by validating data sharing conditions and providing logging and monitoring of all activities on a confidential dataset, it builds trust into the system, and helps to solve the Open Science Dilemma, by both allowing open science and on the other hand also solving sovereignty issues can limit the extent to which data can be openly shared.

Mental health workforce in primary care and the allocation of depression treatment: a data linkage study
Jesper Dros, Netherlands Institute for Health Services Research (Nivel); Christel van Dijk, Zorginstituut Nederland; Koen Böcker, Zorginstituut Nederland; Robert Verheij, Netherlands Institute for Health Services Research (Nivel); Bert Meijboom, Tilburg University; Jan-Willem Dik, Zorginstituut Nederland; Isabelle Bos, Nivel

Introduction: In 2014, major reforms took place in Dutch mental healthcare. The mental health nurse, introduced in 2008 as additional healthcare provider for individuals in need of mental healthcare in general practices, was expected to substitute treatments from general practitioners and providers in basic- and specialized mental healthcare (psychologists, psychotherapists, psychiatrists, etc.). The goal of this study is to investigate the association between the degree of deployment of mental health nurses and treatment by other mental healthcare providers.
Method: We conducted an observational study based on pseudonymized claims data of Dutch health insurers combined with electronic health records from general practices. Healthcare utilization patterns of individuals with depression treated in general practice or basic- or specialized mental healthcare were analyzed between 2014 and 2019 (N=32200). Both the proportion of individuals treated and time to treatment after depression onset were assessed in association with the degree of deployment of the mental health nurse in general practices.
Results: The proportion of individuals with depression treated by the GP (OR 0.94, 95% CI 0.90 – 0.98), in basic mental healthcare (OR 0.85, 95% CI 0.82 – 0.89) and in specialized mental healthcare (OR 0.90, 95% CI 0.87 – 0.94) was lower when mental health nurse deployment was high. Individuals from general practices with high nurse deployment are treated in basic mental healthcare at a later point in time. However, they remain in specialized mental healthcare for approximately the same duration and are not referred back to primary care earlier, irrespective of mental health nurse deployment at the GP.
Conclusion: Treatments for individuals with depression have shifted from the general practitioner and both basic- and specialized mental healthcare to the mental health nurse. Thereby, general practitioners can dedicate more time to other patients. Future research should focus on referral from specialized care back to basic- and primary care, as this anticipated shift has not been found.

On Text-based Personality Computing: Challenges and Future Directions
Qixiang Fang, Utrecht University – Faculty of Social Sciences (UU-FSW); Anastasia Giachanou, Utrecht University – Faculty of Social Sciences (UU-FSW); Ayoub Bagheri, Utrecht University – Faculty of Social Sciences (UU-FSW); Laura Boeschoten, Utrecht University – Faculty of Social Sciences (UU-FSW); Erik-Jan van Kesteren, Utrecht University – Faculty of Social Sciences (UU-FSW); Mahdi Shafiee Kamalabad, Utrecht University – Faculty of Social Sciences (UU-FSW); Daniel Oberski, Utrecht University – Faculty of Social Sciences (UU-FSW), UMCU

Text-based Personality Computing (TPC) refers to automatic personality assessment based on text data (e.g., tweets, essays). Many empirical studies and datasets, plus a few survey papers exist. However, there is not yet a position paper that reflects on the quality of current TPC research, suggests solutions and future directions, and combines perspectives from NLP and social sciences. In this paper, we review existing TPC research and identify 15 challenges deserving the attention of the research community. We organize these 15 challenges into the following 6 topics: personality taxonomies, measurement quality, datasets, performance evaluation, modelling choices, as well as ethics and fairness. Furthermore, in light of these challenges, we offer concrete recommendations for future TPC research:

– Personality taxonomies: Choose Big-5 over MBTI; Try modelling facets and using other taxonomies like HEXACO where appropriate.
– Measurement quality: Pay attention to measurement error in personality measurements, be they based on questionnaires or models; Try to reduce measurement error by design (e.g., choose higher-quality instruments; use better data collection practices); Provide quality evaluation (i.e., validity and reliability) for any new (and also existing) approaches.
– Datasets: Make TPC datasets shareable, which should also contain fine-grained personality measurements and descriptions of the target population;
– Performance evaluation: Report a diverse set of performance metrics; Report R2 for a regression task.
– Modelling choices: Make use of their psychometric properties when modelling personality traits (e.g., use joint modelling; modify the loss function to preserve the covariance information); For even better predictions, try incorporating personality questionnaire texts, applying data augmentation and dimensionality reduction techniques, as well as incorporating more personality-related variables.
– Ethics and fairness: Avoid unnecessary TPC; Apply TPC to clinical, professional and educational settings; Investigate fairness.
– Lastly, engage in (interdisciplinary) research work with survey methodologists, psychologists, and psychometricians.

The Hidden Divide: School Segregation of Teachers in the Netherlands
Rafiq Friperson, VU Amsterdam – School of Business and Economics (VU-SBE); Hessel Oosterbeek, University of Amsterdam – Faculty of Economics and Business (UvA-FEB);Bas van der Klaauw, VU Amsterdam – School of Business and Economics (VU-SBE)

We use Dutch register data to document the understudied phenomenon of teacher segregation. We show that teachers in primary and secondary schools in the four largest cities of the country – Amsterdam, Rotterdam, The Hague and Utrecht – are segregated in terms of their migration and social backgrounds. While segregation by social background is not much higher than what would be expected under random teacher-school assignment, segregation by migration background is substantial even after accounting for randomness. Relating schools’ teacher composition to their student composition, we find in most cases that schools with a high proportion of teachers from a particular background tend to have a high proportion of students from that same background.

Bridging Questionnaires and AI Models for Insights into the Intercorrelations of Big Five Traits
Anastasia Giachanou, Utrecht University – Faculty of Social Sciences (UU-FSW); Yucheng Chen, Utrecht University – Faculty of Social Sciences (UU-FSW); Qixiang Fang, Utrecht University – Faculty of Social Sciences (UU-FSW)

In recent years, the field of personality computing has witnessed significant progress, due to the integration of machine learning and natural language processing (NLP). While traditional approaches have relied on questionnaire-based methods, the advent of NLP has brought new opportunities for automated personality detection by analysing user-generated text. In particular, recent models such as BERT (Bidirectional Encoder Representations from Transformers) have demonstrated impressive performance across various tasks including automated personality trait assessment. Automated models are mainly evaluated based on their effectiveness to capture personality traits, while their intercorrelations are hardly discussed. This comes in contrast to the traditional approaches which have extensively analysed those intercorrelations. For example, positive correlations have been found between openness, extraversion, and conscientiousness, while neuroticism has shown negative correlations with conscientiousness. Comparing the traits’ intercorrelations when predicted by automated models with established psychological findings, we can examine whether the automated assessments align with the existing knowledge in the field. Our study aims to investigate and compare the performance and intercorrelations among the Big Five personality traits using two models: the Robustly Optimised BERT Pretraining Approach (RoBERTa), and Bi-directional Long Short-Term Memory (Bi-LSTM). Our experiments on PAN 2015 and PANDORA showed that RoBERTa outperforms Bi-LSTM in predicting personality traits. However, the performance of the models varied across the two datasets that suggests a diversity in the expression of personality traits. In addition, RoBERTa captured most of the sign (positive or negative) correlations present in the original datasets. We only observed differences in the pairings of extraversion-openness and neuroticism-conscientiousness in the PAN 2015 dataset. This indicates that RoBERTa primarily learned the correlations present in the annotated data. Our findings indicate that there are many limitations in the data collection process that need to be addressed in order to create more robust models for personality prediction.

Farmers’ Adaptation to Climate Change in Different Regions of the World: A Natural Language Processing Systematic Review
Sofia Gil-Clavel, TU Delft – TPM; Tatiana Filatova, TU Delft – TPM

Climate Change is expected to affect the global agriculture in negative ways if farmers fail to incrementally adapt early in the twenty-first century, or transformationally adapt during the second half of the century. Many publications from anthropology, sociology, economics, and geography discuss the underlying mechanism of farmers’ adaptation to Climate Change. However, researchers have not systematically analyzed their findings, mainly because this literature is primarily qualitative and vast. This work aims to analyze the factors that drive farmers’ adaptation to Climate Change in different regions of the world. Specifically, we aim to answer the following research questions: What drives farmers to adapt to climate change? Are there patterns in the mechanisms found in different regions of the world? Are there differences in the mechanisms behind farmers’ incremental adaptation and transformational adaptation? For this, first, we perform an active learning analysis of publications contained in Scopus in August 2022. Second, we analyze the articles’ findings using semi-supervised natural language processing. Finally, we derive a database that can be used as an input for generalized linear models. Preliminary results show that climate change evaluation factors (knowledge, believe, and concern) together with individual costs and benefits seem to be worldwide associated with farmers’ transformational adaptation to Climate Change.

Migration Policies and Immigrants’ Language Acquisition in EU-15: Evidence from Twitter
Sofia Gil-Clavel, TU Delft – TPM; André Grow; Maarten J. Bijlsma

In response to the increasingly complex and heterogeneous immigrant communities settling in Europe, European countries have adopted various civic integration measures. Measures aiming to facilitate language acquisition are considered crucial for integration and cooperation between immigrants and natives. Simultaneously, the rapid expansion of social media usage is believed to change the factors affecting immigrants’ language acquisition. However, only a few previous studies have analyzed whether this is the case. This article uses a novel longitudinal data source derived from Twitter to (1) analyze differences in the pace of immigrants’ language acquisition depending on the migration policies of destination countries and (2) study how the relative sizes of the migrant groups in destination countries, and the linguistic and geographical distances between origin and destination countries, are associated with language acquisition. Results show that immigrants who live in countries with strict language acquisition requirements for immigrants and conservative citizenship policies have the highest median times until language acquisition. Based on Twitter data, we also find that language acquisition is associated with classic explanatory variables, such as the size of the immigrant group in the destination country and the linguistic and geographical distance between origin and destination country similar to the previous studies.

Education, unions, and physiology: explaining the gap between intended family size and completed fertility for the 1974-1984 birth cohort of Dutch women
Rolf Granholm, University of Groningen – Faculty of Behavioural and Social Sciences (RUG-FGMW); Gert Stulp, University of Groningen – Faculty of Behavioural and Social Sciences (RUG-FGMW); Anne Gauthier, Netherlands Interdisciplinary Demographic Institute (NIDI)

Women have fewer children than they intend to in Europe, resulting in a gap between intended family size and competed fertility. The gap is mainly a result of first-pregnancy attempts being postponed to older reproductive ages, when physiological constraints make successful conception and birth increasingly difficult. Expansion of higher education and current union formation and -dissolution trends are considered two important factors behind this first-pregnancy postponement, which is why they are the focus of this study. Understanding the fertility gap is key to understanding why fertility is declining in Europe, as intended family size has been quite stable around two children per woman, whereas completed cohort fertility is declining. Event history analysis techniques commonly used in fertility research are ill equipped to model and explain this gap, because they are unable to explicitly model the physiological constraints on fertility. Our novel microsimulation approach grounded in human physiology overcomes this problem. With our model we measure how much education and recent union trends contributed to the fertility gap of a recent cohort of Dutch women. To do this we simulate the entire individual reproductive life courses of Dutch women born between 1974 and 1984, who recently have or soon will complete their fertility. We use mainly GGS I and LISS panel data for the behavioural inputs in our model. The physiological inputs are based on extensive modelling work by demographers and clinical literature. We validate our model with cohort data from the Human Fertility Database. By comparing our simulation estimates with regression estimates using our simulated data, we investigate whether failing to explicitly model the reproductive process leads to incorrect inferences. Our study brings us closer to understanding the gap between intended family size and completed fertility, as well as measuring how much fertility is affected by its individual determinants.

Synthetic Instagram Post Generation for Social Media Research
Lily Heisig, Maastricht University – Faculty of Science and Engineering (UM-FSE); Sander Lardinois, Maastricht University – Faculty of Science and Engineering (UM-FSE); Thales Bertaglia, Maastricht University – Faculty of Science and Engineering (UM-FSE); Adriana Iamnitchi, Maastricht University – Faculty of Science and Engineering (UM-FSE)

Social media platforms, such as Instagram, are valuable data sources for research. However, these platforms often restrict data sharing, making research replication difficult. Additionally, privacy concerns and ethical considerations complicate data sharing further. Moreover, real data is difficult to collect for infrequent instances, such as undisclosed sponsored content. Synthetic data provides a solution to these challenges, but only if it accurately represents the real data. This research uses large language models to generate synthetic Instagram posts and proposes evaluation measures to assess synthetic data quality. We focus on Instagram captions representative for distinguishing sponsored from non-sponsored content. We use GPT-3.5 to generate captions and assign authorship to such posts. We explore different prompt engineering techniques and suggest best practices for synthetic data generation. Our proposed evaluation metrics include simple statistical analysis, topic modelling, text complexity measures, sentiment analysis and measures extracted from the emerging networks of co-occurring hashtags, tagged users, and authors. We find that GPT-3.5 generates very realistic individual Instagram captions that include appropriate emojis, URLs, hashtags and user tags. However, it stays very generic, thus failing to provide (without explicit prompting) the special cases in which we are particularly interested. More importantly, it fails to output realistic distributions across text properties (such as sentiment) and network properties (such as hashtag co-occurrence). We develop a set of dataset characterisation metrics and a methodology for curating the synthetic dataset to better fit the metrics of real datasets.

Implicit Association Tests: Stimuli Validation from Participant Responses
Sally Hogenboom, Open University (OU); Katrin Schulz, University of Amsterdam – Faculty of Humanities; Leendert van Maanen, Utrecht University – Faculty of Social Sciences (UU-FSW)

The Implicit Association Test (IAT, Greenwald et al., 1998) is a popular instrument for measuring attitudes and (stereotypical) biases. Greenwald et al. (2021) proposed a concrete method for validating IAT stimuli: appropriate stimuli should be familiar and easy to classify – translating to rapid (response times < 800 ms) and accurate (error < 10%) participant responses. We conducted three preregistered analyses to explore the theoretical and practical utility of these proposed validation criteria. We first applied the proposed validation criteria to the data of 15 IATs that were available via Project Implicit. A bootstrap approach with 10,000 ‘experiments’ of 100 participants showed that 5.85 % of stimuli were reliably valid: we are more than 95% confident that a stimulus will also be valid in a new sample of 18-25 year old US participants. Most stimuli (78.44%) could not be reliably validated, indicating a less than 5% certainty in the outcome of stimulus (in)validity for a new sample of participants. We then explored how stimulus validity differs across IATs. Results show that only some stimuli are consistently (in)valid. Most stimuli show between-IAT variances which indicate that stimulus validity differs across IAT contexts. In the final analysis we explored the effect of stimulus type (images; nouns; names; adjectives) on stimulus validity. Stimulus type was a significant predictor of stimulus validity. Although images attain the highest stimulus validity, raw data shows large differences within stimulus types. Together the results indicate a need for revised validation criteria. We finish with practical recommendations for stimulus selection and (post-hoc) stimulus validation.

The Amsterdam Social Network Interventions for Health Simulator (ASNIHS): Building the evidence base to improve the resiliency and health of Amsterdam residents
Jiri Kaan, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG); Kristina Thompson, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG); Annemarie Wagemakers, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG); Spencer Moore, Wageningen University & Research – Wageningen Social Science Group (WUR-SSG)

Recent advances in social network research have highlighted a variety of strategies for leveraging social networks to improve community-based interventions. Nonetheless, the role of social networks in interventions is frequently under-theorized or taken for granted. Furthermore, predicting which type of social network intervention would work best in which neighbourhood is sometimes difficult to test in practice. Therefore, in our study, we will develop an agent-based model to create an intervention planning tool called the Amsterdam Social Network Interventions for Health Simulator (ASNIHS). The ASNIHS will be developed in part in collaboration with community members and policy stakeholders, for example, through network mapping or interviews, to improve the development and adoption of social network strategies in Amsterdam health promotion initiatives. Other data would come from pre-existing data sources, such as ODISSEI and HELIUS. The ASNIHS will enable researchers and stakeholders to virtually test the impact of different social network intervention strategies in neighbourhoods on the resilience, health, and well-being of their residents; either with observed or hypothetical network characteristics. Simultaneously, the model will account for multiscale interactions and bidirectional feedback loop interactions, such as those originating from the communities’ geo-spatial environment. By making the agent-based model available for stakeholders via workshops, we hope to serve the community better and gain feedback that will help us improve the ASNIHS for future use. Understanding the efficacy of different social network interventions across neighbourhoods may contribute to improving population health while reducing health disparities.

Cardiometabolic diseases in autistic adults: research based on a nationwide population-based cohort study from the Netherlands
Yiran Li, University of Groningen – University Medical Center Groningen (RUG-UMCG); Tian Xie, University of Groningen – University Medical Center Groningen (RUG-UMCG); Lin Li, Örebro University – School of Medical Sciences; Harold Snieder, University of Groningen – University Medical Center Groningen (RUG-UMCG); Catharina Hartman, University of Groningen – University Medical Center Groningen (RUG-UMCG)

Previous studies suggest that Autism Spectrum Disorder (ASD) is associated with cardiometabolic diseases, but little is known about the full risk profile of cardiometabolic diseases in autistic men and women in different parts of the lifespan. We are currently conducting a longitudinal, population-based cohort study to investigate the prospective associations between ASD and cardiometabolic diseases including diabetes, hypertension, dyslipidemia, stroke, angina pectoris, myocardial infarction, and heart failure in adults with and without ASD. Additionally, we will explore if the associations differ by sex and age. Using nationwide data from the Central Bureau of Statistics (CBS), adults born between January 1, 1920, and December 31, 1996, are identified without pre-existing cardiometabolic diseases. The study period is from January 1, 2014 to December 31, 2020. Incident cardiometabolic disease events are identified according to diagnoses in CBS. Medication dispensation is additionally used to identify diabetes, hypertension, and dyslipidemia. ASD is identified according to the Diagnostic and Statistical Manual of Mental Diseases 4th edition (DSM-IV) diagnostic code in CBS. Hazard ratios (HR) with 95% confidence intervals (CI) will be calculated using Cox proportional hazards regression model, with ASD as a time-varying exposure. Sex, birth year, highest education level, income, obesity, tobacco use, and comorbid psychiatric disorders will be adjusted for in multiple steps of model fitting. Stratified analyses by sex and age bands (18-34, 35-64, and ≥65 years) will be conducted. The findings may inform the diagnoses, monitoring, and interventions to reduce cardiometabolic risk in autistic adults. On the 2nd of November, we will be able to present the results of this study during the ODISSEI conference.

Data linkage of health care registries to study silicone breast implants adverse events
Annemiek Lieffering, Netherlands Institute for Health Services Research (Nivel); Lotte Ramerman, Netherlands Institute for Health Services Research (Nivel); Juliëtte Hommes, Zuyderland Medical Center; Robert Verheij, Netherlands Institute for Health Services Research (Nivel); René van der Hulst, Maastricht University Medical Center; Hinne Rakhorst, Medical Spectrum Twente; Marc Mureau, Erasmus MC Cancer Institute

Objective: To assess the association of silicone breast implants (SBIs) with non-specific health symptoms and medical specialist care utilization. Design: Retrospective cohort studies (2013-2021) using data linkage with 1) the Dutch Breast Implant Registry, which includes almost all patients undergoing breast implant surgery, 2) Nivel Primary Care Database, containing routine electronic health records from over 500 general practices, 3) the Dutch health insurance claims database, including virtually all medical specialist care, and 4) Statistics Netherlands, providing socioeconomic indicators. Participants: Women with SBIs for cosmetic reasons and several control groups. Outcomes: The occurrence of thirteen non-specific health symptoms presented in general practice one year before implantation until three years after implantation. Visits to medical specialists were examined over a three-years period prior to explantation of SBIs. Results: Women with SBIs had an increased odds of non-specific health symptoms after implantation compared to before implantation (OR second year after implantation 2.13, 95% CI 1.22 to 3.72), as well as compared to women without SBIs (adjusted OR 1.44, 95% CI 1.04 to 1.98). Women who underwent explantation of SBIs were more likely to have visited >5 different medical specialties in comparison with control groups of women who underwent SBI replacement surgery (16.1% vs. 10.5%; p<0.001) and women without SBIs (16.1% vs 3.7%; p<0.001). Among explantation patients, women who underwent explantation because of non-specific health symptoms were more likely to have visited >5 different medical specialties compared to women who underwent explantation because of other reasons (31.6% vs. 14.6%; p<0.001). Conclusions: The research findings suggest an association between women who have SBIs for cosmetic reasons and non-specific health symptoms, that may be related to a high medical specialist care utilization. Data linkage of health care data has been demonstrated to be useful in epidemiological studies.

Evidence for severe mood instability in patients with bipolar disorder: Applying multilevel hidden Markov modelling to intensive longitudinal ecological momentary assessment data
Sebastian Mildiner Moraga, Utrecht University – Faculty of Social Sciences (UU-FSW); Emmeke Aarts, Utrecht University – Faculty of Social Sciences (UU-FSW)

Bipolar disorder (BD) is a chronic psychiatric condition characterized by large shifts in mood, energy, and cognitive functioning. Recently, the conceptualization of BD has shifted from alternating discrete episodes to a model of chronic cyclical mood instability. Recognizing and quantifying this mood instability may improve care and calls for high-frequency measures coupled with advanced statistical models. To uncover empirically derived mood states, a multilevel hidden Markov model (HMM) was applied to 4-month ecological momentary assessment (EMA) data in twenty patients with BD. EMA data comprised self-reported questionnaires (5 per day) measuring manic and depressive constructs using 12 items. Manic and depressive symptoms were further assessed by weekly administered self-reported questionnaires (i.e., Altman Self-Rating Mania Scale and Quick Inventory for Depressive Symptomatology Self-Report). Alignment between uncovered mood states and weekly questionnaires was assessed with a multilevel linear model. HMM uncovered four mood states (euthymic, manic, mixed, and depressive) that aligned with weekly symptom scores. On average, the duration of the states was <24h: patients remained in the mixed state the shortest (8.8h), followed by the manic (9.1h), depressive (16.3h), and euthymic (17.1h) states. States switched more frequently than weekly data suggested. In almost half of the patients, significant mood instability was observed. Large individual differences were observed in state duration and switching. The results suggest that mood instability could be a key feature of BD, which should be considered in theoretical and clinical conceptualizations of the disorder. Quantifying mood instability has the potential to improve the care of patients with bipolar disorder on a very individual scale.

The (Great) Persuasion Divide? Gender Disparities in Debate Speeches and Evaluations
Huyen Nguyen, Utrecht University – Faculty of Social Sciences (UU-FSW)

Do men and women persuade differently? Are they evaluated differently? This research investigates spoken verbal tactics across genders and their impacts on performance evaluations using a novel data set of 1517 speech transcripts, evaluation scores, and demographic data from the highest-profile inter-varsity debate tournaments between 2008 and 2018. Female debaters use a more personal and disclosing speaking style, with more hedging phrases and non-fluencies in their speeches. In their answers to opponents’ questions, women negate less, while having longer and more vague answers. In terms of evaluation, across debates, having a less analytical speaking style and more positive sentiment is associated with higher scores for speeches by women, but not by men. Within debates, except for non-fluencies, there is no robust evidence of gender-specific evaluation standards. These findings suggest that the gender score gap in top-tier debate tournaments is due to gender differences in persuasion tactics, rather than discrimination.

Data visualization for incomplete data
Hanne Oberman, Utrecht University – Faculty of Social Sciences (UU-FSW)

Missing data are ubiquitous in the social and human data sciences. Computational pipelines that accommodate incomplete data require exploration and evaluation of the missingness. The R package [ggmice](amices.org/ggmice) enables data analysts to visualize incomplete and imputed data.

Modeling Interactions in Product Reviews: Utilizing Formal Argumentation to Assess Strengths and Weaknesses of Products Across Different Aspects
Ji Qi, Netherlands eScience Center (NLeSC); Atefeh Keshavarzi Zafarghandi, Centrum Wiskunde & Informatica (CWI), VU Amsterdam – Faculty of Science (VU-Science); Davide Ceolin, Centrum Wiskunde & Informatica (CWI)

Product reviews serve as critical sources of information for potential buyers and valuable feedback for reviewed entities. Recognizing their pivotal role, companies increasingly appreciate the efficacy of helpful reviews as a marketing tool. Extensive interdisciplinary research, ranging from philosophy to artificial intelligence, has investigated factors influencing the prediction of helpful reviews. This study focuses on the creation of a valuable resource for new users by effectively analyzing product reviews and highlighting the strengths and weaknesses associated with different aspects of a product. To achieve this, we employ a multi-step approach. Firstly, we adopt a review segregation technique, dividing each review into discrete chunks, typically clauses, wherein each chunk represents a specific aspect of the product. Next, we apply a modified version of the Textrank algorithm to assign a numerical rank to each chunk, reflecting its relative importance within the overall review. We further employ topic modeling using BERT-based transformers to group chunks into aspects. By merging the chunks back into reviews, we ascertain the aspects covered in each review. Subsequently, we analyze the attacking relationships between reviews pertaining to specific aspects of the product. Leveraging argument mining approaches, we identify more reliable reviews based on the attack network of reviews, enabling us to discern the strengths and weaknesses of the product across different aspects through these reliable reviews. To evaluate the proposed approach, we have developed a series of GUI software tools within the scientific workflow platform, Orange3, and applied them to the Amazon Product Review dataset. The implementation in Orange3 offers intuitive graphical interfaces, tunable components, and visualization tools, enhancing user understanding of the underlying mechanisms and the significance of the output. The proposed framework contributes to enhancing the understanding of product interactions in reviews, aiding both consumers and businesses in making informed decisions.

Big data governance on digital twin technology for smart and sustainable tourism
Eko Rahmadian, University of Groningen, Campus Fryslan

We are witnessing a massive increase in the use of digital technologies. As one of the emerging concepts in artificial intelligence, machine learning, and the Internet of Things (IoT), Digital Twin (DT) technology can predict system responses before they occur. DT can be described as a virtual representation that defines the comprehensive physical and functional characteristics of the life cycle of a product. DT applications have been implemented in many sectors, including smart cities. Considering the rapid growth of new ICT applications in the tourism industry and digitization through IoT, we suggest that DT has the potential to be implemented in smart and sustainable tourism. By utilizing big data and other supporting resources, stakeholders will be able to create a virtual representation of a relevant region both by analyzing the flow of visitor activity and by determining the impact of their geographic and temporal patterns on other aspects and policies as a leverage on the use of big data for statistical products. However, we are also aware that compliance with regulations and communication among stakeholders such as data scientists, data analysts, statisticians, managers, and business analysts have become important issues. There are also numerous concerns about security, privacy, and trust. Therefore, our paper proposes a conceptual framework for DT on smart and sustainable tourism and a documentation framework for architectural decisions as a recommended way of governing such systems. This documentation framework provides benefits that shape how each stakeholder communicates and interacts using the system while adhering to rules and regulations to ensure trustworthiness, accountability, and transparency. With a theoretical case study and three case scenarios on the use of mobile positioning data in Indonesia as examples, we intend to demonstrate the applicability of our work.,

Consequences of delayed care in chronic disease management programs in Dutch patients with diabetes during the COVID-19 pandemic: a registry-based observational study
Corinne Rijpkema, Netherlands Institute for Health Services Research (Nivel); Lotte Ramerman, Netherlands Institute for Health Services Research (Nivel); Isabelle Bos, Netherlands Institute for Health Services Research (Nivel); Robert Verheij, Netherlands Institute for Health Services Research (Nivel)

Background: During the COVID-19 pandemic, chronic disease management programs (CDMP) at the general practitioner (GP) for Dutch patients with chronic conditions, such as diabetes, were delayed/postponed. This may have resulted in patients requiring unplanned care in other healthcare settings due to the development of acute complaints. Therefore, we investigated the changes in contact rates for patients with diabetes in 2020-2021 compared to 2019 regarding 1) CDMP care, 2) additional care in general practices, 3) care at out-of-hours GP services and 4) hospital care. Method: We linked GP data from the Nivel Primary Care Database with hospital data from Statistics Netherlands (CBS), to analyze proportional differences in contact rates across all four types of care, shown for each quarter of 2020-2021 compared to 2019. Results: Preliminary results indicated that for patients with diabetes, the care from CDMP decreased in 2020 and 2021 compared to 2019. In Q1-Q3 of 2020 and Q3 of 2021, GPs were contacted less often for additional diabetes care in general practices, but in the other quarters, there was increased GP care for diabetes compared to the same periods in 2019. During the pandemic, patients with diabetes visited out-of-hours GP services more often for their diabetes, except for Q3 of 2020 and 2021. These patients also had increased hospital care for diabetes in 2020, except for a brief decrease in Q2 compared to 2019. Conclusion: Delayed/postponed CDMP care for patients with diabetes may have led to increased diabetes-related care at out-of-hours GP services and hospitals in most quarters of 2020 and 2021 compared to 2019. It is crucial to maintain monitoring due to potential long-term consequences of COVID-19 beyond 2021.

How do regions adopt new technologies? AI adoption in the Netherlands
Harm-Jan Rouwendal, University of Groningen, Faculty of Spatial Sciences (RUG-FRW); Sierdjan Koster, University of Groningen – Faculty of Spacial Science (RUG-FRW); Tersa Farinha, UNU-MERIT

How do regions adopt new digital technologies: by following general trends or first movers? This paper studies AI adoption in regional labour markets. We use the demand for AI-related skills in online job vacancies for the period 2010-2020 for the Netherlands as a measure of AI adoption in firms. We document a rapid increase of AI related vacancies in the past decade, especially in the professional and technical occupations and in urban areas. Moreover, we find spatial differences in AI adoption that can only be partly explained by sector structure and the effect of agglomeration economies. We hypothesize that firms are more prone to adopt AI if similar firms in same regions already use AI. This would indicate that local knowledge spillovers remain important in the regional adoption of new digital technologies and the subsequent automation process.

Who is misinformed, who is a conspiracist? Investigation of psychological factors associated with susceptibility to misinformation and conspiracy theories with a network analysis approach
Selin Topel, Leiden University – Faculty of Social Sciences (UL-FSW); Ili Ma, Leiden University – Faculty of Social Sciences (UL-FSW); Ellen de Bruijn, Leiden University – Faculty of Social Sciences (UL-FSW)

In today’s digital era, global events like the COVID-19 pandemic, the war in Ukraine, and climate change have the potential to trigger swift dissemination of misinformation through social media platforms. During times of heightened uncertainty, individuals may also resort to conspiracy thinking as a maladaptive way to cope with anxiety and uncertainty. Unfortunately, this widespread misinformation and endorsement of conspiracy theories can have severe negative consequences on both personal and societal levels, including increased polarization, mental health problems, and rising aggression. Despite the evident dangers of misinformation spread and conspiracy thinking, our understanding of specific factors that either escalate or mitigate susceptibility to these tendencies remains limited. Past research has investigated various psychological factors, such as Intolerance of Uncertainty, information processing biases, and personality traits, but often in isolation or with a narrow focus. The multifaceted nature of these tendencies demands an interdisciplinary approach. In this study, we aimed to uncover individual and psychological factors influencing susceptibility to misinformation and conspiracy thinking. We collected data from a sample of 214 healthy adults (ages 18-35) residing in the UK through an online study. Participants completed self-report questionnaires and an information sampling task. By constructing a network based on the various questionnaire constructs and computational model parameters, we gain insights into the interactions between psychological factors linked to misinformation and conspiracy thinking. This integrative approach provides a more comprehensive understanding of the complex relationships among these factors. Our study contributes to the advancement of knowledge in this critical area of research and can potentially inform strategies to address susceptibility to misinformation and conspiracy thinking. Understanding the underlying factors in the context of their interplay is essential for developing targeted interventions to promote a better informed and resilient society.

Better Together: Exploring the complex dynamics of classroom interaction through social network analysis
Nina van Graafeiland, Utrecht University – Faculty of Social Sciences (UU-FSW); Mahdi Shafiee Kamalabad, Assistant Professor; Nienke Smit, Utrecht University – Faculty of Social Sciences (UU-FSW)

The quality of classroom interaction is an important factor in the effectiveness of language lessons, but managing and sustaining interaction in large, diverse classrooms is a complex and dynamic process. This means that the process is constantly changing and is impacted by multiple internal and external components. A computational method that can possibly be used to investigate these complex dynamic systems is social network analysis. The application of social network models to classroom data allows us to quantify how students and teachers interact in real-time and unravel the drivers behind this behavior. The first study in this project compared the assumptions and principles of complex dynamic systems theory as a metatheory and social network analysis as a toolkit and investigated how a specific social network model (i.e., the Relational Event Model (REM)) can be applied to classroom data in order to better explore this complex dynamic system. In this study, we generated synthetic network data to test the use of this model in an educational context, in preparation of real-life data-analysis that will be conducted in the following studies. The REM can provide accurate estimations of how specific drivers (e.g., timing, inertia, reciprocity, and gender) might shape which, and when interactions occur. In turn, it can inform us on the evolution of the system over time.

Rising through the Ranks: Firms and Intergenerational Mobility
Oskar Veerhoek, Radboud University Nijmegen – Faculty of Management Sciences (RU-FdM)

Across the developed world income inequality is on the rise whereas intergenerational mobility is in decline. Current generations will likely experience less equality of opportunity than any generation since the Second World War. Many studies of intergenerational mobility focus on the transmission of human capital from parents to children, but few consider the role of firms. This project is made possible by the ODISSEI Microdata Access Grant, which funds access to CBS microdata. With recently published CBS microdata files, this project sets out to study the effect of firms on intergenerational mobility in the Netherlands. A conceptual model of firm impact on mobility is developed based on insights from sociology and economics. Firms are expected to affect intergenerational mobility through a combination of firm pay premia and socioeconomic inclusiveness of hiring and promotion. The results of this study will add to the burgeoning interdisciplinary literature on intergenerational mobility. They will also have policy implications for both firms and governments.

COMMA: An agent-based micro-simulation model to study mental health outcomes during COVID-19 lockdowns
Eva Viviani, Netherlands eScience Center

The COVID-19 pandemic has led to an increase in known risk factors for mental health problems. This evidence has underscored an urgent need for models that can project mental health outcomes over time, explore lockdown scenarios, and target the needs of specific populations. Here we describe COMMA (COvid Mental-health Model with Agents), a new open-source microsimulation model developed to help address these questions. COMMA takes as input demographic information on age structures, population size, etc; and lockdown policies operationalised as a set of action probabilities. The lockdown policies affect the likelihood of individuals developing mental health issues, notably depression, based on demographic profile. Implemented purely in Python, COMMA has been designed with equal emphasis on performance, ease of use, and flexibility. Users can customise lockdown scenarios and population characteristics and execute simulations on a standard laptop within minutes. In a collaboration between the Netherlands eScience center and Wageningen university, COMMA has already been employed to assess the impact of various lockdown strategies on the mental well-being of a population resembling the demographic profiles of the inhabitants of the Groningen area. 

Setting up a Governance Framework for Secondary Use of Routine Health Data in Nursing Homes: Development Study Using Qualitative Interviews
Yvonne Wieland-Jorna, Netherlands Institute for Health Services Research (Nivel); Robert A Verheij, Netherlands Institute for Health Services Research (Nivel); Anneke L Francke, Netherlands Institute for Health Services Research (Nivel); Marit Tomassen, Netherlands Institute for Health Services Research (Nivel); Max Houtzager; Karlijn J Joling; Mariska G Oosterveld-Vlug

Background: In the nursing home sector, reusing routinely recorded EHR data for knowledge development and quality improvement is in its infancy. A data governance framework (who may access the data, under what conditions, and for what purposes) can help obtaining trust in appropriate and responsible data reuse. Little guidance is available on development and implementation of data governance frameworks in practice. Objective: To describe the development process of a governance framework for the “Registry Learning from Data in Nursing Homes” – a Dutch national registry for EHR data on care delivered by nursing home physicians. Methods: Interviews were conducted with stakeholders representing practices, policies, and research in the nursing home sector. Main aim was to explore perspectives regarding the Registry’s aim, data access criteria, and governing bodies’ tasks and composition. Interview topics and analyses were guided by 8 health data governance principles. Interview results, together with legal advice and consensus discussions were used to shape the rules, regulations, and governing bodies of the framework. Results: Stakeholders saw the involvement of nursing home residents, nursing home physicians, boards of directors, and scientists as a prerequisite for a trustworthy data governance framework. For the Registry, their involvement can be achieved through a consent procedure, transparency, and a position in a governing body. In addition, a data request approval procedure indicates that data reuse by third parties aligns with the aims of the Registry, benefits the nursing home sector, and protects the privacy of data subjects. Conclusions: Stakeholders’ views, expertise, and knowledge of other frameworks and relevant legislation serve to inform the application of governance principles to the contexts of both the nursing home sector and the Netherlands. Engagement of the full range of stakeholders in an early stage of governance framework development is important to generate trust in appropriate and responsible data reuse.

Maximizing Panel Research Project Management with Cutting-Edge Panel Management Software
Arnaud Wijnant, Centerdata

Effective panel research requires the seamless integration of various components, from data collection methods such as CAWI (Computer-Assisted Web Interviewing) and PAPI (Paper and Pencil Interviewing) to robust software solutions, efficient helpdesk support, incentivization strategies, stringent security measures, and comprehensive monitoring of privacy. This paper explores the critical role of panel management software in facilitating successful panel research projects while ensuring optimal data quality and participant satisfaction based on the experiences we have with projects like the LISS panel project. Panel management software serves as the cornerstone of panel research, offering invaluable features to researchers. It enables the creation and management of online surveys using CAWI, allowing for efficient data collection and analysis. Additionally, it facilitates the seamless transition to PAPI when needed, ensuring flexibility in diverse research scenarios. To ensure smooth operations, a helpdesk support system is part of panel management. Assistance plays a vital role in resolving technical issues, addressing participant queries, and ensuring uninterrupted research activities. Incentives are integral to maintaining panelist engagement and motivation. The software enables the implementation of incentives. These incentives not only enhance participant retention but also encourage active participation, yielding reliable and high-quality data. Security and privacy are of paramount importance in panel research. Robust software solutions employ stringent security measures to safeguard participant information, ensuring compliance with data protection regulations. Monitoring plays an important role in maintaining data quality and panel integrity. Effective panel management software offers comprehensive monitoring capabilities, allowing researchers to track participant activity, detect anomalies, and identify potential sources of bias or fraudulent behavior. This monitoring ensures the reliability and credibility of research outcomes. In conclusion, panel management software forms the backbone of successful panel research projects. By leveraging these essential components, researchers can conduct rigorous and reliable studies while upholding participant privacy and data integrity.

The Contagion of Collective Action: Analyzing Spatial Diffusion of Citizen Care Collectives in The Netherlands
Kevin Wittenberg, Utrecht University – Faculty of Social Sciences (UU-FSW); Rense Corten, Utrecht University – Faculty of Social Sciences (UU-FSW)

In recent years, The Netherlands has seen a resurgence of collective action by citizens to take charge and organize the provision of (in)formal care services among themselves. To date, efforts to understand which communities are able to mobilize themselves for such collective action have been scarce and yield low explanatory power. Existing research predominantly takes the perspective that the emergence of collectives can be explained based on static characteristics of residents and their wider community. In this paper, we expand beyond this perspective, and argue that the emergence of care collectives may also be explained endogenously, which means that citizen collectives in one area can shape the conditions for collectives to emerge in its vicinity. We argue that this may be caused by direct contact between citizens and collectives, but also indirectly through legitimization processes and improved understanding between citizens and local government. We take a two-step approach to test our hypotheses. First, we use a data-driven approach to predict the emergence of citizen collective for care based on administrative data. We do so to adjust for potential confounding causes of spatial clustering. We then investigate the spatial correlation of the residuals of this model to test for diffusion of collective action. We differentiate between within-municipality and between-municipality diffusion. Preliminary results indicate that citizen collectives are indeed spatially clustered, even when accounting for demographic, political and geographic characteristics, which supports the notion that citizen collectives shape conditions for new collectives to emerge in their surroundings. The explainability of collectives based on administrative data at the zipcode level is low. We are still investigating the stability of the results under various model specifications, exploring the addition of more data in the predictive model, and alternative measurement of what constitutes a neighboring region.

How Parental Unemployment Influences Children‘s Educational Attainment and Income: The role of Parental Economic Resources and Parental Mental Health
Flora Zhou, Erasmus University Rotterdam, School of Social and Behavioural Sciences (EUR-ESSB)

Although the negative impact of parental unemployment on children’s outcomes has been investigated in previous work, the mechanisms underlying the intergenerational effect still remain unclear and the heterogeneous effect of parental unemployment in families with different income levels is not fully addressed yet. This research examined the effect of parental unemployment on children in the Netherlands, focusing on children’s highest educational attainment and early-career income. We explored whether the intergenerational effect of unemployment is explained by two essential mechanisms, namely parental economic resources and parental mental health. Furthermore, we investigated how the buffering effect of family incomes works in the process of the intergenerational effect of parental unemployment. We linked data from the Longitudinal Internet Studies for the Social Science (LISS) and the Social Statistical Data sets from Statistics Netherlands (SSD), deriving children’s information from the SSD and their parents’ information from the LISS panel. In the end, we have 1283 children who were born between 1989-1996 in our sample. Given that we aim to estimate the effect of parental unemployment on children’s educational attainment and income when they were between 23 and 30 years old, we only measured parental information before children were at their age of 23 to address the potential causal issues. The Structural Equation Modeling (SEM) results indicate that parental unemployment does not influence children’s educational attainment and early-career income after controlling for parental economic resources and parental mental health before the incidence of unemployment. Despite the lack of evidence on the intergenerational effect of unemployment, we found that unemployment significantly decreases paternal income and maternal mental health. Our results also suggested that maternal mental health after unemployment positively influences children’s educational attainment.

The study of the social dimensions of the internet is at an acute crisis point. There has been a general rolling back of data access to studying what happens online, most dramatically recently with the elimination of API-based access to Twitter (now X). This talk will outline various paradigms for empirical research of studying online behaviors of people and platforms, as the field transitions from its reliance on Twitter data to an uncertain future, with a particular focus on the potential for creating an infrastructure utilizing user-sourced data on human and platform behavior.

Neighborhoods embeddings using Street-Level Images and Human similarity perceptions
Francisco Garrido-Valenzuela, TU Delft – Faculty of Technology, Policy and Management (TUD-TPM); Oded Cats, TU Delft – Faculty of Technology, Policy and Management (TUD-TPM); Sander van Cranenburgh, TU Delft – Faculty of Technology, Policy and Management (TUD-TPM)
In the era of rapid urbanization and complex urban challenges, such as social inequalities and resilience, gaining a deeper understanding of urban spaces is crucial for urban planners and policymakers to maintain or improve citizens’ well-being standards. Having an accurate and nuanced understanding of the wide variety of urban spaces facilitates effective neighborhood planning, prioritizing interventions, and resource allocation. However, traditional methods to characterize neighborhoods often fail to capture the multidimensional nature and intricate relationships defining urban spaces. In particular, they do not incorporate human feeling and perceptions into their characterizations, which provide valuable insights into people’s experiences and interactions with urban environments. In this way, human perceptions support the design of inclusive, people-centered urban interventions, enhances livability, and promotes sustainable development. To address this gap, we propose a method to create urban embeddings which incorporates urban human-perceptions. Our method involves three steps. First, the city is divided into small spatial units, each representing a neighborhood with a collection of street-level images (SLI). Second, we conduct a survey using SLI in which participants must compare triplets of neighborhoods (five SLI as a neighborhood) and choose the odd-one-out. Images are drawn from millions geotagged-SLI database we have collected from across the Netherlands. Third, we generate urban vectors using a deep-embedding model. The resulting embeddings do not only encompass observable neighborhood features like building types, street furniture, and transportation infrastructure but also encompass (physically unobservable) human perceptions of neighborhoods. We (preliminarily) find context-aware representations that encompass the spatial, visual, and socio-economic complexities of neighborhoods. It is able to differentiate various sorts of residential, commercial, and rural neighborhoods. By providing insights into urban dynamics, uncovering hidden patterns, we believe our method can support urban planners and city policymakers with evidence-based decision-making to effectively address the complex challenges that cities face today.

Housing crisis and mental health in the Northern Netherlands
Agata Troost, University of Groningen – Faculty of Spatial Sciences (RUG-FRW)
There has aready been research on spatially concetrated deprivation affecting individual mental health (see eg. Visser et al, 2021, for an overview), yet little is known about the specific influence of the lack of access to affordable housing in Northern Netherlands. This research combines the large-scale administrative datasets of the Statistics Netherlands (CBS) Microdata with detailed questionnaires and measurements of mental health and well-being from the longitudinal Lifelines study (https://www.lifelines.nl/). We start with investigating which areas of Netherlands are most affected by housing crisis, or in other words, where in the Netherlands do we encounter the biggest scarcity of affordable housing. We employ various measurements of that scarcity and in doing so consider different predictors of housing (in)affordability, at different spatial scales. For creating the “bespoke neighbourhoods”, or areas of investigation other than the administrative programme units, we use GIS and other programmes such as Equipop (https://www.equipop.kultgeog.uu.se/?languageId=1).
Once we identify areas struggling with the biggest scarcity of affordable housing, considering the needs of the local population, we study the influence of housing scarcity and housing insecurity on mental health, taking into account financial insecurity. In our analysis we use longitudinal models which, in addition to modelling the influence of spatial context (competitiveness of the housing market), control for the household context and individual health history. We also theoretically explore why people settle in competitive housing market areas (eg. because of educational or work-related reasons).

Networks in the market for researchers
Flavio Hafner, Netherlands eScience Center (NLeSC); Christoph Hedtrich, Uppsala University
We study the role of networks in the labor market for young scientists in the United States. Nearly one in five PhD graduates that publish after PhD graduation do so at a university where their advisor has a former co-author; graduates that have such a connection are twice as likely to match with the university, even within fine- grained peer groups. We document a citation premium of 10 to 30% for graduates placed through the advisor’s network. Further analysis indicates that network placements are associated with private benefits on both sides of the market, with no evidence for aggregate productivity effects.

The Sound of Disgust: Differentiating Disgust Vocalisations through Computational Approaches
Roza Kamiloglu, VU Amsterdam – Faculty of Behavioural and Movement Sciences (VU-FGB); Christiaan Meijer, Netherlands eScience Center (NLeSC); Disa Sauter, University of Amsterdam – Faculty of Social and Behavioural Sciences (UvA-FMG); Joshua Tybur, VU Amsterdam – Faculty of Behavioural and Movement Sciences (VU-FGB)
Ew! Yuck! Ugh! These are vocalisations that often accompany disgust, which serves the crucial evolutionary function of motivating avoidance of the pathogens that cause infectious diseases (pathogen disgust). It can be elicited by a variety of sensory stimuli, especially those associated with smells and tastes that could indicate the presence of pathogens. For instance, we might experience pathogen disgust when we smell spoiled food or walk into a dirty public restroom. We might, however, also experience other forms of disgust, such as repugnance of violations of social norms or moral standards. Examples include watching a news report about a company exploiting its workers or witnessing dishonesty like lying or cheating might elicit disgust (moral disgust). Here, we investigate whether the vocalisations accompanying pathogen and moral disgust are acoustically distinguishable, just as their functions. We employed computational methods that allow us to capture pathogen and moral disgust vocalisations using rich audio samples from daily life. The acoustic dataset consisted of 75,504 data points: 88 acoustic features extracted from each of 858 disgust vocalisations compiled from YouTube. Six machine classifiers with 5-Fold cross-validation tested whether the two types of vocalisations can be classified based on their acoustic structure. The AUC (Area Under the Curve) with a value of 0.68, a measure of how well logistic regression model performed, indicated better-than-random classification (0.50) performance. Moreover, a pre-registered perception study demonstrated that naïve listeners (n = 200) are able to infer at better-than-random guessing levels whether the vocalising person was experiencing pathogen or moral disgust. Our approach here shows that pathogen and moral disgust vocalisations are acoustically distinct, and listeners are sensitive to these differences. We demonstrate that applying the tools of machine learning to rich body of audio data can reveal systematicity in complex and high-dimensional behavioural domains like human vocal expressions.